Wikipedia talk:Edit filter/Archive 5

Archive 1

←

Archive 3

Archive 4

Archive 5

Archive 6

Archive 7

Archive 8

How...

was this not picked up by a filter? Surely that's a pretty obvious thing for a vandal to do? HJ Mitchell | fancy a chat? 02:19, 23 January 2010 (UTC)

It might have been one of the 0.47% of edits that trip too many conditions and thus end up being allowed by default. It "should" have tripped at least #50, though perhaps not #9, since the way he wrote out "whore" would make it not seem to be a separate word. -- Soap ^Talk/_{Contributions} 02:25, 23 January 2010 (UTC)

Due to the way the filters are targeted, vandalism is the most likely thing to hit the condition limit. So I suspect Soap is right. Prodego ^talk 04:30, 23 January 2010 (UTC)

My debugger states that this does not match filter 50. It certainly does not match filter 9. Traceback for 50 follows:

Extended content
ok ok backtrack * *backtrack *' *'backtrack *'' *'backtrack *''T *''Tbacktrack backtrack Match attempt failed

The problem appears to be the anchors. Not the entire line was in caps (it started with some original content, the added content was in the middle of the line), so it failed. This raises the question: can the anchors be removed? Off the top of my head (and at 6 AM) I'm a bit worried of false positives - 50 was apparently designed for the entire line to be in capital letters, not just significant portions. --Shirik (Questions or Comments?) 11:08, 23 January 2010 (UTC)

New/modify filter

Could someone please take a look at this vandals contributions, [1], and add whatever it was he did to a filter if possible, may be a new grawp game. Thanks--Jac16888^Talk 18:14, 24 January 2010 (UTC)

I'd have to see more examples. Prodego ^talk 18:16, 24 January 2010 (UTC)

All I've got for now, I'll keep my eyes open--Jac16888^Talk 18:17, 24 January 2010 (UTC)

I think this is the person who vandalized Commons a few days ago. The edits he made over there were such that nobody was able to view the Recent Changes page, because apparently the pages he vandalized were transcluded onto the Recent CHanges page. I don't know if what he did here also prevented RC from being viewed, but I would say I'm surprised that pages transcluded onto a page as important as that aren't full-protected. There is an abuse filter, #139, that protects against something very similar to this, and might have helped us if not for the fact that Tribble/Fleek's edits used a different way of accomplushing the same thing. still, I think we can use this. -- Soap ^Talk/_{Contributions} 18:35, 24 January 2010 (UTC)

Actually, it looks like Template space has been specifically excluded from that particular filter due to a legitimate template that occurred in the past. Though, going through all 2157 hits the filter has produced since it was created I wasnt able to find that particular false positive. I won't do anything yet, but it looks to me like this could be solved either by full-protecting the pages that are transcluded onto Special:RecentChanges, or by adding "position:absolute" to the edit filter, or both (since this CSS trick has been used in many other places besides this one.) -- Soap ^Talk/_{Contributions} 19:02, 24 January 2010 (UTC)

I went ahead and added "absolute" since the testing place seemed to indicate it would have stopped Tribble. If there is some place on the wiki that absolutely needs position:absolute and is frequently edited by non-admins, though, it will need to be added as an exception (though Im not aware of any such page). Note that the false positives early on in the logs from people editing their own CSS pages has been fixed. -- Soap ^Talk/_{Contributions} 19:10, 24 January 2010 (UTC)

As far as I can tell, MediaWiki:Recentchangestext is the only page on recent changes, and all templates on that are fully protected--Jac16888^Talk 19:08, 24 January 2010 (UTC)

Okay. Well, I do know that the edits made by User:Fleek on Commons somehow made the Recent Changes page unreadable. I wasn't here to see what this vandal's edits accomplished over here, but it looks like the intent, at least, was similar. -- Soap ^Talk/_{Contributions} 19:10, 24 January 2010 (UTC)

Just enable cascading protection on the Recentchangestext page and everything will be fine. Just another WoW/Grp wannabe. Nothing special. Triplestop x3 20:52, 24 January 2010 (UTC)

Commons has commons:Template:InterProject transcluded everywhere, so it was obviously vandalised. We can probably also blacklist some words -- zzuuzz ^(talk) 22:04, 24 January 2010 (UTC)

Suggestion for new filter.

I have noticed many unconstructive edits and articles made on Christian Weston Chandler, who is a subject of debatable notability. A filter made to block edits adding his name or articles contructed with his name as a title would be useful until the issue is settled one way or another.66.68.18.72 (talk) 03:15, 25 January 2010 (UTC)

There is already Filter 257, which logs edits containing his name, but it's weak because there are other people with similar names. (Note, for example, that we have an article for a person named Chris Chandler.) As for article titles, there is a blacklist for that too which is more efficient than an edit filter, but it seems to be rarely used, perhaps in part because there's no log produced by an attempt to create an article that hits a blacklist (as far as I know), so it would be very difficult to find out if a particular entry in the list was causing problems. -- Soap ^Talk/_{Contributions} 03:24, 25 January 2010 (UTC)

Careful what you say about private filters Soap. Prodego ^talk 05:07, 25 January 2010 (UTC)

On the subject of contains_any

Here's a fun discussion: contains_any(). Based on a recent false positive report regarding filter 12, I set out on a quest to do a rewrite of the logic behind the filter (while keeping the same conditions). After testing it on the test wiki, I put it in log-only here and it seemed to be working well and restored it to disallow. Now the fun part: It's more complicated and shouldn't have those false positives anymore, but also went from an average of 3.6ms down to what's currently less than 1ms (though I saw it as high as 2ms earlier). This performance increase is interesting, so I just thought I'd throw out the idea that, at least for future work, if you're going to have a call to contains_any with a lot of options, you might instead consider a regular expression that does the same thing; it may be more efficient. --Shirik (Questions or Comments?) 04:29, 28 January 2010 (UTC)

Edit summary repetition

Since filter 292 was created, the filter condition hit rate shot up from about 0.6% to nearly 2%. In light of this, the similarities between 292 and 135 led me to save a few conditions by merging it there, but I'd raise the question of whether it is needed at all. The majority of cases where edit summaries are repetitive have been something like "Replaced content with..." which would have tripped 135 anyway. Is there something else being targeted here? (On that note, regardless of this, this filter is actually just showing us an underlying problem, it is not the problem itself: Of those messages that are getting through, a lot of them seem to be extremely close to the 1000 condition limit, so I think an overall optimization pass might be appropriate. I'll probably take an whack at that tonight.) --Shirik (Questions or Comments?) 17:53, 12 February 2010 (UTC)

• 292 was based on some recent pagemove vandalism that was trying to evade filter 250. I appreciate your merging it into 135 - I would have done that in the first place but I'm awfully leery of the | operator. NawlinWiki (talk) 19:27, 12 February 2010 (UTC)

If you're "leery" of a basic logical operator, you shouldn't be modifying abuse filters. Gurch (talk) 20:17, 12 February 2010 (UTC)

I'm not sure that's a fair statement to say. Certainly the wiki benefits from as many people modifying the filters as possible, especially in urgent situations, so long as those people don't disrupt the wiki (i.e., cause false positives). I'm not opposed to a temporary reduction in performance which benefits the wiki and can then be later optimized by someone a little more experienced. NW was astute enough to make the appropriate filter while avoiding what he was unsure of, and then I took the steps later to maintain that functionality while improving performance. So long as users aren't manipulating things they aren't familiar with, this process benefits everyone. --Shirik (Questions or Comments?) 00:41, 13 February 2010 (UTC)

Filter 46 "Poop" vandalism

For Filter 46, is it really necessary to have "http://www.stationstops.com/2008/03/25/nyc-public-restrooms-the-straight-poop-on-public-restrooms-in-new-york-city-and-beyond/", "http://poopthebook.com/blog/2007/08/14/paradox-public-bathrooms/"? Can't we simply blacklist those...it seems like a total waste of resources to search for those two urls...and a misuse of the edit filter?Smallman12q (talk) 20:34, 11 February 2010 (UTC)

I'm not entirely sure where you're coming from. Are you saying that the performance hit isn't worth it, or are you saying that the URLs are unreliable sources and thus aren't worthy as an exception anyway? If it's the former, I would say not to worry about it. One of the most trivial things the filter does is check for string containment; the performance gains, even as multiplied as the filter is, would be minimal at best (and borders on WP:PERFORMANCE, as much as I disagree with it with regards to the edit filter). If it's the latter, I somewhat agree with you, though I was not the one to put those exceptions in and I don't know in what context they are present. --Shirik (Questions or Comments?) 20:45, 11 February 2010 (UTC)

Well I meant both from the technical performance side, and that the edit filters generally do not include urls as they belong on the blacklist.

Is it so difficult to move them to the blacklist...is it really necessary that the urls be in the edit filter?Smallman12q (talk) 20:55, 11 February 2010 (UTC)

I think you're misreading that filter. The URL is an exception to the filter (i.e., a whitelist). The reason it's there is because the word "poop" is in those URLs, but we want to permit those URLs to go through. Regards, Shirik (Questions or Comments?) 20:56, 11 February 2010 (UTC)

Oh...wow...erm...you're right. Guess I didn't read the "!". Well, I guess if the performance hit is nominal, then its okay...is there any reason why those two urls were singled out?Smallman12q (talk) 21:13, 11 February 2010 (UTC)

Without being the person that added them, I can't be sure, but I would assume they came from a WP:FALSEPOS report. (More info may be in the logs; I didn't look.) --Shirik (Questions or Comments?) 21:18, 11 February 2010 (UTC)

I did a linksearch and found that that url is only linked here and in Wikipedia talk:Edit filter/Archive 4, which means its not needed...The second one appears only in Public toilet and also Wikipedia talk:Edit filter/Archive 4. So from this, I can conclude that the link http://www.stationstops.com/2008/03/25/nyc-public-restrooms-the-straight-poop-on-public-restrooms-in-new-york-city-and-beyond/ can be removed, and as for the second link, its only used in public toilet...so I'm not sure if its worth keeping...Smallman12q (talk) 21:30, 11 February 2010 (UTC)

Perhaps, but that's probably something that should be dealt with at the article-level first. It would be inappropriate of me to use EFM rights to violate an established consensus (or, more appropriately, status quo in this case). --Shirik (Questions or Comments?) 22:00, 11 February 2010 (UTC)

So you won't even remove the link that has no articles linking to it=(.Smallman12q (talk) 22:10, 11 February 2010 (UTC)

I have no objection to removing it if no articles link to it. Is that the case? I didn't bother to check. --Shirik (Questions or Comments?) 22:11, 11 February 2010 (UTC)

You didn't check=[...for this linksearch the only links are to edit filter discussions and none to articles. Also, for links, shouldn't we be using "old_links"?Smallman12q (talk) 23:07, 11 February 2010 (UTC)

Sorry for being unclear. What I meant to say is that for the second one, I didn't check that it wasn't still in public toilet, as it was initially reported. I double-checked that, and it still is, so I'm going to leave that in until someone with more expertise on that subject (with no offense to the authors, can someone be an expert on public toilets?) determines whether or not that link fits in. As far as old_links, there's two problems. Firstly, that would require an additional condition which we are already strapped for. Secondly, that would prevent the link being readded after vandalism such as page blanking, which we don't want to do. So I'd rather leave that URL in as a full exception while it's being used. As for the other one, I'll remove it. --Shirik (Questions or Comments?) 15:58, 12 February 2010 (UTC)

The article toilet already has the word "poop" in it in the "see also" section...so there's no need to have the exception...(A whole discussion about poop @.@)Smallman12q (talk) 13:59, 13 February 2010 (UTC)

Just beause the article already has the word "poop" in it does not mean the exception is not needed. In fact, that exception is most definitely needed for that article. There is no check for existing contents in the article. --Shirik (Questions or Comments?) 17:30, 13 February 2010 (UTC)

Improving filter 98

Filter 98 "Creating very short new article" currently reads:

!("autoconfirmed" in user_groups) &
(article_articleid == 0) & (article_namespace == 0) & 
(new_size < 150) & !('disambig' in article_text) &
!(contains_any(lcase(new_wikitext), "{{surname}}", "{{given name}}", "{{delrev}}", "#redirect", "{{softredirect}}", "{{db-unpatrolled}}")) &
!('disambig' in lcase(new_html))

Would it be better like this:

!("autoconfirmed" in user_groups)
&(article_namespace == 0) 
&(old_size == 0) 
&(new_size < 150) 
&!('disambig' in lcase(article_text))
&!(lcase(new_wikitext) rlike "({{(surname|given name|delrev|softredirect|db-unpatrolled|increation|newpage|under construction)}}|#redirect)")
&!('disambig' in lcase(new_html))

Shouldn't "article_text" be lcase?

I've also added some of the templates from Category:Under-construction_templates.

• Is "old_size" faster/slower than "article_articleid"?
• Is "rlike" faster/slower than "contains_any"?
• In what order are the conditions evaluated? Found at Wikipedia:Edit_filter/Instructions#Controlling_efficiency

Also, is it possible to get some "standard" for the formatting of the abuse filters...it'd be nice if they were more consistent.Smallman12q (talk) 00:09, 13 February 2010 (UTC)

1. No, they should be identical as they are both integer comparisons. However, "old_size==0" is wrong, and will match more than just article creations.

2. Maybe. In this particular case, the answer is "probably not" but I don't want to go into the technical details of PCRE to explain why. After re-reading the situation I change this to "probably" as I missed a few characters that would be relevant. In particular, the fact that almost all of them are templates is relevant. --Shirik (Questions or Comments?) 00:45, 13 February 2010 (UTC)

3. Left to right, as you found.

4. Strongly opposed. Stroustrup has a good discussion of why obsessing over standardization of small details like formatting is counter-productive in TC++PL, but assuming you don't have that book, suffice to say he argues there really isn't a strong reason to force some kind of formatting, so long as a reasonable formatting is present in all cases. --Shirik (Questions or Comments?) 00:38, 13 February 2010 (UTC)

I'm not sure if you answered this...should article_text be lcase?Smallman12q (talk) 12:31, 13 February 2010 (UTC)

Probably not a bad idea .. I bet there's a few pages hanging around that have (Disambiguation) instead of (disambiguation) in the title. —Soap— 12:49, 13 February 2010 (UTC)

I'm a bit worried about running lcase(article_text) as that is a massive amount of string manipulation. There must be a better way, probably by doing article_text rlike "[Dd]isambig". --Shirik (Questions or Comments?) 16:50, 13 February 2010 (UTC)

Would that be faster? To do a regex check rather than lcase? Also could you reply to the "Filter 46 "Poop" vandalism" thread above.Smallman12q (talk) 17:26, 13 February 2010 (UTC)

lcase() relies on manipulating the entire contents. For something like article_text this could take an extremely long time compared to a regular expression which merely looks for a D or d in already-existing text. --Shirik (Questions or Comments?) 17:32, 13 February 2010 (UTC)

Note that article_text is the title of the page, not the contents of the article. If Im understanding you correctly I think you may have confused that. —Soap— 17:37, 13 February 2010 (UTC)

Whoops, yes I am. Though on that note, I wonder if this is inaccurate. Surely there are more disambiguation pages than those that are marked as such in the title. For an incredibly obvious example, see John Smith (though this is not a short article, but ignore that for now). --Shirik (Questions or Comments?) 17:46, 13 February 2010 (UTC)

That would explain the new_html check (though that's a strange one when new_wikitext would probably be just as good without the costs). Perhaps the check on article_text is redundant (unless someone sees something I don't). --Shirik (Questions or Comments?) 17:49, 13 February 2010 (UTC)

You could remove "article_text" as it is redundant.Smallman12q (talk) 02:03, 21 February 2010 (UTC)

suggestion

Could "haha" be added to one of the filter rules? Possibly extended onto the "poop" vandalism filter -- penubag  (talk) 22:27, 16 February 2010 (UTC)

While this should be at WP:EF/R, I'd say I'd be a bit concerned with the number of false positives this might cause. You wouldn't think "poop" would cause false positives, but it has quite a few, and I'm concerned that "haha" would be quite similar. --Shirik (Questions or Comments?) 05:11, 17 February 2010 (UTC)

In order to determine the number of false positives, you can create a test filter. Ruslik_Zero 19:45, 17 February 2010 (UTC)

Would Special:AbuseFilter/135 already cover this? It's one of those densely coded ones that takes a while to figure out (it also is tag-only because of the extremely high rate of false positives).I think that it would not be possible to create a separate filter for 'haha' only because it would conflict with this one, and thus show only this one. —Soap— 23:23, 17 February 2010 (UTC)

It would catch "hahaha" but nothing shorter. As for coding additional filters, I'm growing increasingly concerned about the condition limit hit rate, so I'm not sure how well this will hold up, but I guess it could always be tried. I expect it to eat quite a few conditions, though; conditions which we don't have to spare. --Shirik (Questions or Comments?) 11:55, 18 February 2010 (UTC)

I take that back, I misread it. It would catch "hahahahahahaha" and nothing shorter. (1 + 6) --Shirik (Questions or Comments?) 12:08, 18 February 2010 (UTC)

I don't have rights to view/edit filters, so I cannot test this. -- penubag  (talk) 00:20, 20 February 2010 (UTC)

The edit filter is largely on regex. You can test the "filter" part using regex and some variable text.Smallman12q (talk) 01:48, 21 February 2010 (UTC)

Filter 18

Apparently, this filter has been turned off. (Test edit tag.) In speaking with an edit filter manager Wknight94 ^talk, he informed me why this filter was turned off, saying "At the time, the edit filter performance was poor so I shut a few unnecessary ones off. If things are better now, it can be turned back on. "

So, I propose that if performance has improved, this filter should be turned back on. I ran a search on all articles with the text "Headline text" and got 3,079 results. This is the default text for the button, adding == Headline text ==. to the page. Give that number, and the fact there are other issues with the likes of '''Bold text''' and others, I think it would be helpful to turn this back on. --Avicennasis 01:43, 21 February 2010 (UTC)

There were two reasons I believe, though before we go into the second, let's address that of performance. We're currently at the edit filter condition limit, and there needs to be a lot of optimization done before we can do much more. Otherwise, filters are just going to be kicking each other off the table. So unfortunately, I do have to say the performance concerns still remain. However, there's another issue: I believe part of the problem with this filter was that it was too bitey. Since it's not really "abuse" so to speak, denying edits where users have a lot of useful content but one little mistake from clicking on the bar that they can't find, we'd rather the edit go through. This is just my view here, though, and certainly not any kind of decision, just thoughts. --Shirik (Questions or Comments?) 03:15, 21 February 2010 (UTC)

I know a few of the tags will stop you from saving the page, such as blanking everything will usually not let you save no matter what. However, a lot of the other tags may prompt you once to check your text, and make you press save a second time. At least for me, it usually goes through the second time. Not much of a big deal to me, but I guess I can understand how a new person might get scared off by a big "WARNING: YOU MAY NOT KNOW WTH YOU'RE DOING" at the top of the edit screen. As for tag limits, I was not aware that only so many could be active at one time, although I see the logic in such. In the meantime, I will try to cleanup what I can - I made a search list of common test edits, which so far has provided me with happy hunting. There are a few searches that return 15 000+ test edits, so I may suggest a bot, or try to get approval to write one myself to help. Probably a long ways off for me, though. --Avicennasis 09:26, 21 February 2010 (UTC)

I thought there was already a bot request underway for this, and User:TTTSNB was going to handle it, though I can't find it anywhere on the WP:BRFA page so I might be mistaken. I do know that TTTSNB cleaned up a lot of these hits manually, though, so there was a time not long ago when there were even more stranded test-edit code chunks then there are now. —Soap— 12:07, 21 February 2010 (UTC)

Semantics

When viewing a private filter without rights, you are given the following message:

"You may not view details of this filter, because it is hidden from public view."

Why is there a comma before the because? Shouldn't it read:

"You may not view details of this filter because it is hidden from public view."

Generally, a comma is not used with the because clause, however, it can be used to prevent lexical ambiguity (which does not appear to be the case here).Smallman12q (talk) 01:47, 21 February 2010 (UTC)

Anybody else agree? (Or am I missing something?) Smallman12q (talk) 01:47, 21 February 2010 (UTC)

Agree (as a member of WP:GOCE) but I'm not sure if this is configurable in the MediaWiki namespace or if it's actually part of the extension itself. Perhaps someone else knows better. --Shirik (Questions or Comments?) 03:18, 21 February 2010 (UTC)

I agree, and I have changed the message. Ruslik_Zero 09:44, 21 February 2010 (UTC)

False positive reports

I posted this at WT:FALSEPOS but I figured I'd cross-post here since this place gets noticed more. I revamped the false positives system. False positives are now recorded at WP:Edit filter/False positives/Reports and transcluded into WP:FALSEPOS. That allows us to use a preload template that will automatically record some of the information, helping out new users trying to report false positives. I think this is a good and necessary change for usability. Feel free to throw up any concerns. --Shirik (Questions or Comments?) 03:17, 21 February 2010 (UTC)

I would use '&editintro=' instead of an edit notice. Ruslik_Zero 13:28, 21 February 2010 (UTC)

Interesting. I did not know about the &editintro= capability. As I'm not a sysop I can't change it now (I can add the editintro but I can't remove the edit notice)... but you can ;) I have no objections. --Shirik (Questions or Comments?) 16:42, 21 February 2010 (UTC)

I moved the editnotice to Template:Falsepositive/Editintro and updated WP:FALSEPOS. The main advantage is that now the message is not shown when I edit sections. Ruslik_Zero 17:48, 21 February 2010 (UTC)

Filter reduction

Because we're at the cap on the edit filter, I went through and picked out a bunch of candidates for optimization and left my recommendation here. Any thoughts on these ideas would be most appreciated:

• 275 – Recommendation: Disable – 6 hits in the 2 months it's been active. Has a significant number of conditions, though analysis shows that it should only be about 2 for non-moves. An alternative is to optimize it to place the check for move first.
• 270 – Recommendation: (Weak) Disable – This filter has triggered a lot of false positives in the past and has been annoying people. Don't we have a bot that handles these interwiki links anyway? If that's the case, I'm not sure I understand the benefit of this filter. Totally open to discussion; it might just be my understanding of its purpose more than anything else.
• 263 – Recommendation: Disable or merge – This filter has seen no hits since mid-December. It also might be able to be merged into 179.
• 250 – Recommendation: Optimize – This filter eats up several conditions that it doesn't need to, primarily because it checks the same variable three times when it could do it all at once. (I probably will do this without further discussion because there's really no reason against it that I can see.)
• 213 – Recommendation: Disable – This is a very simple filter, but "a condition saved is a condition earned". Put simply, this filter hasn't had any hits since October.
• 177 – Recommendation: Disable – Hasn't had a hit since December, and I can't tell, but it doesn't seem like it's had any legitimate hits since August (though the filter notes seem to imply differently). Is this still needed?
• 151 – Recommendation: Optimize – This is a pretty complex filter, and it seems to me like it should be possible to optimize this filter. I did some work on testwiki but haven't been able to come up with an optimized solution, but perhaps another pair of eyes can help.
• 148 – Recommendation: Optimize – This filter is not taking advantage of short-circuit logic. This is more for my own notes and will probably be done by the time you read it.
• 139 – Recommendation: Optimize – This filter could probably benefit from some condensation of the last two condition sets (which check added_lines and removed_lines) by turning it into a regex instead of trying to do string manipulation.
• 132 – Recommendation: Optimize – This was discussed a while ago, but this filter probably can still benefit from being transformed into a regex instead of a call to contains_any due to the similarities between many of the arguments.
• 82 – Recommendation: Disable – This filter appears to, at least in part, target the same issue as 278. While this filter has had a lot of hits recently, they all seem to be semi-false positives (I say semi- because they weren't exactly "good" edits, mostly just test edit spam, but that's really not what this filter is designed to detect). I would advocate dropping this (surprisingly complex) filter in favor of the simpler 278. I haven't seen this issue with any other URL, so I don't think that this would be a problem, and of course 278 can always be adjusted to add more URLs if necessary.
• 58 – Recommendation: Optimize – I think collapsing this into a regex will help immensely.

Now I'm not necessarily recommending all of this be done; I'm just throwing out ideas of things that can be done. I'm probably going to go through the optimizations I recommended without much discussion, but I'd rather not be the only person that says "we should disable this" before we go out and do it. That's a bit in excess of WP:BOLD. Any second opinions? --Shirik (Questions or Comments?) 18:19, 21 February 2010 (UTC)

I wish there were a way to sort the filters by condition consumption rate. I am aware of the problem, and have wanted to find a time to go through all of them and see which ones are the most overconsumptive relative to their usefulness, but I haven't had a chance. Is this the result of you having done that and chosen the ones you found most unnecessary? —Soap— 18:24, 21 February 2010 (UTC)

Just a note that I decided to go through all of the filters after all, and found that the number of conditions consumed fluctuates wildly. For example, filter 98 was over 100 conditions a few minutes ago and now it's 10. I'm not really sure what to think. —Soap— 18:47, 21 February 2010 (UTC)

In general I picked out the more complex filters, however there's a few simple filters in this list too. I did focus on filters which are using a lot of conditions for what is supposed to be a simple task, though. An automated capability to sort by average condition usage would be very nice, though with the amount of fluctuation that number sees on a daily basis, I wonder if that's even possible.... To answer your question directly, I did focus mostly on filters that are very heavy and then evaluated their usefulness. --Shirik (Questions or Comments?) 18:27, 21 February 2010 (UTC)

54 probably could be disabled. It's consuming quite a good many conditions for not that much gain. NW (Talk) 18:32, 21 February 2010 (UTC)

Filter 98 change

I think a recent change to filter 98 has introduced a problem. It seems to no longer be catching very short new articles, and now only catches redirects that have been created. I'm not very familiar with filter language, but it seems this change [2] may have introduced this issue. Could someone more familiar with the filter syntax take a look please? Thank you! ConcernedVancouverite (talk) 01:48, 24 February 2010 (UTC)

I apologize for that. One tiny missing character completely inverted the entire test. It's supposed to ignore redirects, etc., and instead it looked only for redirects, etc. Thanks for pointing this out. It should be fixed now. --Shirik (Questions or Comments?) 02:31, 24 February 2010 (UTC)

Thank you for the quick fix! I'm glad it was an easy one to get working again - it is a very useful filter! ConcernedVancouverite (talk) 03:00, 24 February 2010 (UTC)

Filter 289

Isn't necessary to put lcase(article_text) in Special:AbuseFilter/289? Helder (talk) 17:44, 24 February 2010 (UTC)

No. Review the hits. It may be more accurate if lcase were used, but the problem it would detect would be extremely rare, and almost certainly not worth the additional computation time. --Shirik (Questions or Comments?) 02:07, 25 February 2010 (UTC)

It would be easier to change the regex. Prodego ^talk 02:30, 25 February 2010 (UTC)

Question/Proposal

I was wondering what the options are when a filter is activated. Is it simply a case of disallow/tag/ignore? I'm interested because I had an idea for a new filter, if such a thing does not already exist. Basically, the filter would be a simple dictionary check, looking for words (cock, gay, penis, etc) that have legitimate uses on-wiki, but which are also frequently used by vandals. Upon a hit, the filter would post the edit to an IRC stream, which Huggle would draw from in the same way it draws from the recent changes stream. Edits from the new stream, however, would be automatically placed in the front of the cue, so Huggle users would see them right away. The filter might be similar to that used by ClueBot, but stricter. The false positives that result would be dealt with by Huggle users. This would obviously require a Huggle update, but I don't think that should be too big a deal.

Anyways, I'm not sure if this is even possible, but I thought it might streamline vandal fighting a bit, so I brought it here to ask everyone what they think. Thanks Throwaway85 (talk) 08:17, 10 January 2010 (UTC)

The filter pretty much exists; in fact we have several filters doing that, but the technology to create a feed from it which can be viewed in Huggle does not yet exist (I've asked this before). I agree it would be a great idea, but I don't think it's possible at least not yet. -- Soap ^Talk/_{Contributions} 14:32, 10 January 2010 (UTC)

I figured as much. Thanks anyway. Throwaway85 (talk) 12:11, 11 January 2010 (UTC)

I just thought about a workaround. What about creating its own unique tag, and having a bot patrol for those tags then post to an IRC stream? Or just update huggle to give preference to those tags in the recent changes stream? Throwaway85 (talk) 22:37, 11 January 2010 (UTC)

It sounds like a good idea, but I don't know if it's technologically possible to get Huggle to recognize tags. Anyone know? -- Soap ^Talk/_{Contributions} 01:26, 15 January 2010 (UTC)

It would depend on how tags are dealt with. From veiwing the raw IRC feed, it doesn't look like tags are included, so it would require a minor amount of dev support to change that. Were that to happen, it would be a simple matter to make Huggle prefer tagged over non-tagged edits. The nice thing about that is it would be easy to make it work for all of the existing tags as well. Seems that dev support for hg is a little light tho. Throwaway85 (talk) 11:43, 15 January 2010 (UTC)

I once developed a filter to flag edits that were made to pages that were likely unwatched. If Huggle supported tags, then this would be very useful. Triplestop x3 03:29, 18 January 2010 (UTC)

Huggle cannot "support tags" when neither the MediaWiki API nor the IRC recent changes feed support them. Gurch (talk) 15:00, 4 February 2010 (UTC)

In case you don't know, there is a Special:AbuseLog.Smallman12q (talk) 23:13, 11 February 2010 (UTC)

I am quite aware of that. That is not the API, nor is it the IRC recent changes feed. Gurch (talk) 18:10, 12 February 2010 (UTC)

Why can't you just scrape the the log? Is there a policy against that?Smallman12q (talk) 00:12, 13 February 2010 (UTC)

I imagine it wouldnt be as efficient as Huggle. You'd basically have to refresh the page every few seconds to get Huggle-like functionality out of it, and even then you'd be viewing all the filters at once and would have to ignore most of them. (Or else view the logs for each filter independently which would be even more tedious.) —Soap— 00:25, 13 February 2010 (UTC)

Special:AbuseLog doesn't specify which tag(s) were used. It would be necessary to scrape Special:RecentChanges instead. Repeatedly requesting that page at short intervals (rather than connecting to an IRC feed, getting recent changes that way, and using the API for any additional information) is inefficient, requires ugly code, is a burden on both client and server if it is to be done frequently enough to keep Huggle's internal state anything approaching accurate, and is almost impossible to provide consistent error handling for because the whole layout and content of the page, as well as associated error messages and so forth, can be customized by any particular MediaWiki installation. Gurch (talk) 03:18, 13 February 2010 (UTC)

It doesn't give the tags, but it does give which filter from which you can then determine the tag. Why not write a bot that creates an IRC feed from the edit filter log and then use that feed?Smallman12q (talk) 12:35, 13 February 2010 (UTC)

Gurch would know better than me but I dont believe it's possible ... the AbuseLog isnt a "feed" that can be piped into a bot like that unless you have it constantly refresh the page and parse the input. Basically what I said above, although I didnt realize you were talking about a bot at the time. -- —Soap— 13:50, 13 February 2010 (UTC)

Determining the tag from the filter is not really feasible. For public filters, one can view the filter description and see the tags there, but that means either loading the filter description page every time a filter is used, which is not feasible, or loading them all once, in which case if anyone changed the tags a filter was using, one would still be claiming the old tags were assigned. For private filters one can't even do that -- the description page is private. The only way to tell what tags a private filter applies is to look at the log for that filter, which is an awfully kludgy way to do things and of course completely breaks if the filter's tags are modified as before.

Creating a bot would make Huggle reliant on a third-party bot that someone would have to host (I don't have the resources to do so), this sort of dependency is something that has to be avoided. Adding tag information to the actual IRC feed would help, but my request for this has been ignored for over a year now. Gurch (talk) 14:13, 13 February 2010 (UTC)

There is an API for the abuse log at Extension:AbuseFilter for example: http://en.wiki.x.io/w/api.php?action=query&list=abuselog&afluser=SineBot&aflprop=ids .Smallman12q (talk) 01:55, 26 February 2010 (UTC)

Yes, I am quite aware of that, I wrote the documentation you just linked to. Please read the rest of this section. That API does not include change tags. Gurch (talk) 05:54, 26 February 2010 (UTC)

Filter 50

What about using

 & (added_lines rlike "^[A-Z0-9\s\pP]*?[A-Z]{5}[A-Z0-9\s\pP]*$")
 & !(added_lines rlike "#REDIRECT|__((NOEDIT|NEW)SECTION|(NO|FORCE)?TOC)__")

instead of

 !('''"#REDIRECT"''' in added_lines) & (added_lines rlike "^[A-Z0-9\s\pP]*?[A-Z]{5}[A-Z0-9\s\pP]*$")
 & !(added_lines rlike '''"__(NOEDIT|NEW)SECTION__"''') & !(added_lines rlike '''"__(NO|FORCE)?TOC__"''')

? Helder (talk) 23:36, 24 February 2010 (UTC)

That would be better, its pointless to do an "in" check if you're already using "rlike" on the same variable. Shouldn't it be "(#REDIRECT|__((NOEDIT|NEW)SECTION|(NO|FORCE)?TOC)__)", or is the outer parenthesis not needed?Smallman12q (talk) 01:48, 26 February 2010 (UTC)

There is no difference in the outer set of parentheses. It actually will make the match (slightly) slower due to the creation of an additional capture group that is unnecessary. (The edit filter uses preg_match() for its processing.) --Shirik (Questions or Comments?) 18:18, 26 February 2010 (UTC)

New thread about Filter 139

See Wikipedia_talk:Administrator_intervention_against_vandalism/TB2#Filter_139; I dont really know if here or there would be a better location to carry on but since it's there I' ll post a link here. —Soap— 02:09, 27 February 2010 (UTC)

Filter 279

Why is this restricted to the article namespace? I think it would help out with 293. Is there any particular reason against making it global? Note that I don't particularly understand 279 very well, except for the general concept. --Shirik (Questions or Comments?) 07:14, 20 February 2010 (UTC)

The idea is that any new user making too many edits to fast is likely to be a vandal. Ruslik_Zero 19:28, 20 February 2010 (UTC)

Not exactly... but my concern wasn't really about what the filter is for (and I don't really want to go into the details of that in a public venue). Instead I'm just trying to figure out why it only targets the article namespace instead of all edits. It would have helped out with 293 (which is now deleted) and I'm sure it will be useful in the future if that limitation is removed. --Shirik (Questions or Comments?) 16:45, 21 February 2010 (UTC)

You've decided to make my life difficult by not allowing the rest of us to see this filter, however one thing I can tell about it from looking at its log is that it's not only looking at all namespaces, it's even looking at the sandbox. Please stop doing that. Gurch (talk) 08:27, 3 March 2010 (UTC)

Q about #39

This strikes me as a very odd false positive, especially when it's labeled as "possible BLP violation". What were the triggering conditions? Merely that it was an IP edit to a schools page?--Father Goose (talk) 20:17, 4 March 2010 (UTC)

Any edit containing the word "fired" on an article for a school will trigger that filter. It's a false positive in this case because it's the wrong kind of fire, but apparently there is a problem with people vandalizing high school articles saying that various teachers have been fired by the administration. (And note that due to the way the MediaWiki software operates, most edits where a word is inserted into the middle of a paragraph, the software will see the entire paragraph as "new", hence it saw the word fired as new.) —Soap— 20:29, 4 March 2010 (UTC)

Filter 17

Filter 17 seems to be getting some false positives lately that I can't understand. In particular this and this report, which don't seem to match the watchwords that are in the filter, as far as I can see. Am I reading the code wrong? This filter is difficult to work with because even its false positives are often controversial edits I'd be reluctant to endorse. —Soap— 01:33, 8 March 2010 (UTC)

Something seems to be wrong. I tested it out on testwiki as testwiki:Special:AbuseFilter/83 and eventually tore it down to a very small piece which shouldn't be matching. My regex debugger says it doesn't match, but the abuse filter says otherwise. I really don't know what's going on here, it's very peculiar... --Shirik (Questions or Comments?) 02:35, 8 March 2010 (UTC)

I figured it out. For some reason it was matching "Dieudonné donne une 'conférence' dans un car"; the most obvious reason is the accented e looking like a word boundary (and then matching "rence"). I'm not sure what to do about this, because it's definitely not a word boundary. Perhaps the abuse filter doesn't like UTF-8? (While that character can be represented in ASCII, it appears to be represented in UTF-8.) --Shirik (Questions or Comments?) 03:16, 8 March 2010 (UTC)

I've filed a bug report --Shirik (Questions or Comments?) 03:29, 8 March 2010 (UTC)

Filter 279

A recent experiment seems to have shown that filter 279 is not working the way it should be. User:Dr Aaij is a professor engaging in a school project. He recently had all of his students sign in on his talk page. As a result, this filter blocked several students citing "repeated vandalism attempts". When I looked, there were no hits, and each student only posted once. It appears that the throttle for this filter is incorrect. However, I've looked over it and it looks correct. (I will gladly give the details of my analysis in private, I just don't want to discuss the details of a private filter in a public venue.) If anyone could give it a second look that would be great. --Shirik (Questions or Comments?) 20:14, 8 March 2010 (UTC)

As a note, I have temporarily exempted that page from this filter to allow his project to continue while we figure out what's going on here. --Shirik (Questions or Comments?) 20:15, 8 March 2010 (UTC)

I believe it's just the result of the way the filter is set up, and can't really be fixed without either disabling that part of the filter or adding exemptions manually. Sorry to be a pain, because it's a really simple thing, but I'd rather not give out the answer here because it's a private filter and I've been warned before that private filters are supposed to be private all the way down. I did put the comment into the filter comment box though. —Soap— 22:41, 8 March 2010 (UTC)

Bot to help out?

I've been giving this some thought, and I've thought about writing to bot to help out at WP:FALSEPOS. What I'm thinking is the following (written informally according to IEEE 830, so please pardon if it's a bit technical; feel free to ask questions).

Requirements

Definitions Resolved A marking on a section through the use of the template {{done}}, {{notdone}}, or {{resolved}}, case insensitive Level 3 section A wiki-encoded section using the format ===Section name=== Whitespace Any space, tab, line feed, or carriage return character Functional requirements Upon detection of a new level 3 section at WP:FALSEPOS/Reports, if the code : [[]] ({{plainlinks|URL=http://en.wiki.x.io/wiki/Special:AbuseLog?title=Special:AbuseLog&wpSearchTitle={{urlencode:}}|NAME=filter log}}), whitespace ignored, is present in the section, the system shall erase the entire section and post a standardized message to the user's talk page indicating to specify the problematic article in the report. Upon detection of a new block, if the block applies to a user who initially created one or more level 3 sections, the system shall erase the entire section(s). Every 24 hours the system shall move any level 3 section which is marked resolved to an archive area. Non-functional requirements The system shall use the recent changes IRC feed to monitor block logs and new reports. The system shall not operate without the approval of the bot approvals group

My primary justification for this is fairly simple. (1) The false positives page is very useful in identifying problematic filters. We should try to keep it as consistent as possible without biting newcomers, who are most likely to be hit by the filter. (2) I recently changed the WP:FALSEPOS system to be much more user-friendly. As a result, this has helped users (both non-vandals and vandals) to quickly produce false positive reports. Unfortunately, every day there is a significant number of these that come from vandals. Making this system harder to use to avoid vandals is counter-productive. Instead, we should try to reduce our workload in dealing with these obviously problematic posts. (3) Eventually, the false positives page will simply become too long to manage reasonably, so there should be some sort of archiving system to help manage the page.

I suspect #1 will be the most likely to encounter controversy, so let me just point out that, in nearly all cases, if not 100% of the cases, when a title has been missing entirely (not just improperly formatted, but completely missing), this has been indicative of a vandal's nature. However, in the interest of assuming good faith, instead of just outright deleting the section, I think giving the user more detailed instructions on how to proceed may be beneficial to anyone who hits the rare case of making an appropriate, but incorrect report. The other two requirements I feel are self evident.

I apologize to ramble on and on about this, so I'll cut it here in the hopes of inspiring discussion. What does everyone else think of this? Proposals on additional limitations? Any other ideas? Feel free to tear my idea apart. --Shirik (Questions or Comments?) 06:41, 3 March 2010 (UTC)

I dont really have a strong opinion here, but I think the bot is about to archive this so Im just posting a dummy reply to keep it live in case anyone else wants to comment. —Soap— 12:16, 13 March 2010 (UTC)

Filter 129 expected movies

To help narrow it down, it could incorporate terms like "expected" and "anticipated". SGGH ^ping! 13:04, 16 March 2010 (UTC)

A reasonable change, but I would like to first build consensus for this change as I'm not exactly sure what this filter is targetting. Any objections from anyone? --Shirik (Questions or Comments?) 18:32, 16 March 2010 (UTC)

Filter-exempt userright

It is known that significant amount of edits in Wikipedia is made by bots, sysops and other trusted accounts. Currently the only way to exclude their edits from edit filters is to use conditions like !("sysop" in USER_GROUPS). This method is actually inefficient because every filter makes independent checks. I think it would be beneficial to have a userright called, say, 'abuse-filter-exempt'. Any user with such a right would be automatically exempted from all filters. Ruslik_Zero 19:44, 16 March 2010 (UTC)

Just a note that this was discussed a few months ago here and failed to achieve consensus. (Personally, I still have mixed feelings about it just as I did then, and can't really support or oppose.) —Soap— 20:19, 16 March 2010 (UTC)

Captcha

This is a proposal to request from developers to add Captcha to the list of action options. When triggered the filter would ask the user to fill a Captcha to pass. This aim to make vandalism less fun, and consequently less likely to be repeated again. Sole Soul (talk) 19:01, 18 March 2010 (UTC)

Captcha is used to prevent (unauthorized) automated processes from making edits and certain actions. I do not understand how it can help with vandalism. Ruslik_Zero 19:23, 18 March 2010 (UTC)

"This aim to make vandalism less fun, and consequently less likely to be repeated again." Sole Soul (talk) 19:38, 18 March 2010 (UTC)

This would be a misuse of captcha. Ruslik_Zero 19:53, 18 March 2010 (UTC)

Different use is not necessary a misuse. Sole Soul (talk) 20:15, 18 March 2010 (UTC)

I concur. This could be a very effective way to reduce vandalism- it wouldn't stop the more determined (or the incredibly bored) trolls and vandals, but I think it;s worth a try if it can be done. HJ Mitchell | Penny for your thoughts? 22:20, 18 March 2010 (UTC)

Filter for removal of header in Wikipedia:Introduction

There are recurrent discussions about protecting Wikipedia:Introduction, most recently here, on one hand we want to keep it edit-able so that new users can edit for the first time, on the other hand, as you can see, users often remove the header and replace it with various text, so readers don't see the introduction but instead random things. So what do you think of a filter which would check if the header is removed, and warn and/or disallow when it is, but which would let in editing below the header ? This would give users most of the editing experience but prevent the removal of the header and the resulting deception of readers. Cenarium (talk) 04:03, 22 March 2010 (UTC)

I do not think that creation of single page filters is a good idea. Ruslik_Zero 05:25, 22 March 2010 (UTC)

It goes two ways. On one side, it's not really a performance problem. A single-page filter would eat, in the overwhelming majority of cases, an extremely small amount of time and a single condition. On the other hand, this argument could be applied to many different pages and making a filter like this could be a poor precedent. I personally think this page is special; it's one of the first things some new editors see. If it's an extremely recurring problem (like several times a day), perhaps it is worth it. But if we're talking about once a week, I don't think it is. What's the typical rate on this? --Shirik (Questions or Comments?) 06:46, 22 March 2010 (UTC)

A filter will consume 3 conditions at minimum: the condition itself, parentheses around it and the first logical operator. Ruslik_Zero 19:19, 22 March 2010 (UTC)

The removal of the header happens several times a day, from three to six times. I agree it's not an optimal solution, but I don't see what else we could do, besides protection. Cenarium (talk) 18:55, 24 March 2010 (UTC)

"In the overwhelming majority of cases" is the key point. The edit filter uses short-circuit logic, and the check for the page can be the first condition. That means, unless we're talking about an edit on the page itself, it will always only consume a single condition. --Shirik (Questions or Comments?) 03:10, 26 March 2010 (UTC)

Why is this necessary? SoxBot does exactly this, restoring the header when it is removed, [3].--Jac16888^Talk 03:48, 26 March 2010 (UTC)

Yes, but it takes a few minutes most of the time. Cenarium (talk) 13:45, 26 March 2010 (UTC)

So? A few minutes is no big deal, and a page for newbies with no introduction is a lot less hurtful to us than when a person's first experience here is being stopped by the filter--Jac16888^Talk 14:55, 26 March 2010 (UTC)

And any performance savings, no matter how small, are very important to the filter, especially when the work can be offloaded to a bot. WP:PERFORMANCE rarely applies to the edit filter. --Shirik (Questions or Comments?) 17:52, 26 March 2010 (UTC)

The (default) text of the edit window asks users not to remove the header, so they can expect some kind of reaction if they do it, and in any case, the warning can be nice. While a few minutes may seem not much, frequently repeated for a high-visibility page, it becomes quite a deal, with dozens or more readers not seeing the header in a day. I feel it's more important to give an actual introduction to the newbies, who expect it, or we may loose them. Cenarium (talk) 02:35, 27 March 2010 (UTC)

Filter 225

This filter has notes mentioning false positives as a reason for disabling it. I think this is one of the more useful filters. Surely we could fix this up, probably just by deleting whatever is causing it, but I don't see the false positives myself. Can anyone point me in the right direction? --Shirik (Questions or Comments?) 21:09, 31 March 2010 (UTC)

Disabled filters

I went through and disabled a bunch of filters that hadn't had recent hits and were very expensive and didn't seem to have a current need for a filter. We were at about a rate of 3.3% on the filter limit. Just a friendly reminder to everyone to please keep an eye on this number and don't keep making more and more filters without finding ways to fit the conditions in (either by disabling or optimizing old ones). If anyone has a question about something I disabled, please feel free to ask. Thanks! --Shirik (Questions or Comments?) 02:48, 1 April 2010 (UTC)

Requesting filter modifications of 39 and 289

I'm not sure if this is the right place, but I'm requesting two edit filter modifications. For filter 39, I'm thinking of the schools which fall under the Category:Fictional schools to be removed. And also, for filter 289, "New user creating interrogative pages" what about including the words "Who" and "Which" for the heading of the page? Minimac (talk) 17:02, 22 March 2010 (UTC)

The first is difficult to implement, because it will consume lots of resources. The second is not necessary as there a lot of page beginning with "Who". Only "which" can really be added. Ruslik_Zero 19:16, 22 March 2010 (UTC)

I added Which. Ruslik_Zero 13:22, 5 April 2010 (UTC)

3RR filter

Hello, is there a filter for violation of the 3RR? --The High Fin Sperm Whale 04:45, 6 April 2010 (UTC)

It's not really possible to code that kind of logic into a filter. One could theoretically make a throttle-warning filter set at 3 reverts per 24 hours, but distinguishing between reverting vandalism and edit warring would be practically impossible, and such a filter might really cause problems with automated tools, etc., (not to mention annoy recent change patrollers). I don't think it's feasible. --Shirik (Questions or Comments?) 05:26, 7 April 2010 (UTC)

Request for permission

I wished to request for permission to view filters; perhaps an EFM permission (abuse-filter private) to only 'view' filters. The permission would allow me to view and study current filters so that in the future, I may propose/create new filters post non-public discussions and post testing. But the future creation does not mean I need any right to edit currently. When (and if) I do need such a right, I would again request out here for an explicit permission. Warm regards, ♪ ♫ Wifione ♫ ♪ _―Œ ^{♣Łeave Ξ мessage♣} 19:09, 3 April 2010 (UTC) (As I'm out for the weekend, in case you have questions, I would be able to answer them from Tuesday morning onwards as I would be actively editing only from then. In between, if by chance any of you see me on the ACC Tool Server interface, do drop me a note and I'll rush back. Thanks and regards)

The best place to ask for this is WP:RFA. Ruslik_Zero 19:22, 3 April 2010 (UTC)

Hi Ruslik. Stole a look here the moment I got time. I don't think RfA is the right place. The main page of the edit filter page says this > "Presently, requests for assignment of the "Edit Filter managers" group to non-admins should be made at Wikipedia talk:Edit filter, where a discussion will be held for up to a week prior to a decision being made". Do kindly guide me on how to proceed. If not tomorrow, will surely answer on Tuesday to any queries. Warm regards. ♪ ♫ Wifione ♫ ♪ _―Œ ^{♣Łeave Ξ мessage♣} 18:01, 4 April 2010 (UTC)

Edit filter permission is only for those who edit filters. If you want only to view rights you should apply for adminship as administrators have this right by default. (There is no way to grant it separately.) Ruslik_Zero 13:13, 5 April 2010 (UTC)

We have granted EFM in the past for "view-only" use, see Wikipedia talk:Edit filter/Archive 4#Request for permission. Disclosure: I closed the discussion. No comment on present case. –xeno^talk 13:21, 5 April 2010 (UTC)

Umm, so did I make it? Rgds... ♪ ♫ Wifione ♫ ♪ _―Œ ^{♣Łeave Ξ мessage♣} 18:50, 9 April 2010 (UTC)

(Responding due to talkback template left on my talk page) While I hadn't planned to be the one to close this; I would have to say that unlike the request linked above, consensus has not been reached to grant you the EFM flag at this time. If there were trusted users to vouch for you, this decision could be revisited. There is also, of course, the above-suggested RFA venue which is far better trafficked than this page. –xeno^talk 16:35, 11 April 2010 (UTC)

As I've been away for a week, I hadn't seen this, but WifiOne asked me to pop over here! Before making comments, I would like to ask WifiOne to explain why he would want to view the current filters - I read what was written above, but I am still not sure of the need. If you can think of a new filter that you would like to propose, you could still propose it here, surely (or if it is something that you think shouldn't be discussed in a public arena, then you could always email a current edit filter user about it) -- PhantomSteve/talk|contribs\ 09:30, 12 April 2010 (UTC)

Sure. The problem is that many edit filters currently are blocked from view. The permission I am seeking is only view rights and not edit rights. View rights would enable me to not invest my time in suggesting filters that are already available. For example, one particular area that I wish to work on is the NPP area where the tags currently being given to new articles could be improved. In a specific example, a particular newly created page that was created by a vandal was tagged 'large unwikified article' and not tagged as I presumed it should have been. That's only one case. There are others. It's clearly a much extra investment of time if the changes that I think a filter should have (or new filters that I might have in mind) already exist. Thus, the request for view rights rather than edit rights. Warm regards. ♪ ♫ Wifione ♫ ♪ _―Œ ^{♣Łeave Ξ мessage♣} 09:54, 12 April 2010 (UTC)

(p.s: That particular vandal was blocked post my reporting him on aiv. I won't post details here but his username too escaped our filters. In the past, I did try my bit on working out how to stymie some types of persistent vandalism (a.A village pump proposal, b.the related RfC). Thus the request to view... ♪ ♫ Wifione ♫ ♪ _―Œ ^{♣Łeave Ξ мessage♣}

Adding a comment as I don't want this to be seen in any different way. Actually Xeno, who had granted the past view-only edit filter right, had done so then with the line, "the permission should allow him [the user] to assist in developing better filters. And if not, no harm done." So I had presumed that I would in the future able to add to the knowledge base that exists out here in various areas, especially vandalism reverting. In case the view-only right is granted, great. In case it is not, I still have no issues. I'll keep working on vandalism reverting through rollback. Thanks. ♪ ♫ Wifione ♫ ♪ _―Œ ^{♣Łeave Ξ мessage♣} 11:01, 12 April 2010 (UTC)

Wifione has asked me to comment on this request, but I do not have anything to add. Stifle (talk) 11:11, 12 April 2010 (UTC)

Ditto Stifle. Stwalkerster [ talk ] 13:21, 12 April 2010 (UTC)

Wifione has asked me to comment on this request, and i have something to say. From the archive that Xeno linked above i see no consistency is reviewing requests for permission. Those who wish to review what is already in place so that they might avoid creating duplications seem to be among the more scrutinised. Some people are given a quiz while others get a {{done}} pretty much just for asking. Considering the issues raised in that archive page about admins just giving themselves the right it seems odd to suggest going through RfA because as is also mentioned, anyone submitting an RfA simply for filter access will surely fail the RfA. That being said i don't believe Wifione has any malicious intent behind making this request, though i may be the wrong character witness as i have nothing to do with abuse filters, though one called me a "tireless sock" last year. delirious & lost ☯ ^~hugs~ 20:03, 12 April 2010 (UTC)

If someone is an administrator, they can add it to themselves. If not, and they clearly are very capable of using the filters, they will probably get added by someone (e.g. User:Shirik). Personally I don't agree with giving the AFE rights to just view private filters, but consensus seems to be if someone is working in an area where they repeatedly encounter situations where having access to the private filters would be helpful, they may be added by an admin as well (e.g. User:Chzz). None of these things seem to be what Wifione is saying he needs AFE for. If you need to "view and study current filters" the majority of them are not private, and already visible. Prodego ^talk 20:23, 12 April 2010 (UTC)

Actually Prodego, this was the statement Chzz gave on the basis of which the edit filter 'view only' permission request was filed. "I'd like to be able to view the current edit filters; I don't intend to modify anything, but realize that the permission would enable me to do so; per the note 'requests for assignment of the 'Edit Filter managers' group to non-admins should be made at Wikipedia talk:Edit filter' I post this here". I believed that the edit-filter view only permission was in the first place created to help trusted editors view all existing filters (including blocked) and suggest changes to existing filters and creation of new filters (areas where I intend contributing). ♪ ♫ Wifione ♫ ♪ _―Œ ^{♣Łeave Ξ мessage♣} 02:22, 13 April 2010 (UTC)

In any case, you will probably note I extensively disagreed with that, and Chzz commonly needed to use the filter for help requests or something similar, I don't recall exactly. Prodego ^talk 02:51, 13 April 2010 (UTC)

Yes, I noticed that. Chzz had got the view-only right but you had disagreed with the reason... I think the factor that was considered finally was not whether the user wishes to really use the edit filters finally (the user can't anyway with view-only rights) but whether the user is trustworthy or untrustworthy. I do realise that having the account creator, autoreviewer and rollback statuses are no guarantees of my trustworthiness, but I do hope that the request is not considered in bad faith. Warm regards. ♪ ♫ Wifione ♫ ♪ _―Œ ^{♣Łeave Ξ мessage♣} 03:55, 13 April 2010 (UTC)

I think the issue here Wifone is simply that there is no need for you to have it. If you wish to learn about the filters there is plenty you can do without having the right, perhaps if you did what you say you're intending to do for a while for the filters you can view - most of them, and show that you could be helpful, then people would be more open to giving you the right. As it is right now, you just come across as wanting another hat. No offence intended--Jac16888^Talk 03:58, 13 April 2010 (UTC)

No offence taken Jac. That's why I think Chzz's archive was shown by Xeno to describe the simplistic process through which Chzz got the edit filter 'view only' right. In my personal opinion, if trying to view blocked filters is viewed by you as 'an attempt to gain another hat', I would be more than pleased to not be given that 'hat' yet have the permission to view the filters. Look, all of us are on the same side - I (and you hopefully) do fight vandalism whenever we get time as volunteers. I linked my past RfC, Village pump proposal, on anti-vandalism for a better idea to editors about how I have tried to 'be helpful'. If you wish more such links to how I have tried to be helpful, I would be more than pleased to link them up out here or on your talk page. Warm regards. ♪ ♫ Wifione ♫ ♪ _―Œ ^{♣Łeave Ξ мessage♣} 04:10, 13 April 2010 (UTC)

I meant helpful as in helpful specifically towards the edit filter system. Saying you want the ability because you intend to start working in the area seems to come across in the same way as those editors who ask for the acc right because they're going to start creating accounts despite never having done so before or the new user who wants rollback after 12 edits --Jac16888^Talk 04:19, 13 April 2010 (UTC)

Yes surely. Sorry if I'm sounding as being critical. The guidelines are generally met with quite strictly when normal editors apply. When the overseeing admin believes the requesting user is a good editor, the guideline may be overlooked. In my case, I got the AWB permission much before the minimum edit count was reached, the autoreviewer right much before the minimum pages for recommendation were reached, the ACC access without having ever created a single account. At least in the last case, guidelines were followed to the tee I presume (Wikipedia:Request_an_account/Guide#Registering_for_use). (I would also suggest the creation of a separate guideline page for telling future requesting editors the procedure to follow and requirements to be specifically met before applying for the view-only permission. This would go a long way in avoiding future confusion). So my basic premise rests on the fact that if the edit filter 'view only' option has been given in the past case with and without certain requirements (please do go through this once), then it should perhaps be a similar criteria for analysing my request. Thanks and regards. ♪ ♫ Wifione ♫ ♪ _―Œ ^{♣Łeave Ξ мessage♣} 06:56, 13 April 2010 (UTC)

So you say strict guidelines should be applied to "normal editors", but for "good editor[s]" there need be none. What exactly makes an editor 'good' as opposed to 'normal'? Prodego ^talk 14:22, 13 April 2010 (UTC)

Hi Prodego. Good editor... Ok, let me give it a shot - one who doesn't pettifog and turn a blind eye to various admins telling the same thing over and over again? :) Don't worry, I've understood :) ♪ ♫ Wifione ♫ ♪ _―Œ ^{♣Łeave Ξ мessage♣} 18:11, 13 April 2010 (UTC)

help desk enquiry

There's a query on the help desk where an editor says whenever he tries to add any references to any pages, he gets a message about external links being present. Do take a look at this and comment if required. Thanks. ♪ ♫ Wifione ♫ ♪ _―Œ ^{♣Łeave Ξ мessage♣} 04:16, 14 April 2010 (UTC)

That is unrelated to the abusefilter, that's the spam captcha. Prodego ^talk 12:34, 14 April 2010 (UTC)

Thanks. ♪ ♫ Wifione ♫ ♪ _―Œ ^{♣Łeave Ξ мessage♣} 19:44, 14 April 2010 (UTC)

on.nimp.org

Even in non-hyperlinked form, might be a good thing to block, lest someone copy and paste it into their address bar. I'm not sure if it still hosts malware, but we definitely don't want to be sending anyone there. Just came across this diff where it was used. Given that on.nimp.org is known by the name "Last Measure", I think we can be assured that blocking the string won't block legitimate reference to the site. Throwaway85 (talk) 01:02, 16 April 2010 (UTC)

I would say this is probably a better candidate for the spam blacklist instead. --Shirik (Questions or Comments?) 02:35, 20 April 2010 (UTC)

Would that also block it if it was entered in plain text, though, or just as a link? —Soap— 15:47, 26 April 2010 (UTC)

It would also work for plaintext URLs (per the Documentation, it simply builds a regex like /https?:\/\/[a-z0-9\-.]*(line 1|line 2|line 3|....)/Si to do the match). --Shirik (Questions or Comments?) 15:53, 26 April 2010 (UTC)

Request for Edit Filter Access: Tim1357

Hey there, I am here to ask for permission to view and change edit filters. I have been active for 8 months now, and have focused mainly on the technical side of wikipedia. I am a member of the Bot approvals group, and active at WP:DBR and WP:DDR. I am well versed in python and regular expression, and will be responsible enough not to break anything too important. Thanks, Tim1357 (talk) 23:35, 25 April 2010 (UTC)

I've seen your work before, and I think it would be great to have your technical skills helping us out. —Soap— 00:24, 26 April 2010 (UTC)

I've seen Tim's work and think he would be a good candidate to have access. MBisanz ^talk 01:29, 26 April 2010 (UTC)

I agree. Prodego ^talk 03:59, 26 April 2010 (UTC)

Agree with the sentiments above. One of the great things Tim has done for us is quickly code a bot that allows us to track filters; he shows the technical competence for the bit, and I have no reason to think he can't be trusted with the tool. --Shirik (Questions or Comments?) 15:48, 26 April 2010 (UTC)

Per the consensus above, I have granted access to Tim1357. Cenarium (talk) 17:55, 26 April 2010 (UTC)

Does filter 261 need to be private?

Does filter 261 need to be private? Sole Soul (talk) 02:28, 18 April 2010 (UTC)

Yes. Someguy1221 (talk) 06:44, 18 April 2010 (UTC)

I tried to view the filter details, but the filter's details have been marked as private. Boygirl22 (talk) 00:58, 1 May 2010 (UTC)

Heh yep. Private filters are not viewable to the general public to avoid vandals taking advantage of the loopholes. Tim1357 ^talk 03:38, 1 May 2010 (UTC)

Test edits

Filter 18 is supposed to detect test edits. I reviewed a lot of hits by this filter, and it was clear to me that in most cases the incorrect wikitext is being added by mistake. It is not a test edit unless the user is editing with the intention to test. The problem here is the warning shown to the user. A lot of new users are already hesitant to edit in fear of screwing something. They need us to tell them to be bold and not to worry about good-intentioned mistakes. I think this filter should be a tag only filter, or at least, the warning message should be tweaked to address the possibility of mistakes. Sole Soul (talk) 18:32, 27 April 2010 (UTC)

Filter 18 has been disabled for nearly 6 months now for this exact reason. There was discussion a while ago to just have a bot do cleanups, but I don't think that was ever followed through on. --Shirik (Questions or Comments?) 13:06, 29 April 2010 (UTC)

How embarrassing, I did not notice the dates. I came across this edit filter by browsing the warning messages, not the Management page. Thank you Shirik. Sole Soul (talk) 13:39, 29 April 2010 (UTC)

When it's possible for huggle to see the tags that have been placed on edits, and prioritize edits based on those tags, filters such as those may become far mare useful. Someguy1221 (talk) 04:41, 1 May 2010 (UTC)

Filter 1

There is a new filter that traps any new article containing more than a series of consecutive exclamation points. This is not a lways a good thing, because sometimes a legitimate article will be created with a sequence like this in it. For example, this user tried more than a dozen times to create his article while continually being stopped by the filter. While I won't endorse the creation of this particular article, there may be cases in the future in which this filter would stop a legitimate article. I think it would help prevent frustration on the part of users if a more helpful error message was created rather than just the default error message; would someone who is an administrator be able to do this? —Soap— 16:42, 29 April 2010 (UTC)

For the time being, I have limited the filter to only edits, not creations. Tim1357 ^talk 20:14, 29 April 2010 (UTC)

Blocks of HTML escaping

I thought we used to have a filter that blocked edits like this from being made. Does anyone know where it went? I can't seem to find it, so maybe I'm just remembering it wrong. Regardless, thoughts on handling this? We probably could block all codes for the letters A-Z as I can't think of any reason they would ever need to be put in an article without typing the letter directly. (Other codes, obviously, are more likely to be used and would have to be let through freely.) --Shirik (Questions or Comments?) 14:46, 30 April 2010 (UTC)

I can not remember such a filter. Ruslik_Zero 15:35, 30 April 2010 (UTC)

Filter 58

Can anyone see why it didnt catch the recent vandalism to Nishkid64's talk page? I can't see any 'loopholes' in the code that could be exploited to prevent the "disallow" event. —Soap— 11:43, 8 May 2010 (UTC)

I don't get it; the filter tests OK (matches) on those edits, and we're at 0% of the condition limit – something as early as 58 should certainly not be failing out. Anyone else have an idea? --Shirik (Questions or Comments?) 18:51, 8 May 2010 (UTC)

I'm even more confused. The filter stopped the same exact text on Tiptoety's page. Elockid ^(Talk) 14:35, 9 May 2010 (UTC)

Filter 260 and the "your mom" vandalism

Does anyone object to restoring the 'your mom' filter as a separate filter, un-merging it from 260? There are false positives coming in, and I have been adding exceptions for them, but if I understand correctly it would save us processing time if they were run on their own filter rather than being merged with 260, since those false positives are not relevant to the rest of the conditions (and could even produce false negatives by letting vandalism through). See here particularly the last two changes to #260 for examples of what I mean. (Actually, the only exceptions on that filter anywhere are from me). Basically I think it should be separate because the other watchwords have essentially no need for false positives, and I believe that "your mom" does have false positives that are worth accounting for. —Soap— 12:22, 8 May 2010 (UTC)

Note that, because Im going to be busy this summer, I plan on resigning from the edit filter soon and don't want to make a change myself that I won't be able to un-make if it causes problems or simply is a bad idea. —Soap— 18:46, 8 May 2010 (UTC)

We seem to have the conditions under control now; I have no objections to making it a separate filter. --Shirik (Questions or Comments?) 18:47, 8 May 2010 (UTC)

I moved it back to filter 320, since it popped up a false positive I cannot believe I didn't forsee, and will probably be somewhat common. Someguy1221 (talk) 07:42, 12 May 2010 (UTC)

"Pie" filter

Can we create something to that matter? Not sure why but there is a "pie" vandal who inserts "pie" or "i like pie" into random articles. After seeing it so much, I finally began keeping track of all the IPs and usernames who put it in WP for a future SPI but I thought an edit filter might be useful too. Tommy2010 12:06, 12 May 2010 (UTC)

• Already added to filter 260. NawlinWiki (talk) 12:36, 12 May 2010 (UTC)

  It's an Internet meme, so probably not the same person. I've been thinking about a filter for this myself but never got around to actually doing it. —Soap— 12:52, 12 May 2010 (UTC)

What's it mean? Tommy2010 13:13, 12 May 2010 (UTC)

Probably nothing. —Soap— 16:35, 12 May 2010 (UTC)

I already did it. I added something to 260 to match "I like pie" and "I like cheese". Tim1357 ^talk 21:31, 12 May 2010 (UTC)

Pre-emptive warnings

I was thinking about warning User talk:70.72.31.181 which triggered the edit filter about 4X in a row, but so far his/her actions have not been saved into Wikipedia. What kind of actions are standard and is there a policy regarding this? Tommy2010 23:33, 12 May 2010 (UTC)

I don't believe there's a policy. I sometimes use Twinkle to warn them, and then put a link to the filter log in the Comments section in case anyone thinks Im warning them for nothing. Some of the more severe filters are reported directly to AIV, and a user can even be blocked for triggering the filter if the filter they're triggering is one that is designed to detect specific socks of already-banned users. —Soap— 00:29, 13 May 2010 (UTC)

Ahhhh, that's smart. Awesome. Tommy2010 00:34, 13 May 2010 (UTC)

Filter 58, 5/12

• My error. Fixed. Lasted 6 minutes. I shouldn't try to edit when I've just woken up. Already revoked my own EF rights. NawlinWiki (talk) 11:17, 12 May 2010 (UTC)

Lol, it's ok. Was that where the !!* thing came from? Tim1357 ^talk 20:24, 13 May 2010 (UTC)

Requesting abusefilter rights

I was also wondering if I could get abusefilter rights here and help out with this part of the project. Thanks, Tommy2010 14:43, 12 May 2010 (UTC)

I think we need all the help we can get, however if approved I would urge you not to make big changes without consensus. About a week ago I remember you saying that a certain filter shouldn't trigger if the user is autoconfirmed; perhaps that filter really does need to be changed, but I want to be sure that you'd be weighing costs and benefits before making a change like that. —Soap— 16:35, 12 May 2010 (UTC)

Yes that was just an idea and would not make a big change without a discussion Tommy2010 17:16, 12 May 2010 (UTC)

The abuse filter is a powerful and dangerous tool. It has, on several occasions, blocked all editing from editors... and worse. May I ask what qualifications you have that suggest that you can be trusted with this tool? --Shirik (Questions or Comments?) 02:37, 14 May 2010 (UTC)

I apologize for the late reply. After your comments, I will withdrawal my request. I'd rather gain more experience than risk making mistakes. However, you can expect me to participate more in this area of the project as I primarily deal with vandalism. Thanks, Tommy2010 15:17, 14 May 2010 (UTC)

My statement wasn't an effort to get you to back down, however it wasn't a lie either – the abuse filter really is a dangerous tool, and that is why even admins don't, by default, have the capability to edit filters (though they can give themselves the right if they choose to). My suggestion if you aren't confident of your abilities right now is to look over the non-private filters we already have in place and perhaps try out the abusefilter on an installation of your own first. That will give you some experience into how to develop filters and just the kind of power it has. --Shirik (Questions or Comments?) 15:25, 14 May 2010 (UTC)

The AbuseFilter has safeguards built in to prevent precisely the situation -- a filter blocking all edits -- that you claim has happened. Can you point to a specific example? And what can this installation do which is "worse" than blocking all edits? Again, can you point to examples of where this has happened? I agree that the filter is a powerful feature which needs a great deal of respect, but I don't think this level of hyperbole is useful. Happy‑melon 16:04, 14 May 2010 (UTC)

Shirik is probably right here, my only experience is that I love using huggle, although lately I have been checking the edit filter log for vandalism and warning them based on that. I assumed that one needs abusefilter rights in order to write a filter. Tommy2010 16:08, 14 May 2010 (UTC)

I think what Shirik is saying that it has blocked all edits in the past, before the fix was added in. And I believe that even now, that fix only triggers after a hyperactive filter has been running for several minutes, so if it really was blocking 100% of the edits coming in, even for just a few minutes, that would be quite a lot of edits (probably in the hundreds). —Soap— 16:28, 14 May 2010 (UTC)

The abusefilter has blocked all edits before, (and still can). It can only do it until 5% of its current 'sample' (a few thousand edits) trip the filter. So that could be a few hundred edits blocked before the filter switches off. Prodego ^talk 16:31, 14 May 2010 (UTC)

Echoing what Prodego said, generally this is detected within minutes of the problem, so I've seen cases where we've reacted faster than the filter did. I'd rather not point to any particular instance as we'd all like to forget it ever happened, but suffice to say I've seen it happen. And there have been cases worse than blocking all edits – one case actually removed "autoconfirmed" status from all that made an edit. --Shirik (Questions or Comments?) 20:07, 14 May 2010 (UTC)

Filter 316

Filter 316 needs some improvements. The recent Zeals did not trip the filter like Xealking and Xealking2 but have the same MO. Elockid ^(Talk) 19:00, 12 May 2010 (UTC)

These changes are not really subtle, perhaps a new filter? Elockid ^(Talk) 20:11, 12 May 2010 (UTC)

Yeah I've been thinking about this one for a while too. I really wish for some kind of preg_replace functionality that would be able to help, but even if it were to exist (which it doesn't), it would be incredibly expensive. We may just have to revert to whack-a-mole. In the meantime, I know we've still caught quite a few Zeal socks with 316 so I don't think it's a good idea to disable it yet. I'm open to ideas on how to improve it, though. --Shirik (Questions or Comments?) 02:39, 14 May 2010 (UTC)

Just pasting a few more socks: Redzest (talk · contribs), Yellowzest (talk · contribs), Xealking8 (talk · contribs) (and other numbers 1 to 9), Ferventxeal (talk · contribs). —Soap— 17:41, 15 May 2010 (UTC)

Well in the meantime, is it possible to come up with something to detect the (far from subtle) naming pattern? HJ Mitchell | Penny for your thoughts? 21:46, 16 May 2010 (UTC)

Filter 58

Filter 58 seems to be producing a lot of false positives which is quite concerning considering it's flagging up seemingly innocent edits such as this one and disallowing as well as prompting a bot report to AIV. I'd appreciate it if someone could look into it and try to tighten it up a bit. Thanks, HJ Mitchell | Penny for your thoughts? 16:52, 15 May 2010 (UTC)

Prodego disabled it, but I re-enabled it as log-only. I'll look at it later today when I have the time, but since most of the people who hit this should be blocked on site, I think it should at least be logging the edits. Someguy1221 (talk) 21:15, 15 May 2010 (UTC)

Ok. Prodego ^talk 00:30, 16 May 2010 (UTC)

Requesting Edit Filter Manager Permissions

I have a bit of experience with regex and I know how edit filters work here. I'm looking to improve and shore up false positives. Also I'm on vandalism patrol at times and notice how things could be improved. Let me know if I could help out here. --Sidonuke (talk :: contribs) 13:58, 13 May 2010 (UTC)

No objections here; I've worked with Sidonuke in other parts of the project; but I can't speak to qualifications. --Shirik (Questions or Comments?) 02:38, 14 May 2010 (UTC)

Thanks. In terms of qualifications. I have nothing to show to be honest but if you are all willing, I'll be grateful to show you that I'm capable. Sidonuke 09:35, 16 May 2010 (UTC)

If no one else disagrees. Could permissions be granted? Sidonuke 10:42, 20 May 2010 (UTC)

Can you say something about this. While it happened more than three years ago, it still worries me. Ruslik_Zero 11:42, 20 May 2010 (UTC)

I was acting like a little kid back then not knowing what I was doing. I could of made a new account to hide my past but I decided against it because it was a mistake and I wanted to show that I have changed since then. Sidonuke 13:41, 20 May 2010 (UTC)

Special:AbuseFilter/332 is a filter I helped Jéské_Couriano create. It's a simple one. Sidonuke 06:01, 22 May 2010 (UTC)

I wish to make a comment. I did not know abuselogs could NOT be OSed. Sidonuke 06:52, 22 May 2010 (UTC)

I noticed that back in November, and I've been told the developers are working on it. See here (the same page I was linked to when I asked about it). —Soap— 22:29, 22 May 2010 (UTC)

Seems like its been ignored. Sidonuke 14:45, 23 May 2010 (UTC)

N-word not trapped

Any idea why this edit wasn't picked up? I've just found and reverted it. I suppose two days on is too late to put a warning on the user talk page. --Redrose64 (talk) 17:17, 19 May 2010 (UTC)

The N-word potentially has over 2,000 legitimate uses in Wikipedia (see [4]). Remember that before the majority of non-slaves considered the word offensive, it was widely used in literature and other writing, so it appears legitimately in articles on older enclopedic topics. Thanks to Lee Atwater, even tandem repeats have legitimate appearances. Someguy1221 (talk) 05:37, 20 May 2010 (UTC)

I thought we had, or used to have, a filter that would block out repeated uses of the word, but perhaps it only fires when it's capitalized. —Soap— 12:34, 20 May 2010 (UTC)

We do have one. In fact, one that you've edited! Special:AbuseFilter/225. It's probably true that there are no legitimate reasons to type the n-word in all caps. Someguy1221 (talk) 21:34, 20 May 2010 (UTC)

Filter 58

I noticed this filter caught Zealking. Any chance we can enhance it to catch more of these? Elockid ^(Talk) 01:59, 22 May 2010 (UTC)

A recent attempt to do this resulted in blocking all edits for a short period of time. I don't really find value in trying to block his edits as they're easily detected and blocked by hand, whereas attempts to block his edits from a filter will likely just cause him to come up with a new pattern. If anyone else wishes to go after this, be my guest, but keep in mind there are a lot more valid uses of exclamation points than you might think. --Shirik (Questions or Comments?) 22:08, 22 May 2010 (UTC)

I knew I'd seen this type of edit before - see MayorEnthusiam (talk · contribs). Sock maybe? --Redrose64 (talk) 10:14, 23 May 2010 (UTC)

@Shirik: Easily detected, but at least a quarter to half of them are not even caught by recent changes patrollers which I think is alarming. This sock is a prime example: Greyzest (talk · contribs) of how they're circumventing detection. It wasn't even caught by anyone else. Luckily one my earlier blocks seemed to have autoblocked them before this account finally got blocked. I haven't seen any filters catch anything recently except this one either.

@Redrose: definitely Zealking. Elockid ^(Talk) 10:59, 23 May 2010 (UTC)

False positives

Could someone else look at Wikipedia:Edit_filter/False_positives/Reports#76.102.27.141? I'm just not getting why this is even being caught by the filter. Someguy1221 (talk) 06:18, 26 May 2010 (UTC)

Well, it's a private filter (Special:AbuseFilter/242), so I can't see the code, but one thing I do know is that throttle filters often are set up to treat all IP's as the same user, so if several different IP's create different redirects on completley unrelated pages, the throttle might be triggered even so. —Soap— 01:28, 29 May 2010 (UTC)

Thanks for restoring the edit filter access to my account. To me it looks like this filter is working properly; the IP who filed the report had tried to make more redirects than is allowed by the filter. So the filter was doing what it's supposed to; whether it's too strict or not is another issue worth debating, but there doesn't seem to be any bug in the code. —Soap— 11:00, 29 May 2010 (UTC)

What I didn't understand was the first line of code. As I understand it, this IP shouldn't be caught by the filter. Someguy1221 (talk) 23:58, 29 May 2010 (UTC)

I think that all five of the "user_ " variables are valid for IP addresses, they're just set to zero or blank values (with the possible exception of user_name; I think that behaves like normal but am not 100% sure). So a filter that checks edit count, age of account, etc. will still trigger on an IP address, it will just see those things as 0. —Soap— 00:14, 30 May 2010 (UTC)

Edit Filter - hide patrolled recent changes script

Fellow recent change patrollers, can I request your feedback on a script I created (adapated) for those of you who monitor recent changes with tags (like repeating characters or -BLP). The idea was that when looking at pages like this one you will end up seeing any edits which triggered the EF tag, even if that edit is no longer the most recent edit. In many cases patrollers or other editors have already corrected the issue. This script will show only edits which still have a rollback tag visible, and there is a portlet tab to turn the function on and off.

importScript('User:7/hidepatrolled.js');

Granted, a vandal could make 1 edit which triggered the filter and then make a second which did not trigger the filter to hide their tracks... but I'm not sure that will happen to often. Thanks for your input.  7  11:29, 4 June 2010 (UTC)

Test edits

Do we have an edit filter for test pages? Someone created the page Fixedia with the word test but it didn't trip any wires. Marcus Qwertius (talk) 15:58, 7 June 2010 (UTC)

I don't think we've ever tagged the word "test", although there is a filter, 98 , that triggers when someone creates a very short new page, and this user did trigger that filter. It just isn't set to disallow, or even to "warn" anymore, apparently because once in a while a user will create a page with content like "Under construction" and then a minute later go on to fill it with useful content. If it were set to disallow, they might get confused or discouraged. There are also a few articles like this one that are tiny even when they're complete. Also, there was once a filter that tracked test edits in general, but it too was disabled because sometimes a user would add content and test edits (like clicking the toolbar) in the same edit. —Soap— 19:34, 7 June 2010 (UTC)

Speaking of test edits... :) Filter 18 was the one set to watch for test edits. From what I gather, it was turned off for two reasons: Performance, and it was seen as bitey. (I've discussed this before.) I am wondering if the performance issue has/can be addressed? The bitey concern can be easily fixed by removing any warnings and merely logging the edits. This would allow the user to save the edit and be totally transparent to them, while allowing editors to view the log and remove any unwanted tests. Just a thought. ^Avicenna_sis @ 16:29, 9 June 2010 (UTC)

That would be great if it could be brought back. Marcus Qwertyus (talk) 16:45, 9 June 2010 (UTC)

Gay

Is there a filter that should catch the insertion of the word "gay", such as this piece of vandalism? Mjroots (talk) 07:46, 11 June 2010 (UTC)

It'd cause too many F+'s, I believe, from people adding quotes or noting the subject's orientation. —Jeremy ^{(v^_^v Dittobori)} 07:52, 11 June 2010 (UTC)

My tests

{{editprotected}} I was testing ClueBots filters, but I'm not getting a response from his owner...

Here is what I entered that was 1) not removed by ClueBot and 2) didn't trip a filter: 8==> (wang, ding-d0ng, c0ck) $luty H03 B!tch d!ck head cunt fag milf dumb @ss ho3 turd chink p0rn $ex pu$$y fagot puta h4rdc0re $uck !t

Please note, I'm not a vandal, I just wanted to mention these so that they can be fixed, see my original post at: http://en.wiki.x.io/wiki/User:ClueBot/Sandbox#Testing_.28Using_multiple_variants.29

You might want to use {{editprotected}} if you want to add these to the blacklist. Or someone with abusefilter rights can add them to whatever filter. —fetch·comms 19:59, 11 June 2010 (UTC)

Most of those words are on the edit filter in various places, but aren't set to "disallow" because they could be used in constructive ways. But rest assured that we aren't ignoring the problem. I think the 8==> may have once been entirely disallowed, though. Also, I think that while this request isn't "controversial" in the sense that most content edits are, it can't be done immediately either, and so I'm going to unlink the {{editprotected}} template again ... and, anyway, this page is watchlisted by many editors who work with the edit filter, who wouldnt need the template to be able to see this. —Soap— 20:13, 11 June 2010 (UTC)

Removing Copyvio tags

Someone removed the copyvio speedy tag from Keen "Mecca" Cotter but only a (tag:references removed) appeared. Can only one tag be displayed? If so it would seem the copyvio would take precedence. Marcus Qwertyus (talk) 20:05, 13 June 2010 (UTC)

Only one tag will be displayed, yes, but the edit filter still shows them all. I'm not really sure what determines which tag will appear next to the edit if more than one tagged filter is triggered. —Soap— 20:09, 13 June 2010 (UTC)

Part two, when someone removes a hang-on tag that trips the filter as well? Marcus Qwertyus (talk) 20:44, 13 June 2010 (UTC)

Well, {{hangon}} is considered a speedy deletion tag, probably because the only pages with it are those that previously had one of the other tags; so the filter is written in such a way that removing {{hangon}} also triggers the "removing speedy deletion templates" filter, and its associated tag. —Soap— 20:54, 13 June 2010 (UTC)

Filter 313

Please, exclude bots from this filter, if possible. Sole Soul (talk) 20:38, 14 June 2010 (UTC)

 Done (though I'm surprised ClueBot wouldn't edit through a warning). —Soap— 20:53, 14 June 2010 (UTC)

Filter 135

Special:AbuseFilter/135

I really appreciate this filter and understand the necessity of its existence, but however I would like to learn: why does it mark edits like this one with "repeating characters" tag? The corresponing changes have nothing in common with adding hahaha or whatever of this kind. If there is a drawback in the code, please fix it, as it's indeed annoying while looking through marked edits in order to detect vandalism. --Microcell (talk) 16:47, 29 May 2010 (UTC)

I believe it's the presence of the "1000000000000" towards the end of the line. --Redrose64 (talk) 16:54, 29 May 2010 (UTC)

Me too. :) And is there a possibility to sift out such cases to exclude from the filter (e.g. to look strictly at changes, but not at the wikitext)?--Microcell (talk) 18:04, 29 May 2010 (UTC)

Here's another similar one. Edit was just adding a wikilink, but the line already contained 1000000000. WP:Popups highlights only the added characters, so it must be possible. Qwfp (talk) 09:48, 15 June 2010 (UTC)

It really isn't possible to get the edit filter to see just what's been added, because it isn't "intelligent" the way popups is. And if it were, it would be a lot more time-consuming. However, this has been brought up before and I think it wouldn't be a bad idea to add a check to the filter that stops it from tagging an edit if the areas of text being edited already had the repeating characters in it before the edit. —Soap— 17:54, 17 June 2010 (UTC)

That sounds like a good idea to me. Qwfp (talk) 18:04, 17 June 2010 (UTC)

Of course the filter needs to test removed_lines against the same regexp. I have a similar "repeating chars" filter set to "disallow" in another WM project and it works beautifully. — AlexSm 02:33, 18 June 2010 (UTC)

Filter actions

I'm evaluating Extension:AbuseFilter for use on our internal wiki, and trying to learn how it all works. When configuring a filter, under Actions taken when matched, I see a Flag the edit in the abuse log checkbox, but it's disabled (greyed out) so I can't click on it to clear the checkbox, and I can't figure out how to enable it. Is there any way to enable this option so I can clear the checkbox and prevent posting a triggered event to the Abuse log? Or is it designed to always log, and then optionally do one or more other actions?

--obliquemotion (talk) 15:06, 16 June 2010 (UTC)

All I can say is that it appears that way for us too. I think a higher level of access than just being an administrator is necessary to change a setting like that; if it's possible at all, you'd probably have to do it in the site-level installation/configuration of the extension. —Soap— 20:41, 16 June 2010 (UTC)

I always interpreted this checkbox as just a reminder that triggering will be available in logs. — AlexSm 21:10, 16 June 2010 (UTC)

Thank you both, that's a big help. --obliquemotion (talk) 17:31, 17 June 2010 (UTC)

Issue with "removing references" filter

== FYI ==

http://en.wiki.x.io/w/index.php?title=MediaWiki_talk:Tag-references_removed-description&oldid=368173192. Happy‑melon 17:56, 20 June 2010 (UTC)

This is Special:AbuseFilter/61; section title clarified. — AlexSm 03:13, 21 June 2010 (UTC)

Proposal: Duplicate the MediaWiki AF notices to uw templates

Recent changes patrollers who revert solely on the basis of abuse filters would like their warnings to include a message about the tags being automatic and link to a place to report false positives like ClueBot does. I am one of them, and I don't want to blame them, but rather the abuse filter.

A way to do this is to copy the MW AF notices to Template:uw locations. It would be of great help to us. T3h 1337 b0y 22:50, 21 June 2010 (UTC)

No one should be reverting anything solely on the basis of abuse filters. If the edit was allowed, that means human review is required. If it should be prevented altogether the AF would have disallowed the edit. Prodego ^talk 23:05, 21 June 2010 (UTC)

User:SvickBOT/Edit filter effectiveness

Per my request at Wikipedia talk:Database reports, User:Svick generated a report on the effectiveness of warning filters. Sole Soul (talk) 21:35, 22 June 2010 (UTC)

Hey, those are pretty good figures. Higher than I expected, for sure. It's not surprising that the anti-spam ones have low prevention rates; people who come here to advertise aren't going to be deterrred by warnings and usually aren't regular editors anyway. With regards to the "mistake detection" ones, perhaps better warning templates might help; after all these are good faith users, not vandals. Perhaps the people tripping the Skype filter just aren't able to find the strings in the text and so they click Submit anyway. —Soap— 22:51, 22 June 2010 (UTC)

I agree with you on the anti-spam filter, but I add 2 things: 1) it is a throttle filter, and when you do something many times you become bold. 2) the warning template language is somewhat weak "For some kinds of links this may be okay, but it is often a sign of people abusing Wikipedia. If that is not your intent, we apologize". Most spammers think their links are useful "not like other spammers", so talking about intent is misplaced, IMO. Sole Soul (talk) 23:04, 22 June 2010 (UTC)

CIDR checking

I was all geared up to experiment with making my very first edit filter, but got hung up on an issue that I thought was built in. I wanted to check for an IP in the range 86.96.224.0/19 adding "editsemiprotected" to a talk page, but I can't see how to check if the IP is in that CIDR. For those that are curious, WP:Sockpuppet investigations/Brexx should provide sufficient explanation of my motivation. I know it won't work when he's logged in, but he does a remarkable amount of it with his IP address exposed.—Kww(talk) 14:39, 24 June 2010 (UTC)

...ip_in_range(user_name, '86.96.224.0/19')... -- zzuuzz ^(talk) 14:51, 24 June 2010 (UTC)

Just wanted to have an idea

...whether this article creation would have given a tag to np patrollers... Thanks. ♪ ♫ Wifione ♫ ♪ _―Œ ^{♣Łeave Ξ мessage♣} 04:33, 23 June 2010 (UTC)

Although I can't say with 100% certainty, I don't believe it would have. The edit filter log shows no history for this pagename, and those logs should remain even after a page has been deleted. (example) ^Avicenna_sis @ 20:27, 24 June 2010 (UTC)

Alright, thanks. My question was actually an issue on why the new page did not have a filter tagging it... Thanks anyway. ♪ ♫ Wifione ♫ ♪ _―Œ ^{♣Łeave Ξ мessage♣} 13:06, 26 June 2010 (UTC)

Access to editfilter

I have some basic programming experience, and I think that working with the edit filter would be interesting. I think I'll spend some more time trying to become acquainted with how it functions before I do anything, however. My question is will I have to request the ability to edit it, or can I simply grant myself the right when I feel I'm ready? NativeForeigner ^Talk/_Contribs 23:18, 24 June 2010 (UTC)

Admins are considered trustworthy enough to judge for their own sake when they're ready to hold the abusefilter user right (and can remove it from themselves as well). Note though that admins can also read all of the code of the private filters without that userright, because part of the admin package includes the ability to view private filters. —Soap— 23:22, 24 June 2010 (UTC)

OK, thanks. I think I'll start watching what's going on here to learn the ropes. NativeForeigner ^Talk/_Contribs 05:13, 25 June 2010 (UTC)

Edit blocked when two "warn" filters trigger

Still evaluating the Extension:AbuseFilter extension, and I have one more question. I have two filters defined:

• Filter 1 ("orange"): article_namespace == 0 & "orange" in lcase(added_lines)
• Filter 2 ("green"): article_namespace == 0 & "green" in lcase(added_lines)

Both are configured ONLY to warn and tag. If I make an edit that triggers either independently (for example, add a line with the word "green" in it), AbuseFilter behaves as I expect (warns, then when I click Save a second time, it tags it in the edit history).

However, if I add lines that would trigger both filters, only the filter #1 warning displays, and prevents the edit altogether. For example, here is the attempted two-line edit:

• The grass is green
• I love oranges

When I click Save, the filter #1 warning displays above the edit box, and the log shows:

• 15:12, June 25, 2010: Obliquemotion (Talk | contribs | block) triggered filter 1, performing the action "edit" on Sandbox 2. Actions taken: Warn; Filter description: orange (details) (examine)
• 15:12, June 25, 2010: Obliquemotion (Talk | contribs | block) triggered filter 2, performing the action "edit" on Sandbox 2. Actions taken: tag; Filter description: green (details) (examine)

Then, if I click Save again (to override the warning), I get the filter #1 warning again, and the log shows:

• 15:12, June 25, 2010: Obliquemotion (Talk | contribs | block) triggered filter 1, performing the action "edit" on Sandbox 2. Actions taken: Warn; Filter description: orange (details) (examine)
• 15:12, June 25, 2010: Obliquemotion (Talk | contribs | block) triggered filter 2, performing the action "edit" on Sandbox 2. Actions taken: tag; Filter description: green (details) (examine)
• 15:12, June 25, 2010: Obliquemotion (Talk | contribs | block) triggered filter 1, performing the action "edit" on Sandbox 2. Actions taken: Warn; Filter description: orange (details) (examine)
• 15:12, June 25, 2010: Obliquemotion (Talk | contribs | block) triggered filter 2, performing the action "edit" on Sandbox 2. Actions taken: tag; Filter description: green (details) (examine)

It seems like each time I click Save, AbuseFilter adds the same two log entries, but prevents the edit (even though I did not enable the Prevent the user from performing the action in question checkbox on either filter).

1. Does the same thing happen in Wikipedia's Edit Filter? Can you explain this behavior? Is it a bug, or have I misconfigured something?
2. Do the filters always run in the order of their filter IDs (e.g. filter #1 runs before #2), and if so, how do I resequence them?

(BTW, I'm running the 1.15.x snapshot of the extension on a 1.15 MediaWiki server.)

Thanks, --obliquemotion (talk) 21:06, 25 June 2010 (UTC)

This is an old, old bug and I don't think there's anyone who doesn't want it to be fixed. That's all I know. I believe that the filters always run in numerical order, which is why it might be a good idea to create a few small, or even empty, filters at the beginning, which can be filled in with important code when necessary. —Soap— 22:13, 25 June 2010 (UTC)

Yep, they always run in that order, but if you want to resequence some, either create a new one and put the filter you want to run later in that one, and disable the original, or switch the filter code around. The first way might be better, as it leaves a space for later use. Ale_Jrb^talk 11:31, 26 June 2010 (UTC)

Thank you both for speedy replies! So if two "warn" edit filters trigger on Wikipedia, the edit is completely blocked? --obliquemotion (talk) 18:34, 28 June 2010 (UTC)

A side note - but don't resequence the filters on Wikipedia. Once they have had a single hit, that hit is logged pointing to the filter that was tripped. If what the filter that was tripped does was changed, the log of abuse hits gets very, very confusing. Prodego ^talk 21:30, 28 June 2010 (UTC)

article_articleid

The variable "article_articleid" can be used to transform any page title into a number unique to that page. I got a false positive hit that was triggered on a certain article_id which only contained the number of the page that was watched for. Is there a way to work backwards from the article_id to find the name of the article it refers to? —Soap— 22:34, 28 June 2010 (UTC)

Do you mean like this :D Sole Soul (talk) 22:39, 28 June 2010 (UTC)

thank you. I actually found the page I wanted for this particular instance by Googling, but that method will be helpful in the future. —Soap— 22:43, 28 June 2010 (UTC)

Filter 342

This fiter is so consuming. I suggest putting "article_namespace ==0" as the first condition, followed by "Old page size" condition. Sole Soul (talk) 14:01, 13 July 2010 (UTC)

 Done by Tim1357; thanks Tim. —Soap— 22:25, 13 July 2010 (UTC)

Is it intentional for the current version of the filter to mix old_wikitext and added_lines to catch an article inserted below the redirect line but not removing that line? Seems like that might also catch some redirect tagging (e.g. adding {{R to section}} to an existing redirect on a new line). PleaseStand ^(talk) 01:22, 15 July 2010 (UTC)

Edit filter bug?

So, I'm looking at this list of filter hits, and I'm confused. The edits to David Lynch (the merits of the edits themselves aside) seem to be being blocked completely, even though the relevant filters are set to only tag/warn. Wasn't a similar bug recently fixed? Someguy1221 (talk) 09:05, 19 July 2010 (UTC)

I dont believe that bug has been fixed. Im not sure it's even considered a bug ... the only other mention of it I've ever seen is in this False Positive report here. —Soap— 09:28, 19 July 2010 (UTC)

Unless I'm overlooking something, there is no bug.

• At minute 06, the user was warned and he/she did not proceed to save
• At 07 the same thing, again, the user was warned but did not save
• At 08, also the same happened
• At 09, also the same happened, but in addition the "Repeated vandalism attempts" filter was triggered

I don't see 2 edits at the same minute, or a warning followed by a tag of the same filter. Most likely the user did not save after warnings. Sole Soul (talk) 11:20, 19 July 2010 (UTC)

Edit filter bug? (a different one!)

When examining this edit against filter 346, no match is found. However, looking at the filter log for 346 does show a positive hit for that edit... and I'm just confused. Someguy1221 (talk) 07:55, 26 July 2010 (UTC)

Filter 233

I have looked through the recent (last 4 days) edits stoppped by this filter, and they all appear to be false positives. Almost all of them are questions about speedy deletions done by NawlinWiki; even the other 2 appear to be in good faith - one by an anon asking about a message on their IP talk page, referring to an edit they never did; and a user, havingf left a message while logged in, editing the sig after logging out. I think thisd filter should be either disabled or refined. עוד מישהו Od Mishehu 17:18, 12 August 2010 (UTC)

• Not really working well, will disable it. NawlinWiki (talk) 18:13, 12 August 2010 (UTC)

Two naive questions

I am newbie here. (i) Filter logs with a "wide" result window (wider than a usual WP page) freeze my PCs from some 10 to >30 seconds (that makes me hate Mr.Z-bot; Technical details: WP XP or Vista, Firefox and Java are recent; I use old wikipedia interface) (ii) This might be not for this talk, I guess the "possible libel or vandalism" tag is triggered by the word "sacked" which is pretty innocent. This spams my RCP backlogs with such "sacked" edits which are non-revertable. Materialscientist (talk) 07:01, 15 August 2010 (UTC)

Why are some filters private and others public?

Or more to the point, why are some filters private? Sorry if this debate has been had already - but I strongly don't agree with the idea that hiding hte methods used to detect vandalism strenghtens the system. This is known as Security Thruogh Obscurity, and runs counter to many of the principles of the open-source world - not least Wikipedia.

And on a realted note, if it's policy to have filters as private, why are lots of them available for public viewing? Thanks, --Christopher (talk) 17:19, 15 August 2010 (UTC)

They're private because if the code were visible, they would be easier to get around. So essentially, yes, hiding the methods to detect vandalism strengthens the system. Are you saying that it doesnt or just that you think it isnt worth it? —Soap— 17:37, 15 August 2010 (UTC)

I think filter editors should regularly check whether the currently hidden filters need to be private or not. Sole Soul (talk) 17:48, 15 August 2010 (UTC)

A number of vandals employ very specific terms or edit specific topics in particular ways. Since the private edit filters exist to deter such attacks, revealing the source code would defeat the point of the filters. We do not need to make it easy for disruptive editors. The public filters mostly deal with obvious and common vandalism modes that shouldn't surprise anyone, like "poop" vandalism and such. The private filters are visible to those with trusted access, such as admins and edit filter managers (not all of whom are admins). We don't allow free access to deleted content to one and all, nor do we permit everyone to see how they can circumvent anti-vandal/disruption methods. In general, all filters should be (and mostly are) checked to see if they are still needed at all, or if the parameters should be changed. Acroterion _(talk) 17:59, 15 August 2010 (UTC)

Thanks all for the reasponses. I don't have any 'genuine' reason for wanting to see the private filters - just curiosity really. I can see why certain filters would be better off unviewable though! --Christopher (talk) 09:31, 16 August 2010 (UTC)

"Page blanking" tag

A proposed edit filter which tags an edit with (Tag: page blanking) if the edit deletes more than 7500 characters and leaves 0-499 characters inclusive. Such edits will most often be bad ones but should not necessarily be set to disallow. This filter should not be used in the Project or User space as sandboxes may be found there. Otherwise, it may be used anywhere.

I mentioned this specifically after discovering that filter 3 now has a set minimum number of characters remaining to apply the regular "blanking" tag. The proposed one will include those that go all the way to zero, which filter 3 now excludes. mechamind90 00:39, 17 August 2010 (UTC)

Help a sister project out?

English Wikibooks just enabled the AbuseFilter extension not more than two days ago and I am bootstrapping it with filters from Wikipedia. I am an administrator there (verify). I am interested in the private filters 12, 34, 58, 260, 279, and 320. I will ensure that they remain private when enabled at Wikibooks. Would a kind soul be willing to email me the exported versions? Adrignola (talk) 02:13, 17 August 2010 (UTC)

I'm doing it now. עוד מישהו Od Mishehu 13:32, 17 August 2010 (UTC)

Filter 351 tag

It has been requested that Filter 351 tag the edits which it catches. The proposed tag name is "Text added after categories and interwiki" (same as the filter name), but in my opinion it seems rathewr long for a tag name. Any other opinions in the matter? עוד מישהו Od Mishehu 04:44, 18 August 2010 (UTC)

End of page text? ^Avic_ennasis @ 22:20, 10 Elul 5770 / 20 August 2010 (UTC)

Request for permissions archive

I've copied all the requests for permissions that were found in /Archive 3, /Archive 4, and /Archive 5 into Wikipedia talk:Edit filter/Archive (permission requests) with standardized headers similar to those above. Going forward, requests for permissions that have been addressed and have had time for subsequent comments can be manually archived into the new permissions archive. This should help give individuals initiating, commenting on, and closing, requests for permissions a conveniently-accessible frame of reference. –xeno^talk 15:12, 23 August 2010 (UTC)

These filters are really annoying

These filters are annoying and unhelpful. I am for example unable to remove vandalism if the vandalism contains reference tags and remove comments from banned users from talkpages. This is highly frustrating.·Maunus·ƛ· 10:17, 4 September 2010 (UTC)

Are you editing as an IP? Neither of those filters should affect admins, or even ordinary registered users for that matter. The only recent hits I see in your filter log are for the Skype plugin for Firefox, which really is a problem, and which is currently set to "disallow" even for administrators because if it weren't, you or someone else would have to be constantly cleaning up after the code that those edits leave in the page. For example, this edit. As for the others, it sounds like Filter 34 and Filter 61 are being triggered, which only affect IPs and users with low edit counts. This is partly because with an IP, there is no sure way to know whether a given edit is the same person as previous edits from the same IP or not, and partly because IPs and new accounts are often used to blank pages and vandalize. There are some users who work in anti-vandalism and prefer to remain as an IP and thus are sometimes stopped by those filters, but as before, it isn't possible for those users to be exempted because the software will not recognize them with each new edit as being the "same" person. —Soap— 10:22, 4 September 2010 (UTC)

If you havent been editing as an IP I would assume the filter hits are triggering entirely on the Skype filter, which can be solved by disabling the Skype plugin whenever editing Wikipedia. It would be nice if the MediaWiki software could simply strip out those strings before saving an edit, but that isn't possible right now. —Soap— 13:15, 4 September 2010 (UTC)

I don't have the skype extension enabled.·Maunus·ƛ· 13:01, 5 September 2010 (UTC)

Maunus, perhaps you could retry some of your changes from a different computer, or try a different browser. The Skype toolbar problem is explained at this link on a Skype forum. The filter is detecting that you are trying to add 'begin_of_the_skype_highlighting' in these edits. EdJohnston (talk) 13:25, 5 September 2010 (UTC)

I found the problem - it seems that it is not enough to remove the skype extension from the functions tab - it just restarted with the browser - I had to remove the skype extension from the applications altogether.·Maunus·ƛ· 13:30, 5 September 2010 (UTC)

Request for permission: User:Mono

Moved to Wikipedia talk:Edit filter/Archive (permission requests)#Request for permission: User:Mono.

There are 2 pages which should be exempt from the Skype filter

I think that there are 2 pages which should be exempt from the Skype filter (filter 313):

1. Wikipedia:Edit filter/False positives/Reports - mmaintaining that page with the occasional Skype junk shouldnb't be that big a problem; and allowing users to edit there even if Skype adds it may make it easier for them to see the problem.
2. MediaWiki:Abusefilter-warning-skype - this page, the warning for the Skype filter, has the text in it, and that is too likely to trigger false positives which prevent editing.

The solution is simple: add a condition that the page ID (article_articleid) isn't 26813105 or 26204397. These conditions would only be checked for edits which wolld currently be caught by the filter, and would be quick. עוד מישהו Od Mishehu 20:04, 6 September 2010 (UTC)

 Done —Soap— 10:11, 11 September 2010 (UTC)

About creating a filter to watch selected pages

• Wikipedia:Administrators' noticeboard/Incidents/CCI

It would be nice if the filter could check a list of pages (thousands in this case) or perhaps for the reversion of a certain template, to determine if a blanked page, when restored, has more than a 200b difference between the version before the blanking and the version after the blanking is "reverted".

The idea is to prevent someone from shrugging their shoulders at the copy vio problem and just restoring an article.

Is this possible?

It would be nice. Not only for this case, but for other copy vio situations. Especially when we start to loook at user contributions globally... - jc37 16:41, 10 September 2010 (UTC)

Sorry, that would be great, but it's impossible. The filter cannot see page histories, aside from the current revision and the one immediately before it. To watch a large list of pages, the pages would have to be in a particular category and I dont know if such a category exists for this particular purpose. Note that we do have filter 224 , which detects any removal of a copyvio template when made by a user with below a certain edit count. (To those who watch this page I know it seems like my answer is always "no", but I genereally only post when Im really sure that something is impossible. If I'm wrong I will do what I can to help create the filter.) —Soap— 22:47, 10 September 2010 (UTC)

AFAIK, they can be placed in a category.

And even if it just checks to see if the template is removed, that's at least "something". - jc37 00:55, 11 September 2010 (UTC)

I'll have a look at making a bot to do this, but I'm not making any promises ;) - Kingpin¹³ (talk) 05:26, 11 September 2010 (UTC)

Yeah, this would have to be done by a bot. The edit filter, as Soap said, cannot see the page history, only the revisions before and after an edit. - EdoDodo ^talk 06:20, 11 September 2010 (UTC)

Thanks for all the info : )

A quick question: Will the template used by Uncle G for this be covered by 224? and if not, can it be? - jc37 19:23, 16 September 2010 (UTC)

Extended it to catch removal of Uncle G's template. - EdoDodo ^talk 17:37, 20 September 2010 (UTC)

False negative on filter 29

Hello. I was bored and decided to browse through CAT:CSD, and I spotted an article of interest and clicked on it. Realizing that the reason given for speedy deletion wasn't valid, I removed the db tag. However, because the tag began {{db| instead of {{db-, filter 29 failed to catch and tag my edit as a new user removing a speedy deletion template. The filter should be tweaked to fix this, probably by simply changing it to not check for the hyphen. Thanks. 99.139.149.215 (talk) 19:10, 20 September 2010 (UTC)

Fixed. Thanks for letting us know. - EdoDodo ^talk 19:44, 20 September 2010 (UTC)

Long-term abuse edit filters

It appears that a lot of the long-term abuse vandals are being caught by a single filter (47). Would it be better to separate them into their own filters, or to keep everything in one filter? Some long-term abuse vandals also have their own filters, so nothing is consistent here. Separating them into their own filters might cause some performance issues, but I haven't tested it. So... all in one are separate? Netalarm^talk 04:44, 19 September 2010 (UTC)

I think that a general decision about all lon-term abusers is out of place. We need, for each oene, to consider:

1. How much resources would it take to separate it into a separate filter, relative to the work done by this filter?
2. To what degree would it complicate this filter to add the specific person's patterns into it?

עוד מישהו Od Mishehu 09:01, 19 September 2010 (UTC)

Hmm, the reason I brought this up is because there is no easy way to find which vandal is targeted by which filter. Having a centralized organization system would help solve this. Regarding the work needed, I think if we merged them into fewer or one filter it would be better. But the current way of organization is that some vandals have their own filters under their name, some have it under a not-obvious name, and some are lumped together. Netalarm^talk 04:22, 27 September 2010 (UTC)

user:Wallflowers98 edit that passed filters

Wikipedia:Long-term_abuse/Wallflowers98 tells me I should report any edits by Wallflowers98 socks that made it through the edit filters here. So: this edit made it through today. keɪɑtɪk flʌfi (talk) 19:55, 20 September 2010 (UTC)

Re-activated filter 306. Thanks, Shirik (Questions or Comments?) 16:23, 24 September 2010 (UTC)

Skype filter

Shouldn't this have warned the user here User talk:Spartiatis Leonidas and the IP who made these edits? Dougweller (talk) 05:38, 29 September 2010 (UTC)

It did warn him, yes. As of 25 September the filter is no longer set to disallow, so if the user chose to ignore the warning it would only have been tagged with 'Skype toolbar formatting' (which it was) but the edit would have been allowed. - EdoDodo ^talk

Why isn't it set to disallow? Am I right in thinking it doesn't put a warning on the editor's talk page? It's clearly causing work for other editors who have to clean up after it, and one user has been blocked because he keeps doing it (of course, if he'd responded to the warnings editors had been leaving him on his talk page he wouldn't have been blocked, but that doesn't solve the problem of IP editors using the Skype toolbar). And from what I've seen it the toolbar problem may not be fixed soon, but that's just based on this post on a Skype forum. Dougweller (talk) 06:33, 29 September 2010 (UTC)

I would agree with disallow - which is what I had in mind when I requested the EF many months ago.  7  06:42, 29 September 2010 (UTC)

I agree as well, but won't change it back until there's a bit more discussion since I don't want to get involved in a revert war. Also, the warning isn't put on the editor's talk page, it's displayed to him when he tries to click 'Save page' and (if disallow isn't enabled) he can then click 'Save page' again to ignore the warning and save it. - EdoDodo ^talk 11:27, 29 September 2010 (UTC)

I'm here because I ran into it at Recent Changes and left Doug a note. I'm no expert (Skype is a browser also? is that how it works?), but disallowing such edits, warning or not, seems reasonable to me. Drmies (talk) 14:09, 29 September 2010 (UTC)

I believe it's a plugin for Firefox, such that whenever it sees an 8 digit number anywhere, it turns it into a clickable link so that you can call someone with just 1 click. It does that by putting "begin_of_the_skÿpe_highlighting" before the number and "end_of_the_skÿpe_highlighting" after it. (misspelled on purpose) The bug is that it's doing that even when it sees 8 digit numbers in editable text boxes, and leaving it there when the user clicks Save. Furthermore it seems to keep doing it, at least in some cases, even if the user deactivates the Skype toolbar ... it has to be actually uninstalled. —Soap— 14:16, 29 September 2010 (UTC)

Ah ok. I'm glad that when I installed the most recent version of Skype I unchecked all the options--I also use Firefox. Thanks, Drmies (talk) 14:18, 29 September 2010 (UTC)

Filter 354

Wished to add a few more words to the filter to make the tagging more effective; can I mention them here for feedback first? Wifione _....... ^{Leave a message} 18:49, 29 September 2010 (UTC)

I'm beginning to think an internal mailing list would be the most effective method of communication. I myself have a few suggestions related to the long-term abuse filters that I don't believe would be a good idea posting in public. A mailing list would also help in this type of situation and would greatly improve internal collaboration. Proposing idea below. Netalarm^talk 02:15, 30 September 2010 (UTC)

Filter 342 and my bot

Would this filter be amended to avoid my bot User:AFC clerk bot? —I-20the highway 20:33, 16 September 2010 (UTC)

I suppose it would be reasonable to have the filter ignore all bot edits. New code would be:

article_namespace == 0 &
!"autoreviewer" in user_groups &
!"bot" in user_groups &
old_size < 200  &
removed_lines rlike "#[Rr][Ee][Dd][Ii][Rr][Ee][Cc][Tt]" &
!new_wikitext rlike "#[Rr][Ee][Dd][Ii][Rr][Ee][Cc][Tt]"

- EdoDodo ^talk 05:21, 17 September 2010 (UTC)

 Done —Soap— 17:00, 17 September 2010 (UTC)

This user has retired and the bot has been shut down. I do see one other bot running, though (User:Uncle G's major work 'bot), so I'm leaving the bot exception in unless someone disagrees. —Soap— 11:31, 2 October 2010 (UTC)

Miss by filter 81

Can anyone explain why filter 81 misses this edit?—Kww(talk) 23:57, 29 September 2010 (UTC)

bra[zs]il? You could tell Rich... Wifione _....... ^{Leave a message} 05:08, 2 October 2010 (UTC)

Doesn't that match "brazil" or "brasil"? That's what is intended.—Kww(talk) 06:00, 2 October 2010 (UTC)

Lets see:

• bra[sz]il(ian)?[^\w]+(hot[^\w]+100|single|album)
• hot[^\w]+100[^\w]+bra[zs]il

are the matches we are looking for.

And the line is

• |Brazil (Billboard Brazil)<ref>]http://www.hot100brasil.com/chtsinglesb.html/]</ref>

This fails to match because "hot100" does not match "hot[^\w]+100" - perhaps it should be "hot[^\w]*100"?

Rich Farmbrough, 07:15, 2 October 2010 (UTC).

And the other + should be * too.... I have updated the filter and tested it. Then updated it again. You may wish to re-test. Rich Farmbrough, 10:10, 2 October 2010 (UTC).

Thanks Rich :) Wifione _....... ^{Leave a message} 10:12, 2 October 2010 (UTC)

Thanks. For people watching, there's nothing wrong with Billboard Brasil, but people disguise references to an amateur chart by labeling them as being Billboard Brasil. It's probably a naming confusion more than active deceit.—Kww(talk) 15:01, 2 October 2010 (UTC)

Too many private filters

Far too many of the abuse filters in use are private. This does little except decrease transparency and make life more difficult for us.

Arguments about security are nonsense. All of the software that runs Wikipedia, from the OS up to MediaWiki and its extensions, is open-source, except for these filters.

Arguments that hiding the filters prevents vandals from knowing what they do in order to circumvent them are similarly unfounded. Almost all of these filters are directed at the sort of immature vandalism you would expect from bored kids. They do not know where the abuse filters are located, and even if they did, they don't know how to read them.

Not that it matters, when the intent and effect of a filter is obvious anyway from the revisions matched in its log. What value, exactly is there in hiding a filter like filter 323 ("reverting anti-vandal bot")? It is obvious what it does.

No long-term persistent vandal who might know where to look for these filters would be trying to make their edit stay by reverting an anti-vandal bot in the first place. And even if they were... you haven't achieved anything by making the filter private, because it's obvious what it does. So why bother, except to alienate those of us who can't read them?

There are dozens of "private" filters that do not need to be, and more that probably don't need to be but of course I can't tell because I can't see them.

The only losers here are the well-intentioned contributors who aren't able to look at these filters in the event that there is a problem with one of them, or suggest improvements, or perhaps improve their efficiency, or even just educate their writers on the finer points of regular expressions.

The "private" setting of abuse filters should be used sparingly, and solely in cases where public knowledge of the filter's content would be harmful -- for example, if it is looking for a particular username, or a libellous statement, or the edit pattern of a long-term persistent vandal.

Gurch (talk) 14:07, 27 September 2010 (UTC)

Looking at the specific example provided, I also don't know why it's been set private: it's not even a blocking filter. –xeno^talk 14:13, 27 September 2010 (UTC)

I agree. Unless a filter was designed to combat a specific, long-term vandal or group of vandals, all filters should be public (at least to autoconfirmed users). If a vandal know how, or bother to, look for codes he probably cannot be prevented by filters. Filters are not supposed to catch all of vandalism attempts, we are lucky if they prevented most of them. Sole Soul (talk) 06:12, 29 September 2010 (UTC)

,

Well, presently there's no option to make a filter public just for autoconfirmed users. It's either set to private, where only admins and edit filter managers can view it, or public, where everybody including IPs can view it. I would be against a mass switching of all the filters to public, as the code can always be provided to someone who needs for it some specific purpose but does not want the filter-view userright. And yes, there are definitely examples of people trying to work around the filters, such as http://encÿclopediadramatica.com/Abuse_filter#Evading_the_filters (misspelled on purpose), where the code of Filter 72 was leaked so that people could vandalize while managing to evade it (that filter has since been moved and re-written). —Soap— 12:30, 29 September 2010 (UTC)

Umlaut on the y? Looks like filter 72 isn't the only case of people trying to work around the filters... as if I needed any more proof there are too many bloody filters and blacklists in this place.

What scares you so much about new and unregistered contributors that you would be happy having some filters viewable by auto-confirmed users but not by them? Assuming your account didn't spring into life already auto-confirmed, you were one once.

More to the point, that filter was put in place to deal with the edit pattern of a long-term persistent vandal, something I mentioned as a valid use for a private filter. This rationale does not apply to filter 323 or any of the other filters aimed at casual vandalism; for those, knowledge of the filter by long-term persistent vandals is irrelevant. Gurch (talk) 15:56, 29 September 2010 (UTC)

Well, there doesn't currently exist a feature that allows filters to be seen by autoconfirmed users but not by everyone. I was just using that example because Sole Soul did. In theory it might be a good idea, because the longer that someone is here, the more likely they are to be trustworthy. However I would say that autoconfirmed is not really that difficult to get, so for the time being, until someone changes the software, I think it better to keep most of our currently "private" filters as private, since we can always give the source code to someone who makes a believable case for why they need to see it. I have no way of knowing if the "long-inactive sysop account" mentioned on ED is still around, or if it even existed. Also, I know you know this, but for the sake of those who don't, the ÿ on encyclopedia is due to the spam blacklist, not an edit filter, and I believe it's on meta rather than enwiki. —Soap— 16:46, 29 September 2010 (UTC)

Yes, it is due to the spam blacklist. That is kind of my point. I strongly recommend counting the number of ways an edit can be automatically rejected -- it's a lot higher than you think, and I guarantee you'll miss at least some of the methods the first time around. (Hint: Don't forget all those automated bots reviewed by a clique of bot operators and nobody else).

You still haven't provided a reason for keeping filter 323 private. Fears of a "long-inactive sysop account" are baseless, mainly because such an account would be able to see private filters anyway, but also because I am sick and tired of such FUD in the first place Gurch (talk) 17:05, 29 September 2010 (UTC)

I never said I wanted filter 323 to stay private. But you seem to have used it as part of an argument in favor of making a large number of other filters public, which I am opposed to. —Soap— 23:14, 29 September 2010 (UTC)

I used filter 323 as an example of a filter that has no reason to be private. There are many others, obviously. I can't list them all because I can't be sure exactly which ones should and should not be private without being able to see them. But the decision shouldn't be difficult for anyone who can:

1. Does the content of the filter associate a user account with an IP address or a user's personal information, or otherwise disclose personal details? I can't think of why such a filter should exist, but if one does, obviously it should stay private.
2. Does the content of the filter contain, or imply, statements about users or article subjects that should not be made public for e.g. libel concerns (say, a filter looking for the phrase "Gurch is a child molester").
3. Was the filter set up in response to a long-term persistent vandal, and does it look for edits characteristic of said long-term persistent vandal?

If the answer is "yes" to any of these questions, the filter should be private. If it isn't... I'm still seeing no reason that making it private is necessary. Gurch (talk) 09:55, 30 September 2010 (UTC)

Filter 323 doesn't catch all possible cases of undoing edits of these bots (we have no way of dirtectly checking if something is a revert), but nit makes sense to me that it would be private, as it owould be simple to get around if you saw the code. The code here is fairly simple.

To take a few other filters:

1. Filter 9 (the lowest number filter which is currently both active and hidden) - the filter has a long regular expression for "Personal attacks"; should they see the regular expression, iot would be simple to find some loop hole and enter some personal attack not caught by the filter.
2. Filter 260 - what is and what isn't caught by this shpould remain secret for similar reasons.

Additionally, per WP:DENY, several filters which are designed to stop specific vandals have names which don't suggest this. I won't name any specifics here, but they do exist.

And as to the criteria for which filters are hidden, some filters are designed against mistakes (such as filter 313) and common newby mistakes (such as filter 3); these should all be public. Others are actually against abuse, vandalism, clear rule violations, or attacks; these should be hidden if the filter name doesn't describe precisely what action it catches and how to avoid it (such as filter 339). עוד מישהו Od Mishehu 12:39, 3 October 2010 (UTC)

Internal mailing list

Some discussions related to the edit filter would best be done in an internal mailing list, such as those related to improving filters on long-term abuse edit filters and other filters targeting vandalism/abuse. I'm aware that comments can be used, but it gets tedious and hard to follow once many people have commented. Since an internal mailing list is no big deal and would help coordinate activities around here, I propose we apply for one. If there are no objections, I can get this done pretty soon. Regarding the name of the proposed mailing list... (wikipedia/wikimedia)-en- [abusefilter, filter, editfilter]? -abuse is already taken, and I'm leaning torwards -filter to make it shorter and easier to use. Netalarm^talk 02:19, 30 September 2010 (UTC)

Sounds good. I think wikipedia-en-filter or wikipedia-en-editfilter would be best. - EdoDodo ^talk 04:29, 30 September 2010 (UTC)

more secrecy? you guys just never stop do you? Gurch (talk) 09:46, 30 September 2010 (UTC)

I pretty much have to agree with Gurch here. If it can't be made public, use the "email this user" feature to contact the editor/admin responsible for the particular filter. A mailing list is an extraordinarily bad idea. Risker (talk) 00:31, 1 October 2010 (UTC)

Er... yes, I'm aware the "email this user" may be used to contact one user. But it's really hard to talk to everyone involved in a set of filters (all filters related to long-term abusive users). The reason I'm suggesting this is for better organization of the filters related to long-term abusive users so users that deal with them will find it easier to navigate and find the appropriate filters. For example, I'm trying to update the pages on several long-term abuse users with their targets and what not (and also determining if they are active, since edits blocked by the filter do not go through to SPI). I can't find the filters for "User A" because they are under some obscure name, and I can't find the filter for "User B" because it's lumped together with "User C" and "User D". Thus, in order to better organize the filters, it becomes necessary to discuss the contents of the filter. Do that in public? Emails could also be sent to every involved user and they could use "reply all" everytime. But why not just use a mailing list for that? A mailing list is no big deal, since it's basically just a multiple user email. (Of course, only sensitive issues should be discussed there... I'm not advocating for secrecy here, it's just that I see no better solution that would allow a group of users to discuss a set of filters). Netalarm^talk 01:47, 1 October 2010 (UTC)

Good idea for a separate mailing list. Wifione _....... ^{Leave a message} 01:26, 2 October 2010 (UTC)

Those who fail to remember history are doomed to repeat it. Bear in mind that a private "anti-sock" mailing list created such great disruption on this project that one administrator felt compelled to resign, and all who were being copied on the list were tarnished by it. This is unnecessary secrecy. Risker (talk) 02:22, 2 October 2010 (UTC)

I don't think you quite get what I'm proposing here. It's not an anti-sock or anti-abusive user mailing list. This mailing list will not discuss abusive users. It will be used to discus improvements to the edit filters of long-term abusive users and other private filters. No discussion of the users themselves, just how to improve the filter to prevent vandalism. Regarding the "unnecessary secrecy" - this is already happening in the edit filter comments. The notes can only be seen by a select group of individuals and are hidden from the public. All the mailing list will do is move those comments and discussions onto a system that is easier to understand and efficient. I see no problem with that, since overall nothing has changed. Netalarm^talk 03:52, 2 October 2010 (UTC)

Netalarm, with due respect to your proposal; don't disregard what Risker is saying. He's been there, seen that, and his comments should carry phenomenal weight. It'll be great to have his viewpoints, clarifications and further comments. Wifione _....... ^{Leave a message} 04:24, 2 October 2010 (UTC)

Right, I am not disregarding what he has to say. I merely feel that he might have misunderstood my proposal (and so is replying to a different idea), so I'm offering to clarify it. @Wifione. Can you understand where I'm coming from or is my wording complicated to understand? Netalarm^talk 04:27, 2 October 2010 (UTC)

Yes yes. I know the background and support the separate emailing list proposal. Wifione _....... ^{Leave a message} 04:59, 2 October 2010 (UTC)

A mailing list would be neither needed nor helpful.. Prodego ^talk 02:17, 5 October 2010 (UTC)

Request for a formal approval process for edit filter rights to be granted to non-admins

I request we reach a consensus to provide a formal approval process to grant edit filter rights to non-admins. I wish to put across a few issues before such a formal consensus process is finalised:

• Edit filters are amongst the key sifting tools that we have to ensure that interested and responsible parties can keep track of certain critical changes to the dump. The fact is that these are amongst the few tools that also help us track some most critical usage factors (that I'll not name here, but I'm sure most of my fellow abuse filter managers understand) to track some specific and particular kind of users.
• At the same time, we have a huge backlog of requested filters; some great ideas, some perhaps not so great, some impracticable on the server, but still, a backlog.
• We need to motivate trusted users (both systems operators and non) with relevant technical background and expertise to not only be able to contribute to reducing the backlog, but also to provide us with their analysis of the current key filters.
• At the same time, we are up against ensuring that our filters are not accessed/misutilised by non-constructive parties.
• I propose that for non-administrators, we develop a formal time-bound approval process (it could be a month-long, fortnight long... that's open to discussion) where consensus is developed on allowing/disallowing an applicant.
• However, I suggest that the consensus-process be unlike standard consensus platforms and should involve/allow ivoting only by current administrators/edit filter managers; as this would be key to ensuring that we involve only highly trusted editors into the fold.

I put forward this proposal to discussion. Thanks and regards. Wifione _....... ^{Leave a message} 07:33, 17 September 2010 (UTC)

Regarding your last point, the community doesn't like self electing groups. Cenarium (talk) 17:23, 20 September 2010 (UTC)

I've re-ordered the threads on this page to have permissions under a =level 1 header= and =other discussions= below. Slightly more formal, I guess =] –xeno^talk 13:59, 21 September 2010 (UTC)

IMO AFE should only be given to admins (who add themselves), or people who would clearly pass RfA and could significantly help with the AF. This tool allows for implementing page protection and blocking, it isn't to be bandied about. Prodego ^talk 00:24, 9 October 2010 (UTC)

If there were a usergroup that provided only 'view' access instead of 'view' and 'modify', would you be more lenient? —Soap— 00:31, 9 October 2010 (UTC)

Pagemove throttle for new users

Filter 68 seems to be producing quite a few false positives, prompting a few inaccurate AIV reports from Mr.Z-bot. Obviously there are extremely good reasons to employ such a filter, but perhaps it might be a good idea to change it to warn and tag rather than disallow so it doesn't disallow good-faith page moves just because the editor has made quite a few already. Anybody got any thoughts? HJ Mitchell | Penny for your thoughts? 22:25, 8 October 2010 (UTC)

Request for permission: User:Sidonuke

I have returned from wikibreak where I had a still open request for edit filter. I have actually got better in my regex editing and I hope to keep up with managing false positives and helping with requests. Also I would love to be able to fix past filters, improve current ones. For the most part I talk to many of the active edit filter managers on IRC, watching what they do and how they react to most things. I have experience with edit filters on personal wikis and would I would test thoroughly any filters before enacting them. Sidonuke 01:19, 29 September 2010 (UTC)

I still would like to keep this open. I have a few people claiming neutral support for the most part. I'm willing to answer any questions that may come my way. Sidonuke 17:04, 7 October 2010 (UTC)

I don't doubt your skills, but it seems you have relatively few non-automated edits, and you seem to have a history of taking long breaks between short periods of editing. This is not going to make me oppose, because we have other filter managers who aren't actively editing at all, but I can't support because I don't have anything visible to base my assessments on other than that you've helped edit a filter in the past. —Soap— 00:40, 9 October 2010 (UTC)

Well thanks for comment, I'll try contributing more to various places on the abuse filter that I can to show what I can do. Sidonuke 14:52, 13 October 2010 (UTC)

Filter 81 again

For some reason, Burn One Down, One More Payment and a few more articles by me triggered filter 81. I think it's because I put "Canadian RPM Country Tracks" and it's picking up Canadian Country because the R&R/Nielsen CAN charts are under WP:BADCHARTS. Is there any way to make the filter not report RPM peaks as false positives? Ten Pound Hammer, his otters and a clue-bat • ^{(Otters want attention)} 21:07, 8 October 2010 (UTC)

I looked at this the day you posted it and couldnt find any answers. Im just leaving a dummy reply now so that this doesn't get archived. —Soap— 07:08, 23 October 2010 (UTC)

Request for permission: User:Genjix

I'd like the ability to flag vandals. Sometimes I see them and all I can do is leave them one message asking them nicely not to do it while they run free on their train of disruption. I have reviewer rights and have edited for several years. Genjix (talk) 01:24, 18 October 2010 (UTC)

What do you mean by "flag vandals"? Can you explain the purpose of the edit filter? Note: User has also requested autopatrolled here Netalarm^talk 01:28, 18 October 2010 (UTC)

I just re-read what this is about and was about to delete this. I thought it was the right to place a template on a vandals user talk page warning them, which a bot would pick up and alert an admin to review. Genjix (talk) 01:30, 18 October 2010 (UTC)

No problem, if you want an easier way for warning vandals, consider using WP:Twinkle. Netalarm^talk 01:31, 18 October 2010 (UTC)

Great! That's what I was looking for. Thanks. Why aren't some of these tools on by default?? :) Genjix (talk) 02:13, 18 October 2010 (UTC)

Some users may be here to create content, fix grammar errors, etc., so such tools won't help them. That way if a user needs a specific tool, he can just activate the ones he needs. There may be more in there you'll find useful, such as WP:Friendly. Netalarm^talk 02:33, 18 October 2010 (UTC)

 Not done User withdrew request. —Soap— 21:36, 25 October 2010 (UTC)

Looping/substring

Hello there, I am user Eptalon, from Simple English Wikipedia. I had a quick look at the language used for abuse filters, and i found the following "extensions" might be useful:

• Commands like "getting the last ten editors" should return a "list" (or list-like structure), with the possibility to get the size, the first element, and the rest of the list
• Alternative would be to have a way to split a string, or get a substring

One of the problems we want to handle (and we seemingly can't) are the following:

• New user / IP (repeatedly) reverting a trusted editor or bot when editing an article

For this to be possible, we need two things:

1. A way of detecting a reversion/undo (that is not based on the edit comment, as this may be bogus/empty)
2. A way of getting each of the last ten editors in turn, and looping through them (no, a "member of" is not enough)

Any help would be appreciated. --Eptalon (talk) 09:46, 27 October 2010 (UTC)

I think everyone here will agree that those things would be great, but my understanding is that the edit filter is designed to run as fast as possible, and some things that might seem like obvious necessities cannot be included. I am not involved with the development of the software, though, so I may be wrong. —Soap— 10:09, 27 October 2010 (UTC)

I agree that this would be useful, although as Soap mentioned there is speed concerns. A variable to detect if a change is an undo would definitely be very useful, and probably wouldn't require too much performance lost. Looping might be more of a concern, however. - EdoDodo ^talk 15:54, 27 October 2010 (UTC)

Behavioural guideline?

This is marked as a "behavioural guideline" but I am having a hard time imagining what the guideline might be; "don't piss off the edit filter?". Retagging may be appropriate. Skomorokh 16:57, 30 October 2010 (UTC)

Request for a formal approval process for edit filter rights to be granted to non-admins

Hi, everyone.

I've found this discussion and think that it requires more attention.

My understanding is that editors wishing to request this permission flag, should make it at "[[Wikipedia:Requests for permissions/Edit filter manager]]" or another especific page like that.

In my opinion, requests may take in general at least three days to process, and this period can be extended if no consensus is reached. Mainly programming experience, knowledge of regular expressions and technical issues should be considered on the comments. If no problems are found, an admin can grant the flag.

Ruy Pugliesi^◥ 15:07, 31 October 2010 (UTC)

Request for permission: User:Barras

Hi all!

I'm Barras (talk · contribs) mainly from the simple English Wikipedia and meta (sul). We've recently enabled the abuse filter on simple and I'd like to take a look at the filters here on enwiki to see if we can import them to simple. I'm a crat and sysop on both, the simple English Wikipedia and meta, as well as a checkuser and oversighter on simpleWP. I won't make any changes to the filters here, I'd just like to have the ability to view the filters. Thanks in advance, -Barras ^talk 21:02, 25 October 2010 (UTC)

I trust you, and I think it would be great to have people co-ordinating back and forth between Simple and En.wiki. I think the case of User:Vito Genovese who received edit filter access here a few months ago for the same reason is a good precedent for granting this request. (Archive of Vito's request for reference.) —Soap— 21:33, 25 October 2010 (UTC)

I concur. Wifione _....... ^{Leave a message} 10:55, 30 October 2010 (UTC)

• Any news if I can get the flag or not? -Barras ^talk 16:31, 3 November 2010 (UTC)

You'll have to wait till more editors comment... Wifione _....... ^{Leave a message} 18:02, 3 November 2010 (UTC)

I've added the bit. This isn't the busiest page, but there has been some support, no objections, and it's possibly the clearest and most focused request I've ever seen on this page.  Done -- zzuuzz ^(talk) 22:39, 3 November 2010 (UTC)

Filter 368

Resolved

 – Wifione _....... ^{Leave a message} 10:47, 30 October 2010 (UTC)

Cross-posted from Wikipedia:Edit filter/False positives/Reports —Jeremy ^{(v^_^v PC/SP is a show-trial!)} 17:10, 29 October 2010 (UTC)

The whole freaking thing. Why does it even exist? What possible use does it serve, other than to needlessly harass new and unregistered contributors? Gurch (talk) 21:24, 28 October 2010 (UTC)

I agree with you but you should probably bring this up either at Wikipedia talk:Edit filter or talk with the creator of the filter, User:Hersfold. It doesn't seem to be on the requested filters page. —Soap— 22:13, 28 October 2010 (UTC)

Well I see you've already talked to him about it. But still, the people who work here are generally not the same people who create the filters and complaints (or comments) about a filter that aren't false positives would be better directed towards the authors of the filter or Wikipedia talk:Edit filter. —Soap— 22:33, 28 October 2010 (UTC)

• Unless there's any additional worry Gurch has with respect to filter 368, I presume this thread stands resolved. Wifione _....... ^{Leave a message} 10:47, 30 October 2010 (UTC)

• So this is resolved... how exactly? Because you moved it here? And then pretended the problem magically went away? There is no real agreement on exactly what to mark as minor, nor does marking an edit as minor aid in any way in hiding vandalism, so what advantage is there to harassing contributors in this way? Gurch (talk) 16:18, 30 October 2010 (UTC)

Dear Gurch, I marked this as resolved as I presumed you'd taken up the issue with an editor on his/her talk page. If you believe that's not sufficient, you can remove the resolved tag from this discussion. Otherwise, let it remain. Up to you. Sincerely. Wifione _....... ^{Leave a message} 10:15, 1 November 2010 (UTC)

Request for permission: User:Wayne Olajuwon

I've been fighting a lot of vandalism for a while now, and I would like to have to the abusefilter right because I would like to try to do this and be good at the abusefilter tool. WAYNEOLAJUWON 18:26, 7 November 2010 (UTC)

I wouldn't be comfortable adding edit filter permissions to your account, as although you're definitely a net positive, it seems your talk page is full of sections for mistakes you've made and you're often seen asking other people for help with questions about what to do in a given situation. So while your intentions are good, I'm not sure you're really ready for something like this. —Soap— 20:58, 7 November 2010 (UTC)

Same. I have a fair bit of technical knowledge, but am generally uncomfortable editing filters. It's a much more touchy thing, in my opinion, than even blockign etc. NativeForeigner ^Talk/_Contribs 23:35, 8 November 2010 (UTC)

Filter 18

A while back we had a filter to catch any edits which add text from the edit bar, such as "'''Bold text'''. It was disabled because "a lot of times people will make edits that have helpful information and "toolbar spam" in the same edit" (as Soap put it). I think that we should re-enable it, in the "warn+tag" mode. I think that this filter is more needed than the "Skype Toolbar Formatting" filter (filter 313). עוד מישהו Od Mishehu 19:59, 7 November 2010 (UTC)

Request for permission: User:Addihockey10

Hi; I'd like the edit filter permission to view the edit filters and modify them if needed, I'm interested in learning how they work and how to improve them. Thanks for your consideration! --Addihockey10^e-mail 17:05, 9 November 2010 (UTC)

• Oppose I don't really see you doing anything that indicates skill with edit filters. You can learn how they work without having the modify permission. More to the point though my judgment of you has really declined since Wikipedia:Requests_for_adminship/Addihockey10_3 and I would like to see you go at least a year or so without seeking additional permissions in the absence of a demonstrated need for them. —Soap— 10:19, 19 November 2010 (UTC)

I don't have much skill with the edit filters, but as you can see in this diff New Manchester High School, it was up for two hours. If we add to current filters or make a new filter this would probably crop up somewhere in Huggle or Igloo and speedily reverted. I had to manually revert this and luckily it was on my watchlist otherwise I'd doubt it would be reverted. If I do at one point end up modifying the filters, I'll contact a more experienced user to me to see if I had done it correctly. --Addihockey10^e-mail 13:21, 19 November 2010 (UTC)

That's not indicative of all of why you should have edit filter rights. Oppose. NativeForeigner ^Talk/_Contribs 02:18, 20 November 2010 (UTC)

Other discussions

edit filter for new pages

I am interested in writing an edit filter to prevent an editor from re-creating one of these recently deleted articles:

• David Cote (film director)
• David Cote (actor)
• Dave Cote

The editors involved seem very persistent and I expect that they will continue to find additional variations on the name. I think this filter would be sufficient to trap most potential name variants, with minimal false positives:

 article_text rlike "Dav.*Cote.*(actor|director|film)"

Would it reduce the cost of the test to also restrict it to new accounts (less than 24 hours old) and to article creation? I did not find an easy way at mw:Extension:AbuseFilter/RulesFormat to limit an edit filter to new articles. —Tim Pierce (talk) 18:19, 27 October 2010 (UTC)

I believe that for this case salting the page would be more appropriate. - EdoDodo ^talk 18:33, 27 October 2010 (UTC)

Another editor is pursuing salting for pages that have already been created and deleted. I am looking into an edit filter to guard against future attempts, e.g. David Cote (filmmaker), Dave Cote (actor), Dave Cote (director), Dave Q. Cote (film actor), etc. etc. etc. Is that not considered an appropriate use for edit filters? —Tim Pierce (talk) 18:39, 27 October 2010 (UTC)

Salting is more appropriate, as EdoDodo has already mentioned. Wifione _....... ^{Leave a message} 10:41, 30 October 2010 (UTC)

This has been done now, and it still hasn't stopped him, so there's now a proposal for an addition to the new page blacklist: MediaWiki_talk:Titleblacklist#Dave_Cote —Soap— 00:37, 2 November 2010 (UTC)

For future reference, please see the header at Wikipedia:Edit filter/Requested for what shouldn't be done using the edit filter. This seems to be a perfect example of a MediaWiki:Titleblacklist situation. עוד מישהו Od Mishehu 19:53, 7 November 2010 (UTC)

Thanks. I didn't realize that Mediawiki:Titleblacklist supported regular expressions, or I would have gone there first. Little embarrassing, that. Sorry to waste everyone's time. —Tim Pierce (talk) 17:34, 9 November 2010 (UTC)

filter 172

The edit filter didn't seem to pick up this edit. Odd. --Ixfd64 (talk) 23:45, 13 November 2010 (UTC)

Too many filters?

Of the last 1,198 actions, 30 (2.50%) have reached the condition limit of 1,000, and 15 (1.25%) have matched one of the filters currently enabled.

Of the last 699 actions, 17 (2.43%) have reached the condition limit of 1,000, and 14 (2.00%) have matched one of the filters currently enabled.

Of the last 1,840 actions, 54 (2.93%) have reached the condition limit of 1,000, and 31 (1.68%) have matched one of the filters currently enabled.

Of the last 687 actions, 16 (2.33%) have reached the condition limit of 1,000, and 7 (1.02%) have matched one of the filters currently enabled.
These are just random screen grabs, not intended to exaggerate the numbers. I believe this means we have too many filters running and that that is counterproductive because edits that actually should be stopped are going to slip through. Meanwhile there are still dozens of open requests on the Wikipedia:Edit filter/Requested page. If we are going to be able to address them we probably will first need to find out how to address the filters we already have. —Soap— 23:34, 20 November 2010 (UTC)

Can any of them be combined? So many of them start the same: !autoconfirmed, namespace==0, etc. Or would that not have a large effect on the conditions? Someguy1221 (talk) 00:47, 25 November 2010 (UTC)

Killed lots of questionable ones. Prodego ^talk 01:10, 25 November 2010 (UTC)

One of them that you killed are from ED trolls who have been steadily trying to vandalize and spam articles such as Manchester High School (Virginia), List of people from Charlottesville, Virginia, List of people on the autism spectrum, and Ruckersville, Virginia, to name a few. Unless it is being suggested that the only thing we can do is semi-protect all affected articles and be prepared to delete a whole bunch of new pages, that needs to be turned back on. –MuZemike 01:29, 25 November 2010 (UTC)

Feel free to do it, I can't identify what is useful and what isn't perfectly, and it is unfeasible to ask for every filter. Prodego ^talk 06:50, 26 November 2010 (UTC)

New message for Filter 344

If someone triggers Filter 344 and is disallowed, this should be the message instead (if any modifications should be made, go ahead)

The false positive line probably won't be used too much, but in case a typo is made in the filter, it would still be necessary. mechamind90 06:48, 26 November 2010 (UTC)

It isn't possible to modify the disallowed message, only the warning. Prodego ^talk 06:51, 26 November 2010 (UTC)

So how would it be done with MediaWiki:abusefilter-warning-urlobfuscation? I would think some type of modification of Filter 344 should be possible. I notice that Filter 306 doesn't have a "warn" action on it, even though I'm restricted from viewing the source. mechamind90 15:05, 26 November 2010 (UTC)

The way to do it is to give it a separate "warning" from filter 3, and have that other warning contain the message above (or some variation). Note that when we do "warn+sisallow", it will alternate between the warning and the standard disallow message. As far as I can tell, Filter 306 is just a disallow - so there is no warning set for it. עוד מישהו Od Mishehu 14:42, 28 November 2010 (UTC)

Section Break

For all of the following, we can do extensive normalisation, regex matching, length comparison and regular comparisons (less than, greater than, equal to) matching, combining different filters with boolean logic.

Just to note, this is little more than wiki-knowledge... nothing in that statement makes any logical sense and it'd be best to remove it. To explain, it sounds like you simply used wikipedia to get some "keywords" then mashed them together to form a geeky sounding statement... I mean "boolean logic"??? "comparisons"??? thats basic programming logic...

But why should you listen to me? After all, I'm a known vandal. I've vandalized: The sandbox, my talkpage, my signature, my own messages, and various talkpages (Though, it's amazing how easy it is to get banned by vandalizing the sandbox). I speak more than I edit because the "WikiProject" will never succeed until every edit is monitored by "Smart" people... not just "trusted" people.

Eitherway, edit filter is neat... but really more of an annoyance. There are a finite amount of filters compared to an infinite amount of cursing methods.

Test:Federal Urchin's Country Klub

Test:f.u.c.k.

• seriously? you forbid that, but not the easiest bypass?

Four letter obscenities aren't really the primary target of the edit filter, but even so, I think that the kind of things you're speaking of are not common in vandalism anyway. Few vandals are determined enough to spend time working out ways to get around the filters. —Soap— 21:44, 20 December 2010 (UTC)

Question about the "possibly autobiographical article" filter

Does the "possibly autobiographical article" filter (I can't remember its number at the mo') catch the repeated use of I, my, we and our, or does it just catch the use of the revision user's username in the article? If not that should be added to the filter, as some might try writing an article about themselves in the 1st person. Or maybe it deserves its own filter, to catch that outside of new articles. Sorry if this question is dumb, I don't know much about the edit filter. I only flip through Special:Tags looking for vandalism and speedy deletion tag removal, and that's the limit of my expertise. Thanks! --- cymru.lass ^{(hit me up)}⁄_{(background check)} 17:24, 30 December 2010 (UTC)

There seem to be two filters, 148 and 188 , and neither of them checks the article text, they only look at the title. Why we have two filters that do the same thing, I dont know. —Soap— 19:19, 30 December 2010 (UTC)

Okie doke. Thanks! --- cymru.lass ^{(hit me up)}⁄_{(background check)} 15:02, 3 January 2011 (UTC)

Filter this?

Shouldn't a newly-created article like the following be disallowed or tagged when created?

Here is an example:

[name redacted] is the mother of [name redacted] who has incredibly crusty hair. She likes to be shagged in the anus and she likes me to spaff on her massive tits !!!!

And another example:

[name redacted] is a Badass. Born 1988 he set the world record for fetal penis size. This lead him to much popularity in his teenage years. He laid pipe. A lot.

He has a friend named [name redacted] that love his anal with his gay lover ernesto. [name redacted] catches, ALWAYS. Always down for a good time call him @[telephone # redacted].

His friend [name redacted] loves long walks on the beach and sucking a lot of black penis. If he had one day left to live he would spend it giving butt hole pleasures and playing with his maltipoo effectionatially named rape me.

And a third:

A teenager of Tigard Oregon, [name redacted] (Pronounced: Ass-hole [name redacted]) has lead led a prosperous but hard life, becoming a heroin addict at the age of 5, but then moving on to destroy the shadow realms of Garaato'ol the following year. After reciveing the "Douchebag Of The Week" award in 2008, he was almost assasinated by Lee Harvey Oswald. But sensing the incoming danger, he roundhouse kicked the bullet toward a crowd of monkeys, causing the Great Monkey Massacre Of Super-Canada.

Early Life

After being born into dark depressing land of Crack Dealertopia, [name redacted] had to sell his mother to the Ice-Breathing Dragon mentioned in the hit new video game: Super Mario Brothers, for the money he needed to start his traveling super circus known as "Glee".

(Obviously, I did tag these page for speedy deletion under G10 and blank them!) Reaper Eternal (talk) 14:03, 5 January 2011 (UTC)

Do you mean you want a filter that checks if a new page has more than a certain number of words from a particular "bad words" list? I'm not sure that's possible. I promise to look at this if no one else does, I don't want you to get the impression that I always oppose new filters, but if it isn't possible to do it the "right" way then I think a filter that works half of the time and denies good pages the other half is worse than nothing. —Soap— 15:16, 5 January 2011 (UTC)

Maybe a filter like the following (I don't know if it will be too expensive). This probably isn't how an edit filter's code should look, but I don't really know edit filter code.

if (user_groups !in (confirmed | autoconfirmed)) // maybe should be "if (!(confirmed in user_groups))"?

if (article_articleid == 0) // only new pages should be affected

if (new_size < 5000) // assuming the var is in bytes, and most vandals don't make pages greater than 5k

if (count("anal", new_wikitext) + count("anus", new_wikitext) + count(" ass ", new_wikitext) + count("asshole", new_wikitext) + count("cunt", new_wikitext) + count("douche", new_wikitext) + count("faggot", new_wikitext) + count("fuck", new_wikitext) + count("gay", new_wikitext)) + count("homosex", new_wikitext) + count("nigger", new_wikitext) + count("penis", new_wikitext) + count("poop", new_wikitext) + count("rape", new_wikitext) + count("shit", new_wikitext) + count("tits", new_wikitext) + count("wank", new_wikitext) + count("whore", new_wikitext) > 3)

disallow

end

end

end

end

I probably forgot some four-letter-words in here. I have ass enclosed in spaces because ass goes into a large number of other words like passing. Reaper Eternal (talk) 18:06, 5 January 2011 (UTC)

On second thought, this might be better to code as a bot because then I could check for addition of infoboxes (probably not vandalism then), '''bold text''' or ''italic text'' (possible vandalism), and telephone numbers / email addresses / names in addition to the cussing (probable attack page). I could also increase the score of the new page a lot more if "high-level" curse words (like nigger) are used and increase the score only a little if "low-level" curse words (like ass) are used. Reaper Eternal (talk) 18:20, 5 January 2011 (UTC)

Thank you. Unfortunately because of the way the edit filter software works, a filter like that would have to run through the whole page individually for each of those checks above. So that would be 18 filters on every new page. We either have to do it all at once or not at all. The edit filter does allow us to combine key words, so there could be a filter with code like "count(anal|anus|ass|etc)", but I wouldn't feel comfortable activating such a filter without being sure beforehand that it will work as intended. When I get a chance (tomorrow night should be good) I'll look at it more to see if it's possible.

Also, filtering "poop" isn't necessary, as that word is considered so bad that it has a whole filter all to itself, and it is set to strict Disallow, unlike words like fuck which are merely considered borderline. (Yes, I'm serious. It seems deliberately crazy but it makes sense in a backward sort of way because those other words all have legitimate uses, whereas most people ... adults at least ... have no use for the word "poop" in everyday usage, at least not in Wikipedia articles, so the word "poop" is nearly always vandalism unless it's a quote of some sort. Some of these are vandals using it on the assumption that a word that toddlers use is not going to be caught by an obscenity filter. A good example of why you can't say poop but you can use grownup words such as fuck is pages like this.) —Soap— 03:17, 6 January 2011 (UTC)

Filter 380 is up; I set it to log only since I expect it will need quite a bit of work before it goes live. Indeed, in the 1 minute that it's been up, there's already two false positives, so this definitely needs more work. So far the false positives seem to consist mostly of large edits where the words were already present in the original text. Usually we get around this by checking the article twice ... I was hoping that wouldn't be necessary here, but it probably will be unless we decide to restrict it to just new pages. —Soap— 00:57, 7 January 2011 (UTC)

I'm disabling the filter for now, as it still needs work, and it seems to have grown away from the original purpose you had in mind, and because although it is tagging a lot of vandalism edits, many of them are caught by existing filters already. However I do want to work on this some more. A check to see if the "bad words" were present in the original text would be a nice change. —Soap— 22:30, 7 January 2011 (UTC)

This doesn't mean a bot is a bad idea though, if you're up to it. —Soap— 01:08, 7 January 2011 (UTC)

EternalBot (talk · contribs · deleted contribs · logs · filter log · block user · block log)

OK, the bot is now up and reporting at ##until_it_sleeps-bots ^connect. I'll continue tweaking it to reduce its false positive rate (currently less than 10%). Reaper Eternal (talk) 17:37, 17 January 2011 (UTC)

294

...Seems to be catching quite a few false positives. Three times today, I've cleared bot-generated AIV reports triggered by (different) editors leaving (perfectly civil) messages for (two different) admin querying the deletion of some non-notable article or other. This is quite concerning given that the filter disallows edits that trigger it. Is there a way it could be tweaked without allowing much of the crap it's trying to prevent? HJ Mitchell | Penny for your thoughts? 04:08, 16 January 2011 (UTC)

That filter is quite overgrown, I've asked NW to fix it (that filter is largely his). Prodego ^talk 06:36, 16 January 2011 (UTC)

247 (Adding emails in articles)

I've recorded a false positive in connection with use of the newsgroup citation template. If this could be considered as part of the filter, that'd be great. Thanks. --trevj (talk) 11:50, 14 February 2011 (UTC)

New userbox

I've designed a userbox for edit filter managers, you can view it here, I designed it for no other reason than one doesn't currently exist. Is it worth including in the main space? Pol430 talk to me 23:09, 22 February 2011 (UTC)

Nice looking! Reaper Eternal (talk) 23:25, 22 February 2011 (UTC)

I've moved it to the main space as Template:User wikipedia/Edit filter manager, if any edit filter managers want to create the, presently red-linked category, Category:Wikipedia edit filter managers, then that would be appreciated. Pol430 talk to me 00:24, 24 February 2011 (UTC)

Agreed, nice userbox. I un-redlinked the cat and am testing the box out on my userpage. 28bytes (talk) 03:07, 24 February 2011 (UTC)

Looking good on your userpage there, 28bytes :) Pol430 talk to me 20:45, 24 February 2011 (UTC)

(moved from above)

I had previously requested urgent help regarding an automatic filter.

My page had been sabotaged and I needed to reinstate the correct text but an automatic filter prohibited me from saving. The page had to be radically altered back to its original state in a matter of hours. Panic was on!

After spending time I didn't have trying to get some help and advice, I went back a few hours later and input the amendments one bit at a time and the the filter seemed no longer to restrict the saving. Yet the total text was exactly the same that it has previously refused.

What happened in between?

So, in case this issue arises agin, which it undoubtedly will, could someone please explain why an automatic filter should ban someone from editing his own page and then hours later accept eveything totally. If I can understand this, then I can perhaps be ready for the problem when it next comes up.

Thanxxx..

Will Golden^(talk) 15:30, 22 March 2011 (GMT)

Your account became autoconfirmed (meaning that you have been here 4 days and have 10 edits), and the 'Common vandal phrases' filter does not trigger on autoconfirmed accounts. Reaper Eternal (talk) 15:56, 22 March 2011 (UTC)

Archival of False Positives

The false positives page (Wikipedia:Edit filter/False positives) was inching toward 250k, so I took a look at why it wasn't being archived. Turns out, the archive page (linked from the Sept-Oct 2010 link on the page) actually goes from November through Mid February - and find that the single archive has bloated to over 2 Meg of records, some apparently duplicated repeatedly. The page barely loads for me, and I can't make heads or tails of what actually happened. Anyone want to take a crack at fixing it? Thanks! UltraExactZZ ^Said ~ Did 14:25, 23 March 2011 (UTC)

Near as I can tell, Cluebot III was repeatedly archiving the same 84 or so records - see [5] from 27 Feb, and the previous [6] and [7]. UltraExactZZ ^Said ~ Did 14:29, 23 March 2011 (UTC)

I've culled most of the more recent duplicates from the list; it's still 585k, and I'm unsure how to proceed. I've pinged the bot's talk page; Cluebot is one of the more reliable bots, so I'm not sure what happened here. UltraExactZZ ^Said ~ Did 14:59, 23 March 2011 (UTC)

Maybe you should try MiszaBot II? It archives many WikiProject talk pages. Reaper Eternal (talk) 16:06, 23 March 2011 (UTC)

That's what I'm thinking, too. Cluebot stopped archiving this one in late February, likely due to the over-bloated archive. A project for this evening, I imagine. UltraExactZZ ^Said ~ Did 18:41, 23 March 2011 (UTC)

Filter 18

I'm trying to test this filter against the page Sylver Logan Sharp, since that's one of the pages where an IP added "== Headline text ==" [8], but no hits are showing up. Is there a problem with the filter, or am I testing it wrong? I'm trying to make sure this filter doesn't generate false positives before enabling it. 28bytes (talk) 19:07, 23 March 2011 (UTC)

The filter itself looks correct, but I cannot see the test interface so I do not know whether you are testing correctly. Reaper Eternal (talk) 19:35, 23 March 2011 (UTC)

The test interface only has access to the last 30 days of changes. The change you are looking for is 12 months old. You will need a more recent test case. —UncleDouggie (talk) 11:45, 24 March 2011 (UTC)

Ah, OK; that helps! Thanks, 28bytes (talk) 13:28, 24 March 2011 (UTC)

Filter 394

Can anyone think of why this edit, which was caught by filter 98 , was not caught by 394 as well? The filters are identical except that 394 has (new_size < 70) where 98 has (new_size < 150) clause; the edit in question had a new_size of 30, so I'm curious why it wasn't picked up by both filters. What am I missing? 28bytes (talk) 18:17, 24 March 2011 (UTC)

Of the last 9,875 actions, 81 (0.82%) have reached the condition limit of 1,000... If an edit hits that limit, checks stop. 394 is a high numbered filter, 98 is not. That edit probably had more than 1000 checks on it, and once the 1000th was made, execution stopped. Note I didn't look at the filters to see that they were really the same, but if they are, that's why. Prodego ^talk 21:44, 24 March 2011 (UTC)

Thanks for the explanation, that makes perfect sense. Related to this, would it be better if I created a new filter instead of enabling filter 18 , so that it would never cause a more important filter (i.e. anything between 18 and 395) not to execute? 28bytes (talk) 22:39, 24 March 2011 (UTC)

If you change the purpose of a filter, then you should create a new one instead. If what you are doing fits better into an existing filter (disabled or not) then use that instead. Prodego ^talk 23:29, 24 March 2011 (UTC)

OK, thanks. 28bytes (talk) 23:39, 24 March 2011 (UTC)

Request for permission for Wikipedian2

 Not done Wifione _....... ^{Leave a message}

The "Edit filter" tool would grant me the privilege to extend my continual work in anti-vandalism, and mould in with the existing work using Rollback in Huggle, and my reviewing on Pending Changes to increase the capacity utilisation of my edits - improving faster identification, analysis, and removal per WP:Vand. The edit filter would give me a transparent and accountable means to retrieve destructive edits using the criteria of user specifics and patterns, combined with title examination.

The setting of an abuse filter will be done with exceeding care, checking and double-checking against relevant impacts of the Wikipedia community, and using consequential weighting techniques to create a filter that will prevent abuse, but will not prevent good faith edits. Wikipedian2 (talk) 21:16, 9 April 2011 (UTC)

Transcluding the entire WP:Vand page in your request was not an encouraging sign. Without looking any further I'd have to say you don't appear sufficiently careful enough to be trusted with this flag. 28bytes (talk) 21:24, 9 April 2011 (UTC)

That's a use of shift I won't be doing again! Wikipedian2 (talk) 21:31, 9 April 2011 (UTC)

How good is your knowledge of regular expressions? Reaper Eternal (talk) 22:57, 9 April 2011 (UTC)

Regular expressions (or regex) is the computerised matching and pairing of a string of words (or characters) against a given variable. A randomised basic example could be the phrase variable 'asd', the word 'asd' could then be compared by the parser in a variety of contexts, including "sdjasd", "purple asd", or "asd" on it's own. If you would like me to go into more detail, please feel free to ask. Wikipedian2 (talk) 23:22, 9 April 2011 (UTC)

• Comment You don't really have enough edit history for me to base an opinion on, since you've only been actively editing for about two weeks. I have to default to oppose, though, as with MacMed above. If this request doesn't get consensus, I would keep it in mind and try to work on edit-filter-related things for a few months and then we would better be able to see how edit filter access would help you. —Soap— 00:14, 10 April 2011 (UTC)

Not now Wikipedian2. Prodego ^talk 00:21, 10 April 2011 (UTC)

Proposed Filter

I thought of this after seeing this article (deleted through AFD, admins only). I speedied it under A7, and it was then recreated with a {{hangon}} tag on it. I was wondering about a filter that warned about such an edit and then directed them to ask either the tagger or deleting admin why their article was deleted so they understand what they need to fix. Here is the code I was thinking of.

article_namespace == 0

& old_size = 0

& lcase(added_lines) rlike "\{\{hangon\}\}"

Any suggestions/thoughts on if it's worth the efficiency hit or if there's problems with the code? Regards, MacMed^talk_stalk 00:48, 31 March 2011 (UTC)

Hi MacMed. There's an open request for such a filter here. I've also seen pages created like that while doing new page patrol. I wonder how useful a "canned" warning message would be in a situation like that, though? 28bytes (talk) 00:54, 31 March 2011 (UTC)

If they were disallowed from the edit and told in the message to click on the username of the deleting admin and ask why the article was deleted and how they could fix it, perhaps with an explanation from the admin and possible user-fication of the article, the IP can improve it and submit it (as long as it's not a notability issue). Maybe even get a new editor (be optimistic!) :) haha. Maybe just set it up as a logging filter first and see if there's enough hits to warrant a filter in any case? Regards, MacMed^talk_stalk 01:03, 31 March 2011 (UTC)

I'd say it is best to not block or warn for such an edit since it notifies new page patrollers that the article has already been speedied, and probably will need speedying again. Reaper Eternal (talk) 01:29, 31 March 2011 (UTC)

Maybe a tag with "New page created with {{hangon}}" would be better? Then NPP can find them and leave a message for the editors. Regards, MacMed^talk_stalk 01:30, 31 March 2011 (UTC)

Hangon taggings cause the article to be in the speedy delete category already, so an admin will review them anyway. Graeme Bartlett (talk) 10:37, 31 March 2011 (UTC)

But do we want a new user to get the feeling their article is being continually deleted, to the point that they get fed up and leave or vandalise? Or would we rather they ask how to fix it, and learn how to be a welcome part of the WP community. Regards, MacMed^talk_stalk 17:05, 31 March 2011 (UTC)

I'm wondering if a bot (like SDPatrolBot) might be better suited to this task? A bot could examine the page log and put a friendly notice on the editor's page telling them which admin to contact, with a link to their talk page. 28bytes (talk) 17:26, 31 March 2011 (UTC)

Request permission for Reaper Eternal

 Done by Prodego. Wifione _....... ^{Leave a message}

Hello, I would like to request the 'abusefilter' privilege to help working against vandalism. I would mainly use it for the edit filters used to disallow the blatant talk page attacks—commonly by banned users—that show up every now and then. I would also use the ones for detecting the prolific sockpuppetteers, as I commonly see groups of puppets when Huggling. I have been editing for around 6 months. My knowledge of regex comes entirely from teaching myself for the purpose of creating ReapETbot (talk · contribs). Thanks all! Reaper Eternal (talk) 03:11, 11 April 2011 (UTC)

I see no issues that would make me oppose, though I have to admit it's not easy to page through 25000 edits of mostly Huggle so I just looked at your talk page archives. I've seen your bot at work on the IRC network and have always been impressed with it. —Soap— 21:15, 11 April 2011 (UTC)

From what I've seen of Reaper Eternal, he's knowledgeable and cautious. I trust him not to break anything. 28bytes (talk) 00:08, 13 April 2011 (UTC)

Reaper Eternal, you are a very clueful candidate for EFM however I am just asking if you could possibly give us a little more insight into regex such as its meaning and how it works. mauchoeagle 00:14, 13 April 2011 (UTC)

The word "regex" is a shortened form of "regular expression". It is a method of finding patterns in strings. I won't go into a full dissertation on how regular expressions work, as that would take up an enormous amount of space. The [ ] operators delineate a set of characters, any of which may match the target string. However, appending a ^ to the front of the characters (e.g. "[^fsag]") matches all characters not in that charset. (Ex. "d[ou]g" matches "dog" or "dug".) The | is an or operator, and the target string can match either of the strings it separates. (Ex. "cat|dog" matches "cat".) The ( ) operators are used just as they are in arithmetic—to group elements. (Ex. "pine(apple|nut)" matches "pineapple" and "pinenut".) The { } operators indicate number of elements to match. (Ex. "(the){3}" matches "thethethe".) The \b is used for word boundaries, the \w is any alphanumeric character, \d is any digit, \D is any nondigit, \s and \S are whitespace and nonwhitespace respectively, \n is a newline, \r is a carriage return, \x## (where ## are two hex chars) matches that explicit ASCII character.... There is far more that I have not even mentioned. Reaper Eternal (talk) 01:18, 13 April 2011 (UTC)

Wow, I am studying computing and that is the definition my professor would give me. Impressive enough for me to most definitely Support. mauchoeagle 01:26, 13 April 2011 (UTC)

• I don't normally comment on these, but I'll make an exception, here. Any admin closing this should bear in mind that I retain the EFM bit for the passive benefits only and have no idea what I'm doing with filters and weight my comments appropriately. Reaper Eternal is someone I know fairly well. We've collaborated on a few tasks and had a few shared interactions. More importantly, though, I've seen the work he puts into writing, reviewing and his excellent vandal fighting. He seems to have picked up a good knowledge of how filters work and has his fair share of experience dealing with the kind of crap they're meant to combat, so I think he'd do a grand job with the EFM bit. Just in case it's not obvious, I'm supporting! HJ Mitchell | Penny for your thoughts? 00:16, 13 April 2011 (UTC)

Thanks for the vote of confidence. :) Reaper Eternal (talk) 01:58, 13 April 2011 (UTC)

Other discussion

Fairly simple filter desperately needed

Could someone look at creating a filter to deal with this nonsense? Preferably disallowing the edit and having the bot report it to AIV. Thanks, HJ Mitchell | Penny for your thoughts? 19:37, 12 April 2011 (UTC)

Yup, let me take a stab at it. 28bytes (talk) 19:40, 12 April 2011 (UTC)

Much obliged. HJ Mitchell | Penny for your thoughts? 19:44, 12 April 2011 (UTC)

No need. User's gone for 6 months. —Jeremy v^_^v ^{Components:V S M} 19:50, 12 April 2011 (UTC)

Filter 398 created. 28bytes (talk) 19:53, 12 April 2011 (UTC)

A trifecta! User(s) blocked, page protected and an edit filter. Can probably unprotect that talk page at least. 28bytes (talk) 19:57, 12 April 2011 (UTC)

If any EFM wants to copy the strings from 398 into a more appropriate filter and kill 398, that's fine with me. 28bytes (talk) 19:59, 12 April 2011 (UTC)

It would probably go well in 58. The filter will be useful if he manages to evade the IP block (or for the five minutes between its expiry and it being reapplied!). Thank very much, 28b. I'll unprotect the talk page in a mo. HJ Mitchell | Penny for your thoughts? 20:20, 12 April 2011 (UTC)

This is just part of a larger scheme; see Wikipedia:Sockpuppet_investigations/TERRY_PAULI_PREDICTED_THIS. I'm not sure if it's going to be possible to keep on top of the many strings he's been typing, apologies to 28bytes that his diffs have been basically all deleted now and can't be read anymore. —Soap— 20:31, 12 April 2011 (UTC)

No worries. Figured a stop-gap filter would do for the time being, feel free to disable and delete it if it's no longer needed. 28bytes (talk) 20:36, 12 April 2011 (UTC)

Here's a diff showing some of the content. [9] Reaper Eternal (talk) 21:22, 12 April 2011 (UTC)

Filter for links that are piped to weird targets

I think a new tag/filter would be useful to catch links that are piped to suspicious targets or vice versa, like [[Joe Shmoe|douchebag]] and [[douchebag|Joe Shmoe]]. — Preceding signed comment added by Cymru.lass (talk • contribs) 01:59, 13 April 2011 (UTC)

Filter 24 used to do that, but it was turned off in part because of edits like this which are legitimate but appear to the filter as if they were vandalism. In other words it essentially prevented any new editor from editing a paragraph with a piped link to a page such as fag even if it was legitimate. It may be possible to revive the filter by checking twice to see if the words are already present in the paragraph, but even with that, edits like the above would still be caught because in that case the visible link was spelled with periods in the name (F.A.G.). So there would still be at least some false positives. —Soap— 02:17, 13 April 2011 (UTC)

In that case, subpattern matching could be used to check for separators within and after the string. — anndelion ※ 03:24, 13 April 2011 (UTC)

Well, I'm not talking about disallowing those edits. I'm talking about just catching them via a tag on Special:Tags, much like 29 catches edits that remove speedy templates and tags them, giving them their own recent changes page here. That way, people can check through them to see if they are indeed improper removals of speedy deletion templates, instead of just having the filter disallow speedy removal entirely. (Sorry if this sounds confusing. I'm going to blame it on the jet lag, although I suspect it's just general cymru.lass craziness...) — Preceding signed comment added by Cymru.lass (talk • contribs) 04:09, 13 April 2011 (UTC)

Wait, I found the right place to request a filter. Never mind! :) — Preceding signed comment added by Cymru.lass (talk • contribs) 07:31, 13 April 2011 (UTC)

• I think this could be a good idea, especially if it's tag only There aren't many any?) legitimate uses for piped links to targets like douchebag. Whisky drinker | HJ's sock 09:51, 13 April 2011 (UTC)
  • I've put in a request over at Wikipedia:Edit filter/Requested#Misleadingly piped links — Preceding signed comment added by Cymru.lass (talk • contribs) 05:26, 14 April 2011 (UTC)

Querying account name in filter

Looking through the documentation, I can't see how to apply a regex against the account name. I'm considering a log-only filter against "username ends in 69 or contains 'rocks'" and "article contains "Eurovision" to help with Category:Suspected Wikipedia sockpuppets of Xtinadbest and Category:Wikipedia sockpuppets of Xtinadbest. The Eurovision part is easy, but the username check doesn't seem to be.—Kww(talk) 19:08, 16 April 2011 (UTC)

I'll make one. Could you give me some examples of edits you want to target? Reaper Eternal (talk) 19:24, 16 April 2011 (UTC)

• http://en.wiki.x.io/w/index.php?title=Kristina_(singer)&diff=prev&oldid=394963790
• http://en.wiki.x.io/w/index.php?title=Rockefeller_Street&diff=prev&oldid=424187225
• http://en.wiki.x.io/w/index.php?title=Horehronie_(song)&diff=prev&oldid=390470597
• http://en.wiki.x.io/w/index.php?title=J%C3%B3hanna_Gu%C3%B0r%C3%BAn_J%C3%B3nsd%C3%B3ttir&diff=prev&oldid=318311744
• http://en.wiki.x.io/w/index.php?title=Eurovision_Song_Contest_2010&diff=prev&oldid=312337072

Thanks.—Kww(talk) 19:33, 16 April 2011 (UTC)

Created at Special:AbuseFilter/328. I may have to turn it off due to time consumption, although I think that keeping the page text check at the end will keep this down. Reaper Eternal (talk) 20:01, 16 April 2011 (UTC)

Actually, it appears to be one of the leaner filters. Reaper Eternal (talk) 20:05, 16 April 2011 (UTC)

Thanks. I expect this to be a low hitter, but let's try keeping it around for a month or so to get a track record. This has been going on for years, and the target area is so wide I can't keep effective watch over it.—Kww(talk) 20:33, 16 April 2011 (UTC)

Possible filter

Hi, I was wondering if it would be possible to create a filter which prevents non-free images being added to article for which they don't have a fair use rationale. Since each image must have a separate rationale for each article, it would presumably be easy to check, each time an image is inserted, whether it includes a "non-free" template, and if so, whether the article in question is listed on the file page. Any thoughts? ╟─TreasuryTag►without portfolio─╢ 13:07, 28 April 2011 (UTC)

I believe you'd need a bot for that, since edit filters aren't able to analyze the contents of another page (i.e. the image description page, to look for the FUR) when evaluating an edit. 28bytes (talk) 13:21, 28 April 2011 (UTC)

Proposed filter

After far too many sockpuppets and a request from a CU regarding Wikipedia:Sockpuppet investigations/Dr.Mukesh111, I wrote up filter Special:AbuseFilter/406. However, it has been awhile since I have done any edit filter work and I would prefer that another filter editor review my coding before the filter is enabled. I cobbled it together from a couple of other filters and common traits of the sockpuppeteer. It is not perfect, but it may help the situation. Thanks. -- Gogo Dodo (talk) 07:11, 29 April 2011 (UTC)

Great to see another (returning) filter manager! I've left some comments in the filter itself. Feezo (send a signal | watch the sky) 12:17, 29 April 2011 (UTC)

Thanks for the feedback. I made some changes. Can you look at it again? -- Gogo Dodo (talk) 00:43, 30 April 2011 (UTC)

Yep, it looks good. :) Feezo (send a signal | watch the sky) 00:58, 30 April 2011 (UTC)

Archiving discussions at Wikipedia:Edit filter/Requested

There are a lot of old discussions at Wikipedia:Edit filter/Requested. It looks like the last time any discussion was archived was in January 2010, about a year and a half ago. Does anyboy mind if I archive all the discussions from August 2010 and older? Thanks. - Hydroxonium (T•C•V) 00:54, 22 April 2011 (UTC)

Changed my mind. There are a lot of older requests that still need comments, so I'll archive the current "Completed requests" and "Denied requests" then move the unsorted completed and declined to there appropriate sections. - Hydroxonium (T•C•V) 01:30, 22 April 2011 (UTC)

 Done, everybody is free to revert, especially if I screwed up anything. I'd like to mention that there are a bunch of requests that are several months old that haven't been completed or denied. Should these be revisited or archived as stale or something else? - Hydroxonium (T•C•V) 07:29, 22 April 2011 (UTC)

I've archived all the completed and denied requests from 2010 at Wikipedia:Edit filter/Requested - Hydroxonium (T•C•V) 14:42, 3 May 2011 (UTC)

Inexplicable triggering of filter 52

How did [10] manage to trigger Special:AbuseFilter/52? -- King of ♥ ♦ ♣ ♠ 10:44, 30 April 2011 (UTC)

I've changed the code of the filter now. Before, it was triggering based on something that isn't in the code any longer ... see the last line of the notes box for details. —Soap— 11:58, 30 April 2011 (UTC)

Request permission for Porchcrop

The following discussion is closed. Please do not modify it. Subsequent comments should be made in a new section.

 Not done Based on competence issues (as well as incivility, but that's a different story), at the present time the tool will not be granted to Porchcrop. Eagles 24/7 (C) 19:55, 4 May 2011 (UTC)

I am requesting to use the edit filter manager permission. I have been a constructive editor to the project. (While I am topic banned, I am being mentored by User:Worm That Turned and I have been making good progress with him.) -Porchcrop ^{(talk|contributions)} 09:36, 3 May 2011 (UTC)

Are you competent at reading, writing, and modifying regular expressions? – anna 10:15, 3 May 2011 (UTC)

Hi Anna. Yes I am familiar with regular expressions. -Porchcrop ^{(talk|contributions)} 08:36, 4 May 2011 (UTC)

Sorry, I'm going to have to strongly object to this request. I don't think it would be appropriate for a user currently under competency-based editing restrictions to have this right. 28bytes (talk) 09:00, 4 May 2011 (UTC)

Well I'm not sure what this has to do ANYTHING involving the restrictions in my topic ban. I am good at regular expressions, and to be honest with you, I have done well with my mentor. And I definitely have improved alot in the areas I got banned, and when the ban expires, you'll see alot of changes from me. So please do not cause trouble here. -Porchcrop ^{(talk|contributions)} 09:07, 4 May 2011 (UTC)

To be fair to Porchcrop, he has made significant progress whilst working with me. I'm not certain why he wants or needs the right, but I hold no objections. If it helps, I would be willing to request the same right, keep an eye on what he's doing - I am more than competent at regular expressions, holding both a mathematics degree and a computer science one, working in an area that uses regular expressions ... with some regularity. Worm^TT · (talk) 09:17, 4 May 2011 (UTC)

(edit conflict)This is nothing personal against Porchcrop. But this is a user right that even administrators do not have by default, so aside from familiarity with regular expressions, some demonstrated proficiency at not breaking things is required, and I haven't seen that from Porchcrop (yet.) I'd be happy to consider supporting once his topic bans expire. 28bytes (talk) 09:23, 4 May 2011 (UTC)

Porchcrop, you were subjected to disruption/competency sanctions less than two months ago. I don't think this is the right time, and agree entirely with 28bytes (talk · contribs). ╟─TreasuryTag►condominium─╢ 09:22, 4 May 2011 (UTC)

Seriously, I'm not sure what you mean about competence. The restrictions in the topic ban has ABSOLUTELY NOTHING to do with this right. 1) I know regularly expressions, put it in your head 2) The competence and disruption is ONLY for the admin areas, not for the edit filter manager areas 3) I was a bit younger, but now I am more careful, competent and clueful. Get all these things right. -Porchcrop ^{(talk|contributions)} 09:27, 4 May 2011 (UTC)

The restrictions in the topic ban has ABSOLUTELY NOTHING to do with this right. Pardon me, but the editfilter right is a position of responsibility, and since the thread linked above suggests that you were unable to understand simple policies and processes such as WP:CSD and WP:UAA, I'm not convinced that you are (yet) ready for that. Your SHOUTING CAPITALS above and in your edit-summary also do not demonstrate patience and maturity. I know regularly expressions, put it in your head – incivil and mis-spelt proof by assertion isn't particularly useful here. I was a bit younger... Two months? Seriously? ...but now I am more careful, competent and clueful. Well... Get all these things right. Perhaps not. ╟─TreasuryTag►First Secretary of State─╢ 09:30, 4 May 2011 (UTC)

I am very serious, the right has no damn thing that gives me any abilities to go against the ban. And please do not say I am being uncivil, there are other editors and administrators that talk like this. -Porchcrop ^{(talk|contributions)} 09:34, 4 May 2011 (UTC)

I will gladly "STOP" as per your request but please be aware that you are being incivil, and that more than one person disagrees with your assessment of this situation, so it may be worth you reconsidering it. ╟─TreasuryTag►District Collector─╢ 09:36, 4 May 2011 (UTC)

I forgot to mention to you though that you are off topic. I HAVE gained alot of competence and clues. The right does not give me any abilities to violate the ban. -Porchcrop ^{(talk|contributions)} 09:39, 4 May 2011 (UTC)

Say, could you explain this oppose? You had said "ridiculous", which that was against WP:NPA, and you are telling me to abide by the civility policy. -Porchcrop ^{(talk|contributions)} 09:53, 4 May 2011 (UTC)

This is getting way off-topic. If you have objections to TT's oppose in your RfA, this isn't the place to air them. 28bytes (talk) 10:01, 4 May 2011 (UTC)

Oh please assume good faith. First of all, he said that I am uncivil when many other editors and administrators talk like this. Second of all, he made a personal attack, which was against a policy, and he's telling me to abide by a policy. -Porchcrop ^{(talk|contributions)} 10:05, 4 May 2011 (UTC)

It's not a violation of Wikipedia:Assume good faith to say that something's off-topic. Or is it? ╟─TreasuryTag►co-prince─╢ 10:07, 4 May 2011 (UTC)

Porchcrop, this right is about trust, not just about ability, as 28bytes has pointed out above. As you are editing under a ban, the community is not likely to trust you with sensitive areas such as this. 28bytes has confirmed that he will reconsider once your topic ban expires but for now, I do not think that you are likely to gain this right. Worm^TT · (talk) 09:39, 4 May 2011 (UTC)

Worm, I know that. But this is just a small topic ban, not a full ban or a block. -Porchcrop ^{(talk|contributions)} 09:43, 4 May 2011 (UTC)

Yes, but the fact it exists casts doubt - topic bans (even ones as limited as yours) have repercussions. Have a little patience, sit it out, keep working on what we're working on, and then re-request. Worm^TT · (talk) 09:46, 4 May 2011 (UTC)

Okay. Provided that all these people support and accept me to use this right at 12 February 2011. -Porchcrop ^{(talk|contributions)} 09:48, 4 May 2011 (UTC)

I think the best you can ask for is that they evaluate your request in good faith at the time, and I am sure that they will. Worm^TT · (talk) 09:58, 4 May 2011 (UTC)

I have some competence-based concerns as well from a scan of your edits, apologies. "when the ban expires, you'll see alot of changes from me" -- what's stopping you from enacting these (presumably positive) changes now? – anna 12:50, 4 May 2011 (UTC)

• oppose No. Not for someone who is so clearly desperate for "adminship lite" by any forum possible. Andy Dingley (talk) 14:29, 4 May 2011 (UTC)

The discussion above is closed. Please do not modify it. Subsequent comments should be made on the appropriate discussion page. No further edits should be made to this discussion.

Archive run needed

Wikipedia:Edit filter/False positives/Reports needs an archival run, and soon. We're at 180 sections. —Jeremy v^_^v ^{Components:V S M} 16:26, 5 May 2011 (UTC)

 Done - Hydroxonium (T•C•V) 06:19, 7 May 2011 (UTC)

Danke, Hydroxium. —Jeremy v^_^v ^{Components:V S M} 15:22, 7 May 2011 (UTC)

Request permission for lustiger_seth

Hi!

I'm a admin, restricted to stuff concerning link spamming, especially WP:SBL.

I'd like to use the permission in order to amend some WP:SBL-tasks. -- seth (talk) 20:57, 9 May 2011 (UTC)

No objections from me. I don't see any complaints about or reverts of your changes to the MediaWiki namespace; you obviously know what you're doing with regular expressions. 28bytes (talk) 21:32, 9 May 2011 (UTC)

Done, feel free to use EFM with no restriction to SBL. Prodego ^talk 21:41, 9 May 2011 (UTC)

thx! -- seth (talk) 19:04, 10 May 2011 (UTC)

Other discussion

Possible filter for LPC

Any time an edit at LPC involves the regex lun\w*\s+pe\w*\s+char, it is vandalism; "lun peh charh" (among many transliterations) is a Hindi and/or Urdu expression meaning something like "go sit on a dick". Said edits are infrequent but quite persistent. It seems like an opportunity for positive use of filtering, if anybody's interested. —chaos5023 (talk) 11:57, 7 May 2011 (UTC)

searching the filter

Hi!
Is there any comfortable possibility to search the edit filter's conditions and notes? This could be helpful to avoid redundant filter rules. -- seth (talk) 19:09, 10 May 2011 (UTC)

I'm going to guess your question is: "How can I search the edit filters?" It is not possible to search the edit filters as far as I can tell. Reaper Eternal (talk) 22:46, 20 May 2011 (UTC)

Yes, I want to search the edit filter (especially the parts "condition" and "notes"). Maybe I should file a feature request at bugzilla. -- seth (talk) 06:05, 21 May 2011 (UTC)

Oh, I can use the api for that. Thanks anyway. -- seth (talk) 17:28, 22 May 2011 (UTC)

revoke autoconfirmed status

I have not noticed any filter, including deleted ones, that revoke an users' autoconfirmed status. I have seen one that blocks a user from obtaining autoconfirmed status though. Is there any filter that revokes an users' autoconfirmed status? Crazymonkey1123 (Jacob) ^{T or M}/_{Sign mine} 21:46, 21 May 2011 (UTC)

I don't think that non-abusefilter managers can see that deleted filters even exist. Reaper Eternal (talk) 00:06, 22 May 2011 (UTC)

All users (including IPs) can see the list of filters, deleted or not; they just can't view the details on the private ones. 28bytes (talk) 00:16, 22 May 2011 (UTC)

What are you talking about Crazymonkey? I don't think there is any way to revoke autoconfirmed after 10 edits/4 days. —GFOLEY FOUR— 00:08, 22 May 2011 (UTC)

Yes, edit filters can be configured to revoke autoconfirmed status. 28bytes (talk) 00:09, 22 May 2011 (UTC)

(edit conflict) YesAs 28bytes says, you can. If an edit filter manager sets the 'Block autopromote' flag on an edit filter, the triggering user's autoconfirmed status will be removed by the MediaWiki software. (This is why there is a "Autoconfirm this user" feature.) Here is an example of a user getting de-autoconfirmed. Reaper Eternal (talk) 00:11, 22 May 2011 (UTC)

The Sleeper Socks II filter also blocks autopromote. That is the only other one that I have found. Crazymonkey1123 (Jacob) ^{T or M}/_{Sign mine} 16:12, 22 May 2011 (UTC)

Condition limits

We're hitting the condition limits hard right now (~7.4%). Any objections to temporarily turning off, say, 3, 50, 123, and 163? Feezo (send a signal | watch the sky) 20:23, 22 May 2011 (UTC)

Leave NawlinWiki a note to clean up his filters. I'll go through the newer ones (250+) and see what we can disable. Prodego ^talk 21:58, 22 May 2011 (UTC)

Disabled: 408, 407, 413, 399, 341, 389. Prodego ^talk 22:07, 22 May 2011 (UTC)

Also took out 123 and 406. Limit is around 1.3% now. Feezo (send a signal | watch the sky) 22:40, 22 May 2011 (UTC)

Rename

For filter 384, the description is "adding one bad word and nothing else" but I checked the log, and edits that add more than 1 bad word. I think that the description should be changed for "adding a bad word" for the reason. ~~EBE123~~ ^talk_Contribs 19:56, 28 May 2011 (UTC)

Notice the edit_delta parameter. That is why it is called "adding one bad word and nothing else". Reaper Eternal (talk) 22:45, 28 May 2011 (UTC)

Deleting blatantly bad-faith false positive reports

Awarded to all who work here. —Soap— 02:33, 7 June 2011 (UTC)

I have been marking the blatantly bad-faith reports as vandalism and not needing attention. I was wondering if we could just remove them as they make the page very large. Thanks! Reaper Eternal (talk) 00:28, 6 June 2011 (UTC)

No objection from me; it says at the top of the page that bad-faith reports may be removed, and I sometimes do so. 28bytes (talk) 00:37, 6 June 2011 (UTC)

That was my doing, thanks for noticing :) I suggested it on the false positives talk page and added the warning when no one objected. By the way, it looks like ClueBot III's not properly archiving that page; it moved the last one to ...Archives/ 1, which made it not show up in the archive box (it was actually the second archive). Feezo (send a signal | watch the sky) 01:12, 6 June 2011 (UTC)

I have now removed 53 kilobytes of blatantly bad-faith reports. Pheww! Reaper Eternal (talk) 01:34, 6 June 2011 (UTC)

Good work! Feezo (send a signal | watch the sky) 14:02, 6 June 2011 (UTC)

Sounds like a good idea to me. —Soap— 02:33, 7 June 2011 (UTC)

Request for permission for Ancient Apparition

I've been learning regex for awhile now, I've got a high enough competency level to understand the filters and will avoid making changes that might break Wikipedia and will always defer to more experienced EFMs. I'd like to take a stab at managing the filters. I'd be happy to answer any questions in relation to regex or current filters. —James ^{(Talk • Contribs)} • 7:08pm • 09:08, 15 June 2011 (UTC)

 Not done. There is no visible support to your request. Therefore, marking as not done. Thanks and best. Wifione _....... ^{Leave a message} 00:18, 23 June 2011 (UTC)

Other discussion

User warnings

Hello, I created a user warning for reporting vandalism as a false positive. Do you have any suggestions as to whether or not we should use it or what tweaks should be made? Reaper Eternal (talk) 20:22, 8 June 2011 (UTC)

If the "false positive" they're reporting is actually vandalism, I usually just give them a vandalism warning for whatever vandalism they were trying to insert. I figure that's the "root cause", and if they knock that off then we don't necessarily need to tip them off that their report was what led to their vandalism being noticed. 28bytes (talk) 20:28, 8 June 2011 (UTC)

Filter 402

(New article without references)
Apparently this filter is tripped when a redirect is created. A redirect does not need a reference, so can the filter be tweaked so that it does not trip when the #REDIRECT command is used. Mjroots (talk) 10:07, 22 June 2011 (UTC)

The filter already checks for "#REDIRECT" and ignores the page if it contains that text. Could you perhaps provide an example of it tripping on a redirect? Reaper Eternal (talk) 10:32, 22 June 2011 (UTC)

I may have been mistaken in thinking it was a redirect rather than a shipindex page. {{Shipindex}} has just been added to the ignore list per my post at Wikipedia:Edit filter/False positives/Reports. Mjroots (talk) 10:41, 22 June 2011 (UTC)

Yeah, I added it. :) Reaper Eternal (talk) 12:01, 22 June 2011 (UTC)

ENGVAR tagging filter

Possibly this has been brought up before, but is there any interest in a filter that would tag changes between accepted varieties of English? It obviously wouldn't be comprehensive, and would probably just look at small edits by new users, but I think it might be useful. Any thoughts on this? Would it be too slow? Feezo (send a signal | watch the sky) 12:05, 26 June 2011 (UTC)

You wouldn't be able to (easily) disambiguate between changes between "switching between styles of english" and "adding the wrong type of english" or "already had both types of english". (The only way I can think of to do it would make it way too slow.) That doesn't necessarily mean it's not useful, since it would only be a tag filter, but it's something to keep in mind. There is a basic implementation I can think of that will have a few false positives (but probably marks articles that need attention anyway) and wouldn't be too slow. Whether or not it's desirable is really another question that I'm not prepared to answer. --Shirik (Questions or Comments?) 17:53, 26 June 2011 (UTC)

A test filter for date format changes is at Special:AbuseFilter/421. I'm actually interested in seeing how prevalent this is. I think the filter should be okay the way I wrote it, but feel free to disable it if it causes problems. I'll also turn it off if it doesn't get any hits. Feezo (send a signal | watch the sky) 14:28, 27 June 2011 (UTC)

Filter 16

Can someone explain filter 16? It's way too complex for its own good, and it just seems to be spitting out false positives, which is pretty severe when we're talking about "disallow and report to AIV". Additionally it's pretty slow for an ineffective filter. --Shirik (Questions or Comments?) 17:54, 26 June 2011 (UTC)

I'd ask NawlinWiki, as he is the only one who edits that filter and I don't think he watches this page. Considering that it is eating about 7% of the condition limit by itself and the only "true positives" I could find are vandals who would have been reverted anyway, I'd say disable it. Reaper Eternal (talk) 18:09, 26 June 2011 (UTC)

Reading the condition count on a single filter is unreliable at best, and personally I think it's just better to estimate the counts manually. However, there's no denying it's expensive. --Shirik (Questions or Comments?) 04:36, 27 June 2011 (UTC)

It's to catch socks of WP:Long-term abuse/Runtshit, and there is more info in the notepad of filter 17 . I haven't had much time lately to keep up so I don't have an educated opinion of whether it's still doing its job well or not. I know that the reason it's on a throttle is because of the many false positives. —Soap— 11:59, 27 June 2011 (UTC)

I'll turn it off for now - Runtshit is a lot less annoying now that we have revdel. NawlinWiki (talk) 20:52, 27 June 2011 (UTC)

Filter 368

Some vandalism on Wikipedia is by removing a large amount of content, but when I revert the vandalism, it always stops me from proceeding because "I was making large edits when marking it as minor". Can someone modify the filter so we can revert this type of vandalism faster? Bryce53 | talk 01:02, 30 June 2011 (UTC)

I mean, it's getting extremely aggravating now. I am always stopped when I try to revert vandalism, because it is marked as minor by default. Bryce53 | talk 06:23, 30 June 2011 (UTC)

I'd like to point out that it only affects relatively new users (under 500 edits), while you currently have 387. That having been said, I think that we probably should have the filter fixed for this purpose (just like it doesn't stop bots). All what needs to be done is add an other condition, saying !("[[WP:TW|TW]]" in summary). עוד מישהו Od Mishehu 09:25, 30 June 2011 (UTC)

I'm a newbie; I have over 15,000 edits on Wikia. I agree with what you're saying, and any edit marked with TW, or is a clear vandalism revert, should not be stopped by the filter. Bryce53 | talk 10:23, 30 June 2011 (UTC)

I've modified the filter to ignore vandalism reverts. Reaper Eternal (talk) 10:28, 30 June 2011 (UTC)

Should probably extend it to cover WP:STiki and WP:HG then, since those are also anti-vandal tools and rollback has no edit count requirements, only the ability to recognise vandalism and familiarity with policies and guidelines are needed. —James ^{(Talk • Contribs)} • 2:45pm • 04:45, 2 July 2011 (UTC)

I've already changed it to ignore vandalism reverts; see my post above. :) Reaper Eternal (talk) 02:36, 3 July 2011 (UTC)

Filter and wiki philosophy

Hi, I was victim of a false positive, which is being investigated here.

Here I would like to raise a general point about abuse of such filters. As a long-time IP editor (dynamic, before you ask) I have constnantly been bullied by registered users and admins alike. That sucks, but it's not too bad, since one can always find a way to reason with them.

Bots are different. Computer says no. The end.

Please ensure that the bot rules are up to standards, and where in doubt remove the rule. Thank you. 220.100.14.12 (talk) 00:51, 10 July 2011 (UTC)

Hello, I'm sorry you were affected by a false positive. However, the reason these edit filters are in place is to catch the massive amount of vandalism that Wikipedia constantly sustains. You can see a log of the events at Special:AbuseLog. Especially look at the logs for filters 3, 11, 31, 39, 46, 50, 139, 294, & 384. This will give you a hint of all the vandalism that vandals constantly try to attack Wikipedia with. Cheers! Reaper Eternal (talk) 01:20, 10 July 2011 (UTC)

Note, though, that ClueBot NG will only revert once for each editor per 24 hours on the same page, so if it gets a false positive, just undo it. The edit filter is a different issue, and you may have to report if it says "disallowed". Reaper Eternal (talk) 01:25, 10 July 2011 (UTC)

Thanks, I did not realise that. I did try and undo ClueBot, but the filter blocked me again, so I got depressed. Shouldn't the same rule ("once for each editor per 24 hours on the same page") apply to the auto-filter, for the same reason - namely there exist false positives? That would be good enough. In fact, why can't the auto-filter rules be incorporated into ClueBot? 220.100.14.12 (talk) 01:44, 10 July 2011 (UTC)

In fact, undoing the ClueBot is specifically targeted by this filter rule, which in my opinion should be removed. 220.100.14.12 (talk) 01:49, 10 July 2011 (UTC)

Ah no, that's just a "Tag" apparently, so that's OK. I was disallowed on "Mary is a b*tch", so I tend to agree that this was probably a once-in-a-blue-moon false positive after all. Thanks. 220.100.14.12 (talk) 01:54, 10 July 2011 (UTC)

Filter for Bold and Italic Text experimentation.

Right now I'm trying to work through the test edits/vandalism of the addition of ''Italic Text'' and '''Bold Text''' to various articles. I asked on the Help desk for suggestions and one of them was to come here. I'm not quote sure what criteria are available but a couple of ideas. "No user with less than 500 edits may add a string consisting of ''Italic Text'' to the mainspace" or to be even more specific. "No user with less than 500 edits may add a string consisting of ''Italic Text'' to the mainspace as either the last or the first text in an article." (and the equivalent for Bold Text). Thank You.Naraht (talk) 15:53, 11 July 2011 (UTC)

The consensus at the previous village pump discussion was not to impede these types of edits because the user may be making useful additions to the page in addition to the edit tests. I have a bot running that undoes edits that are solely edit tests. 28bytes (talk) 15:57, 11 July 2011 (UTC)

Edit Filter help

Is there any way to find out all of the Edit Filters set by a specific user? --79.68.110.245 (talk) 13:26, 24 July 2011 (UTC)

Go to Special:Log and type in the name of the user. You can see all of times they modified an edit filter. Reaper Eternal (talk) 12:17, 25 July 2011 (UTC)

You'll probably want to choose Edit Filter log as the log type. That'll give you a query like //en.wiki.x.io/w/index.php?title=Special:Log&type=abusefilter&user=Reaper+Eternal. — Waterfox ^~talk~ 00:10, 29 July 2011 (UTC)

This way seems better (more descriptive): go to Special:AbuseFilter/history and enter the user. That'll give you a query like //en.wiki.x.io/w/index.php?title=Special:AbuseFilter/history/&user=Reaper+Eternal. — Waterfox ^~talk~ 00:12, 29 July 2011 (UTC)