User:Tbc~metawiki/Wikipedia configuration management

From Meta, a Wikimedia project coordination wiki

See also: a second call for configuration management at Wikipedia-L


13 September 2001

There was a thread on the Wikipedia list that got me thinking about configuration management for Wikipedia.

I agree with the need to accommodate Wikipedia in citations. I also agree with the need to somehow "freeze" articles.

So here's an outline of my proposal:

  1. Wikipedia is, first and foremost, a perpetual work in progress. I think the w:September 11, 2001 Terrorist Attack article is evidence of the value in this principle. Wikipedia contributes to the universal body of human knowledge, offering the unique benefits of Wiki technology: global participation, hypertext, up-to-date information, etc. Others may want to flesh out this mission statement. We must never compromise this key contribution of Wikipedia.
  2. Wikipedia also strives to be academically relevant by enabling periodic releases of encyclopedia versions. The period may be much shorter than a conventional encyclopedia, but most information providers seem to be settling on annual updates. This also builds on the strong tradition of hardcopy w:almanacs. The 2001 edition of the Wikipedia needs to get out soon if we go this route.
  3. Wikipedia articles can be versioned even more often than the whole Wikipedia. We already do this thanks to UseModWiki, but the calls for "freezing" articles are begging for a baselining system where an article can, to some degree, be called "ready." They'll never be "finished," but many are "ready" for "release."

So that brings me to my overall vision. I work in the field of software process engineering, and from my perspective Wikipedia needs a w:configuration management system. We have a w:revision control system for individual articles, but the 10,000 article mark seems a good point to release our first version of the Wikipedia. Eventually (maybe very soon?), Wikipedia will need a supervised release process.

This model is analagous to other w:open source projects. For instance, w:Debian released version 2.2r3 on April 17th, 2001. To be a part of that release, developers of all parts of w:Linux had to "get on the release train," as it's sometimes called.

Here's a start to get discussion of Wikipedia's configuration management system going. Current URLs at www.wikipedia.com would continue to reflect the current understanding of the "wild and wooly world." Articles are "live." But some new links can be added to each article. Every "released" article could have a link to view diffs compared with older released versions of the article, not just the "older revisions." We still show visitors the live copy, but we have a link to "release 2001" or release "September 2001" and so on. These named releases are frozen for all time. So the world will know what Wikipedia in 2001 had to say about the universe. There isn't an "edit" link on released pages, but there's a "visit the live, ongoing work-in-progress version of this article," which still has the edit link.

I still have a lot of vocational work done today, so enough avocational time for today. Talk amongst yourselves...

tbc


I think the idea of "relesed" versions of Wikipedia is good, and the released versions would be very nice to send along with Linux distributions. If they were slightly rigged, so that you can browse it locally, but get redirected to the main Wikipedia site when you try to edit, it would probably draw people here as well. --Pinkunicorn


Is the proposal simply that we release sets of articles that are "approved" (somehow) on a periodical basis? We can't do that until we have set up an approval process. --LMS

Well, I tried also to describe how we present "released" versus "in progress" versions on the wiki, but I guess you've summed it up. We need an approval process so we can get on with the work of releasing articles for the 2001 Wikipedia. After reading about how Eric Weisstein evolved MathWorld, I now wonder if I'm too optimistic. Maybe Wikipedia isn't mature enough to release a 2001 edition. But I guess we'll never know how close we are to a "release" until we establish criteria for when an article is "done." tbc