QUOTE(EricBarbour @ Sun 19th December 2010, 10:25pm)
If anything destroys Wikipedia in the long run, it might be this attitude. For example, if a developer far smarter than anyone at WMF devised a wiki-like framework that made Wikipedia look like Windows 3.0, that could be the death knell for WMF's "developer system" (if you can dignify it with such a name).
Just hacking MediaWiki to use HBase as a back end would likely yield a substantial performance improvement. (
Edit: Apparently Mahalo has already done this.) That, and writing a parser engine that doesn't suck (but that is hard to do while remaining compatible with the abortion that is MediaWiki markup). However, the real problem here is that, by and large, people have come to realize the rather substantial limitations of wikis, and so the only "large wiki" out there is, and will always be, Wikipedia.
There are some pretty sweet looking public MediaWiki installs out there (
This Might Be A Wiki is one of my favorites), virtually all of them are either extremely tightly controlled sites that are just using MediaWiki as a quick and dirty CMS (public editing is either disabled entirely or very tightly controlled), or else wikis that service an already-established community with norms, standards, and community discipline practices already in place. And none of them come even remotely close to the database size, view rate, or edit rate of even a midsized Wikimedia wiki.
The real gruntwork for wikis today is for internal communication; if you look at the extensions registry you'll see that's where most of the work on MediaWiki is going. Very little of what comes out of that activity ends up being useful for Wikimedia. That's the problem with Jimmy's "parasitic milkman" approach to development: Wikimedia's needs for the development of MediaWiki are pretty much
sui generis to them; the only way they'll be met is if they push the development themselves, which they are too cheap to do.