User:Nathan/FlaggedRevs vs. NPP

Flagging, patrolling and other options
The above is a copy of a discussion on WT:RFA. I've reproduced it here as background to my argument that the comparison of FlaggedRevs to new pages patrol is illogical. The systems are not analogous, the implementation and maintenance work required not similar, and the backlog of NPP can't be directly compared to any theoretical backlogged caused by Flagged Revisions.

Flagged Revisions
The purpose of the FlaggedRevisions extension is the protection of quality content against vandalism. FlaggedRevs (hereafter, FR) is a system that requires established reviewers to "approve" additions to articles, either all articles or some subset, before they can be seen by the general reading public. The process of approving revisions is usually described as "sighting" or "flagging" them. Revisions will automatically be sighted if they are made by editors in the reviewer usergroup. All other revisions, either by new users or IP addresses, will need the attention of a reviewer.

There are a number of arguments for an against FR that I won't reproduce. My own position is that biographies of living people are the most serious problem facing the encyclopedia and its public image. Famous people, which are ideally the people who have biographical articles on Wikipedia, are by and large tastemakers. Lawsuits with financial motivation incent litigants to search for a history, a pattern of abuse. Lawsuits + publicity + poor past practices = serious trouble for Wikimedia.

The process for FR is pretty simple - if you edit an unsighted version, as a reviewer, that version becomes sighted. Possibly you can't even edit the sighted version without either approving or explicitly reverting unsighted revisions that don't meet the standard (I'm not sure if that is how it works or not). More importantly, this system applies to existing articles - many of which will be on the watchlist of a current editor. If an article is likely to see non-vandalising editing, it means someone has an active interest in it - this active interest makes it much more likely that the flagging will be kept up to date. It's also important to recognize that the flagging process is not based on an obscure, bureaucratic page listing tens of thousands of unflagged revisions. It's right on your article, built into your editing process. Compare that to NPP, which I'll describe below.

New Page Patrol
New page patrol (abbreviated NPP) is designed to catch bad or deeply flawed new articles as they are added to the encyclopedia, because its highly likely that pages about obscure people, obscure or invented topics or with unusual titles will never be seen again by a regular editor. NPP catches these articles before they sink into the database abyss, and this is a valuable service. The problem, as recognized by many people but most particularly by its most devoted contributor DragonflySixtyseven, is that NPP has huge backlogs.

The only way to patrol an article is by using the new pages log; after a period of time, a few weeks or so at the moment I believe, unpatrolled pages scroll off the log and may never be seen again. Using the log, you pick off a new page (highlighted in yellow for unpatrolled), read its content and click on "Mark as patrolled." Patrol marking does not return you to the log, and typing in Special:NewPages does not return you to your place in the log. If the article is substandard in some way, you tag it with a CSD tag or some other problem template - and that, still, does not return you to the log.

So the process is inefficient, the volume of submissions to check is huge, and keeping up with it requires attention to the specific task in multiple steps for each new page. It isn't, and can't be, integrated into the normal editing process like FR. It does not have a built in constituency of prior editors like many articles do, which facilitates review using the FR extension.

Flagged Protection
It's not clear how FlaggedProtection will actually work, but here's my assumption of the best possible design. In ideal circumstances, FP works like FR - but is applied actively to articles where there is an identified need, as with other forms of page protection. This has the potential to be useless and duplicative of semi-protection; if the standard for becoming a page reviewer in FP is set too low, then it will be no more difficult to circumvent FP than it is for semi-protection. Some edits, a few days, autoreviewer status and bam - vandalism spree. Happens all the time, and only stops those vandals that are the least determined and Wikipedia-savvy (as in, "Ur gay faggot!" geniuses). We do pretty well with those folks already with just Huggle.

If the bar on reviewer status for FP is set high enough, say 1000 mainspace edits (in no more than the last 12 months) and 3 months of editing, then it will be helpful for all of the identified problem articles. It doesn't help the thousands of uncategorized or poorly monitored BLPs out there, nor will it for the tens of thousands or hundreds of thousands of others in the same boat that aren't explicitly BLPs.

Patrolled revisions
This option offers all the work of FlaggedRevisions and none of the payoff. I'm not sure whose idea it was, or why anyone thought it was a good one, but it seems like a monumental waste of time and a bit of a shadow game. We get to go through all the agita of flagging every revision of every article, but not having done it won't make any difference - the vandalism and defamation edits will still be immediately public. We also get all the inevitable bureaucracy and problems associated with selecting reviewers. (Arguments that the process will be straightforward and not controversial fall flat for being obviously silly - many proponents of the rollback usergroup said giving and removing rollback would be no-drama too. Since then, we've had countless arguments on the rollback requests page and one desysopped administrator.)

With this option we will have huge backlogs - if only because no one will really see the point in keeping up with them.

Current poll (March 2009)
The current poll switches on Flagged Protection and Patrolled Revisions. In my mind, this is totally counterproductive. It grants the illusion, comforting for some, of having taken definitive action. It comes in the form of a trial, but the poll was initiated before any element of the trial itself was established. It proposes an implementation of flagging and patrolling that will ultimately cause more harm than good - by creating a great deal of work without solving any problems it will give the unavoidable impression that FlaggedRevisions (in true form) is not worth doing.

It's been supported by some with the argument that anything is better than nothing - while I wish this were true, in this case it isn't. If the 'anything' that sneaks through turns out to be (1) controversial because of a poorly attended poll (2) composed mostly of meaningless makework and (3) burdened by bureaucracy that the makework and its limited impact can't justify, then the whole thing will set the concept of FlaggedRevisions, and thus the effort to improve our treatment of BLPs, back by miles. I've opposed the poll for this reason, and hopefully it won't make it through the WMF gatekeepers of major en.wp changes.