Wikipedia:Pending changes blocks

A pending changes block is a proposed technical editing restriction imposed on a user, that is to traditional blocking what pending changes protection is to traditional protection. Any edit by a pending changes blocked user would not be visible to readers, it would be pending until reverted or reviewed, as if the edited pages were on pending changes protection but such protection would disappear after the revert or review is done. This permission would be assigned to administrators as well as users in a new usergroup, and may be made available to the edit filter.

Pending changes blocks may be imposed on new or unregistered users performing edits that are vandalism, spam or BLP violations, provided the criteria for a traditional last warning are met, at minimum. This means that pending changes blocks can be used one step ahead of traditional blocks by a larger group of users, improving our reactivity and our mitigation of vandalism. For dissuasive effect, the pending changes blocked user would be warned that their edits are now reviewed before being published, and that further inappropriate edits will result in a full block. Any edit by a pending changes blocked user would be reported at AIV by a bot.

Rationale
The rationale for introducing this new technical tool is twofold. First, the excessive amount of time it takes for administrators to block the users reported for performing those unacceptable edits, a concern that is only amplified by the widely held view of a shortage in administrators. Second, the criteria for blocking are quite strict, with a requirement of several warnings in most cases, in the meantime the damaging edits cannot be mitigated preemptively, except when picked up by edit filters set to disallow, but those are often evaded. In addition, since the edits are not saved, they are not apparent to recent changes patrollers, and so users can be unhindered in their attempts for long periods. Among users reported at WP:AIV, both the total number of edits performed between last warning and AIV report, and performed between AIV report and block, are significant. The noticeboard is chronically backlogged, some users have enough time to makes dozens of edits, and sometimes are blocked only hours later when they are no longer active. Some of their edits remain visible for several minutes, up to an hour or more on occasion. Had those users been pending changes blocked by non-admins, their edits would not have been visible to the public. In addition, recurrent sockpuppeting vandals and spam bots could be immediately pending changes blocked. The edit filter cuts a substantial portion of vandalism, but most of the flagged edits are allowed due to significant false positive rates. They are tagged, but many can still be missed, sometimes for days. Deferring those suspicious edits for review would ensure that they are taken care of. In addition, it may reduce the number of edits allowed per minute (8 for new or unregistered users). The situation would therefore be much improved by providing this mild form of blocking to non-administrators, anti-vandalism bots and the edit filter.

Impact analysis
The impact on the backlog of unreviewed edits should be small. The simple fact that the user's edits are no longer published without review should be quite dissuasive on its own. If a pending changes blocked user persists, any edit would be listed at Special:PendingChanges and tagged, and the user would be immediately reported at WP:AIV by a bot. In contrast, on pending changes protected articles, despite being among the articles most subjected to this type of edits, a very large portion of all edits are still acceptable, and they end up forming the bulk of the backlog. The pending edits that are vandalism, spam or BLP violations are dealt with quickly. At the moment, most pending changes are reviewed within one hour, and it takes very rarely more than a few hours, this small increase can be managed.

The impact on general editors should be minimal. Unlike pending changes protection which remains on articles continuously for significant periods of time, and thus is visible and potentially interfering during that time, the ephemeral nature of the disruption caused by pending changes blocked users ensures a very limited distraction.

The impact on the workload of administrators should be positive due to the dissuasive effect of pending changes blocks, meaning less need for traditional blocks. While administrators would also be involved in the occasional promotions (or demotions) of users to the new usergroup, those would form a very small part of their burden compared to their regular duties.

The impact on user participation should be negligible. Since the use of this feature is straightforward and restricted to obvious vandals, spammers and such, misuse or abuse potential would be small. If it does happen, the user would still be able to edit and this would come to the attention of an administrator due to the AIV report. Further, the vast majority of contested admin blocks are of established editors, but it would be technically impossible to pending changes block autoconfirmed users. Pending changes blocks of IP ranges would also be impossible.

Policy issues
Pending changes blocks may only be used for vandalism, spam or BLP violations, provided the user meets the requirements for a traditional last warning before a traditional block, or for a traditional block as determined by the blocking policy. Pending changes blocking thus serves both as :
 * a preventive and dissuasive measure that can be applied before the requirements for a traditional block are met,
 * in case of persistence of the disruption, or if the criteria for a full block were already met, as a temporary measure until an administrator can intervene.

The new usergroup would have in addition to this new userright those of rollbackers and pending changes reviewers. The community should agree on a name for the new usergroup; suggestions include moderator and pending changes blocker. Administrators would possess all of the userrights afforded to this new usergroup, and could grant it or remove it. The criteria and process for adding users to this usergroup would be discussed later by the community. A good anti-vandalism experience would be expected. It may involve a short community discussion with consensus determination by an administrator. The modalities of removal of this usergroup should also be discussed. A new noticeboard for reports for pending changes block may be created. Pending changes blocked users may appeal on their user talk page, with only administrators allowed to review appeals.

Bots may be assigned to this new usergroup based on a successful BRFA. For example, regular anti-vandalism bots may use it every time they give a last warning or report a user at AIV. The only active anti-vandalism bot at the moment is. A new type of anti-vandalism bot may directly pending changes block any user committing a specific pattern of serious abuse, as a complement to the edit filter.

Edit filter
Currently, numerous filters target vandalism, spam or BLP violations, the user is usually either prevented from performing the edit or warned, in the later case they can still confirm and save the edit, which is simply tagged. Those edits are often not reverted in a timely manner, in addition the users often find ways to get around filters set to disallow. In order to correct those deficiencies, it may pending changes block a user tripping the filters three times, so that edits not picked up by the edit filter are also deferred. It may do so at second hit, or first hit for certain filters targeting serious patterns of abuse, such as those of sockpuppeting vandals or spam bots.

Those filters tag the edit, and for most of them, warn the user, they could possibly be set to pending changes block after three hits :

Those filters prevent the action, but users often find ways to get around the edit filter, so they could pending changes block after three hits or less :

Technical details
The naive implementation would be for MediaWiki to place any article that gets edited by the pending changes blocked user under pending changes protection, put the new edit in an unaccepted state, as well as edits by the same user immediately preceding it (only the last revision by another user would be accepted, so a sort of pending changes rollback), then remove the protection as soon as a new revision is accepted (when the user gets reverted for example). The new revision would thus appear at Special:PendingChanges along with unreviewed revisions from pending changes protected articles. They should also be tagged with a wording such as "deferred for review".

The pending changes block interface would be integrated in the block interface, and pending changes blocks logged with the block log. 'Moderators' would have block links like admins, but would have the 'pending changes block' option checked by default and grayed out (on the other hand, admins would have it unchecked by default), and possibly the option to disallow account creation for registered users only (enabled by default). There would also be a notification in the user contribs similar to the block notification, with another color to differentiate it (maybe yellowish).

If a pending changes blocked user edits an article, a notice similar to MediaWiki:Revreview-locked, but more prominent, would be shown but instead would read something like :"Note: Your edits are subject to review (help) for the following reason : ." If the edit gets saved, and thus pending changes protection intermittently enabled, users neither autoconfirmed or confirmed who may edit the article in the meantime would be shown the standard MediaWiki:Revreview-locked with a generic log in details.

Since the pending changes block would no longer have any effect if the account becomes autoconfirmed, it may have to prevent autoconfirmation. Rate limits may be made stricter (for example, 4 edits per minute instead of 8).

The following limits would apply to pending changes blocks by 'moderators', but not by administrators :
 * IP ranges cannot be blocked
 * autoconfirmed or confirmed users cannot be blocked
 * maximum block duration is limited for IPs.

It should be discussed whether to extend the effect to all namespaces, as much as technically possible.

The edit filter already has the ability to block users, it wasn't implemented on the English Wikipedia because of too great a risk of false positives. The extension could be tweaked to support pending changes blocking.