MediaWiki talk:Abusefilter-disallowed

Abuse filter suggestion
Hi, currently the message given to users when they hit an abuse filter is MediaWiki:Abusefilter-disallowed. This in my opinion is rather unhelpful, because I can see a new user getting hit and not knowing where to go from there. I suggest adding a link to this page (or somewhere where a user can easily seek help). To contrast, here's the same message on Wikibooks (which admittedly isn't that much better, but that's because we often make use of custom messages). The same applies to MediaWiki:Abusefilter-warning BTW. Leaderboard (talk) 14:48, 6 November 2023 (UTC)


 * Would it be worth including the offending string so the message becomes e.g.:
 * MediaWiki:Abusefilter-disallowed "bogus.example.org"
 * Would that be helpful to block evaders as well as to new users? Pashley (talk) 15:53, 6 November 2023 (UTC)


 * Offending strings are sometimes best kept secret. Giving advice on how to proceed if the edit was in good faith would really be helpful. Ideally We would give the advice as the edit turns up in the filter, but it seems we are quite bad in monitoring these logs. –LPfi (talk) 18:41, 6 November 2023 (UTC)


 * I think we had this discussion before. I don't remember where, and it seems nothing got done. Or was it another message? –LPfi (talk) 18:42, 6 November 2023 (UTC)
 * I thought the reason for the disallow message was to ensure sockpuppets wouldn't know why their edits were disallowed. If they know what identifying edits to avoid as would be the case with a transparent filter, they could avoid the filter and go undetected. --Comment by Selfie City (talk) (contributions) 18:47, 6 November 2023 (UTC)
 * Yes, exactly. If we tell that "Scunthorpe" was blocked as "indecent words" then that might hint on what we are blocking. It is better that just the admins check for the problem, by themselves or when alerted. (Some filters are indeed transparent, such as the "blanking by new user" one). –LPfi (talk) 20:12, 6 November 2023 (UTC)
 * I vehemently disagree with not writing the offending string – it's a bad user experience for the vast majority of innocent users who usually get caught up in such a filter and isn't how most warning messages are written (except notably Apple's message when your Mac restarts – which reads like "Ooh woo, welp, something went wrong, but we don't care the slightest."). -- SHB2000  (talk &#124; contribs &#124; meta) 11:38, 7 November 2023 (UTC)
 * I'd be in favor of providing some information, but if you think about a couple of long-term abusers, you hardly want to say here's the exact regex we use to detect you, and oh, look, if you double the first letter, you'll be able to post your nationalist spam without triggering this filter. In some cases, the information provided might need to be the name of the filter, as in "If you believe your contribution was valid, please tell an admin that the 'blue-green widget filter' was triggered." WhatamIdoing (talk) 17:15, 7 November 2023 (UTC)
 * Not even that is needed. If you notify us under your user name, we can check the abuse filter link of your user contributions. We get much more detail than any non-admin user can give. They can of course tell us that Scunthorpe is a real place, if that's the issue. I know it is frustrating when one gets error messages without info, but e.g. for login failures, that's usually necessary not to give an attacker hints on what to exploit. What we can do is to point them to the Pub and assure them that we care. –LPfi (talk) 17:24, 7 November 2023 (UTC)
 * , I get that - all I am recommending is that you provide a valid link so that a new user that got caught can seek help (such as this page). Leaderboard (talk) 11:03, 7 November 2023 (UTC)
 * Yes, I answered to Pashley's suggestion. Let's see if somebody has any comment on yours. Pointing to the Pub would be easy, and I think that's the appropriate forum – technical details can be discussed privately among admins, but there is no reason not to raise the issue in public. –LPfi (talk) 11:18, 7 November 2023 (UTC)
 * I think we should have much softer language, directed at the false positive cases. We can afford apologising that we disallowed vandalism, when the filter hits as expected. "Identified as harmful" suggests that the edit indeed was bad. What about something like "Our filter assumes that this action was harmful, and therefore disallowed it. If you think it isn't, […]"? –LPfi (talk) 18:10, 7 November 2023 (UTC)
 * (I think a native speaker should suggest a suitable wording. At least tonight I cannot come up with anything good.) –LPfi (talk) 18:12, 7 November 2023 (UTC)
 * Yes, an apology would be good – since it reassures the user we actually care. -- SHB2000  (talk &#124; contribs &#124; meta) 20:39, 7 November 2023 (UTC)