eekim.com: EEK Speaks

Wed, Aug 03, 2005

WikiMania Hackfest Day 3 #

Today's tidbits: (JLI)

WardCunningham arrived this morning and gamely participated in the afternoon session, despite the long flight. At one point, he asked the MediaWiki developers what the consequences of a wrong architectural decision would be. The answers were revealing. These guys are faced with a tough problem thanks to scaling issues. By necessity, they have to optimize their architecture based on user behavior. However, this severely restricts their flexibility, because if their predictions about user behavior are wrong, it is extremely expensive to rearchitect, reimplement, and reconfigure. Moreover, MediaWiki really can't develop agilely, at least not where new user features are concerned, for the same reasons. If they were part of a large company with plentiful IT resources (e.g. Google), then it would be a different story, but they're not. This also makes ExtremeUsability somewhat of a pipe dream. (JLJ)
Funny observation from a MediaWiki developer (Mark) on downtime: They're the only site that profits from downtime. When Wikipedia goes down, folks assume it's short on resources and tries to donate stuff. (JLK)
At the FLOSS Usability Sprint last February, we framed the following conflict between OpenSource developers and usability practitioners: With usability, the mantra is, "You are not your user." With OpenSource, the mantra is ScratchYourOwnItch. Those two philosophies seem contradictory. However, it's not so simple. Many new features in OpenSource projects are user-driven, although the degree to which this happens largely varies. My impression with MediaWiki from watching the developers work and from talking to various members of the community -- developers and users -- is that the developers are open to feature suggestions, but are not particularly enthusiastic about user-driven design. Developers will implement suggestions, provided they are posted to Bugzilla. That's a poor way of doing things, in my opinion. I'm picking on MediaWiki, but it's a very common attitude in the OpenSource world -- it came up several times at the usability sprint -- and really, in the entire software development community. With MediaWiki, I'm particularly disappointed, because they've got this huge, wonderful user community. As I said yesterday, the purely technical problems MediaWiki faces are fascinating in and of themselves. However, the opportunity to evolve some really cool user features should be just as fascinating. (JLL)
Then again, perhaps MediaWiki just isn't the place for this to happen. As I said above, doing ExtremeUsability would be especially hard, because the high overhead imposed by the architecture makes it difficult to develop agilely. Maybe the place to experiment is with smaller communities, and MediaWiki can steal features accordingly. That, after all, is one of the goals behind the BlueOxenCollaboratories -- explore cool ideas in real communities and teams, and then propagate them so they are stolen far and wide. The only problem with this is that it's almost impossible to duplicate the social effects of scale seen on Wikipedia. (JLM)
I mentioned a "MySQL guy" yesterday. His name is JanKneschke, and while he works for MySQL, he's really here to help MediaWiki with performance issues. One way he's doing that is by advising them on migrating to his high-performance web server, lighttpd (pronounced "lighty"). lighttpd is a relatively new implementation of an old idea: rather than prefork or thread a web server, as Apache does, use a single event loop, which saves you tons of memory and CPU, especially at scale. It's got some other cool performance tweaks as well. Jan is very sharp, and lighttpd seems to be taking off. Another high-profile site that uses it is RubyOnRails. One thing I find amusing is that FastCGI, which has been around forever, seems to be making a big comeback. RubyOnRails, PHP, Sympa, and many other high-profile tools use it. It's amazing how old things can fly under the radar forever, then suddenly find new life. (JLN)
Over and over again, I see that proficiency in one area doesn't necessarily translate into others, even if they seem like they should. Folks in the Wikipedia community are incredibly knowledgable about online, emergent communities. However, judging from the conference design, they know very little about facilitating similar emergent patterns at face-to-face gatherings. Again, I'm picking on them, only because I'm here, but this is a widespread problem, and it further validates what BlueOxenAssociates is trying to achieve. Two things that the organizers here have done very, very right: Bring together great people and establish a wonderful culture of openness. (JLO)

/events | Posted at 4:45pm

Comments

Comments disabled until future notice. If you'd like to contact me, use my i-name (=eekim).

EEK Speaks

A blog about collaboration, community-building, and the various goings-on at Blue Oxen Associates, with occasional digressions on food and other vital matters.