Talk:Community health initiative/Allow CheckUsers to set User agent-based IP Blocks

The WMF's Anti-Harassment Tools team would greatly appreciate user feedback on this proposed feature. Specifically around:

1. Does this fit into the workflows of Wikimedians with CheckUser permissions, or is this cumbersome?
2. Does the placement at the bottom of Special:Block make sense? Or should it be moved next to the username / IP address field? Or somewhere else?

Thank you! — Trevor Bolliger, WMF Product Manager 🗨 23:27, 26 March 2018 (UTC)

This could be useful in dealing with disruptive sock puppets who are active on busy IP ranges. I don't really have any opinion on placement; the mockup looks fine to me. Obviously, regular expressions will make this much more useful. One problem with that, however, is the complexity of regular expressions. I wonder if there's an easier alternative. Google's web search would not have taken off if it required a computer science degree to find useful results. I guess you could assume that checkusers will probably have at least a passing familiarity with regular expressions, though. NinjaRobotPirate (talk) 02:36, 27 March 2018 (UTC)

@NinjaRobotPirate: We'd love to add Regex if UA blocking proves useful. I've created phab:T190889 to track this. I think a simple regex would probably work, such as just including 'Chrome' would block all users with 'Chrome' anywhere in their UA string. It'd need to be handled with care, just like any other IP range block. — Trevor Bolliger, WMF Product Manager 🗨 23:15, 27 March 2018 (UTC)

I think your example is a good idea. Checkusers with more experience than me might have interesting stories about times when they might have needed a more complex regexp, but I can't think of anything personally. NinjaRobotPirate (talk) 00:24, 28 March 2018 (UTC)

The requirements currently state that all such blocks will be displayed as blocks of a user (registered or unregistered). So the idea is that you can only block the latest user agent used by the user and then an autoblock is triggered and logged for any user sharing the same user agent, similar to the IP-based autoblock?

It's important that the blocks be visible to all users, even if the details are not. The blocked users should also be provided with information on how to discuss the block; the block number may be enough, but mentioning their current user-agent would be friendlier for innocent users. --Nemo 07:00, 27 March 2018 (UTC)

@Nemo bis: UA blocks will only be possible for IP or IP range blocks, not usernames. Autoblocks only function with username blocks, but our team is building cookie blocking from IP users in phab:T152462. These cookies will only drop with the same logic of "if a user cannot edit a page, drop a cookie."

In this implementation of UA blocking, CheckUsers will have to paste an exact user agent into a text field. They will likely chose the UA associated with the most recent edit, but could also select a different UA if they so desire.

These types of blocks will show the same message to the blocked users. The default block warnings show avenues to request unblocks and most wikis appear to have updated the block templates to include more details on how to request an unblock. We will be double-checking these block templates to make sure they are all informative and accurate. Does that address your concern, or did I misread your comment? Thank you. — Trevor Bolliger, WMF Product Manager 🗨 23:15, 27 March 2018 (UTC)

Support for regex

Thank you for your work: this tool can be very useful, and completely fits the workflow of checkusers.

I reckon that the support for regex is quite important/urgent, because often LTA update their browser, but not the OS. So, it would be very useful to block a vandal by referring to a partial match of the UA (e.g. "Mozilla/5.0 (Windows NT 6.1"). --Ruthven (msg) 15:59, 30 March 2018 (UTC)