Today at the Chandler staff meeting we were trying something new, little demos of what we've been working on. I was supposed to talk about the unit testing work that I've been doing on the Chandler repository. I attend the staff meeting via telephone, a disembodied voice to the rest of the group, although I would have flown down for today's meeting and a stopover at Macworld ;-). Today there was something wrong with the technology. I could hear what was going on, but nobody could hear me. Bad news for making a presentation. So when it came time for me to talk, and it became clear that the phones were busted, we yanked an IRC channel up onto the laptop being used for the demos and projected it. I then made my remarks via IRC. I've yet to go to one of those conferences where there's an IRC backchannel being displayed while someone is speaking, but I guess this is what happened next. It was weird to be typing and hearing audio reactions while it was happening. I guess if I had thought a little more clearly we could have fired up the iSights, but I don't know how loud the sound would have been. As the price of things like iSights comes down, I hope that we'll see more interesting applications of this technology to decentralized software development. Too bad Apple didn't announce the rumored multi person iChat AV today.

As far as the details of what I talked about: I've got tests for most of the repository API, there are a few areas that still need to be done. This should help us "keep the bar green" as we make additional changes to the repository, and will help us to find repository problems faster. Also, the tests are a decent set of examples for how to use the repository API's.

The other thing that I've been working on is a stress/performance test for the repository. Right now it consists of a bunch of RSS feed data that I downloaded onto disk and then jam into the repository. As a first cut it is interesting because it turned up some bugs. There's also a bit of work left to be done here to make real performance tests that yield commit throughput rates and so forth. We also need to do tests of different usage scenarios. But as of a few days ago, we can read about 700 RSS feeds (~20MB), containing almost 11000 RSS items and commit them into Chandler without falling over. I'm in the process of assembling a much larger set of RSS data - about 20000 feeds, that should get us closer to the state of the repository when someone has been archiving their RSS data.