Summer of Code 2011/Mentors' Summit

Here be dragons (pink & plush)

Max's notes

Code review

Yeah, we're deep in shit. Some projects that have zero paid developers nevertheless have review backlog for both commits and submitted patches no longer than 1 day - e.g. Drizzle. One idea, already implemented by don't-remember-whom, is that NOBODY COMMITS ANYTHING WHILE THERE ARE UNREVIEWED COMMITS. Sounds a bit OTT, but could be implemented in a slightly less extremist variation, e.g. "... while there is unreviewed stuff older than 3 days".

Mentoring strategies

Conclusions we reached at the NOLA Hackathon look just right, others also suggested some kind of ice-breaking events, culminating in plans to have all students and mentors play MUD together for several hours. Overall, different people and projects used quite different approaches, some going as far as reqiring all students to post daily reports to their main dev list.

Comparison of wiki architectures

I came to MoinMoin 2.0 talk mostly to troll :P, but all the other visitors were developers of other engines with the same or similar intentions (:P), so we actually had an interesting talk about how each of us does this or that. Summary of interesting stuff others have:

• Unified blob storage: if it has MIME type of text/x-wiki, it's page text, if image/jpeg - it's an image. We store these things separately and independently for a number of good reasons though.
• Support of different parsers: per-wiki or per-page(?), some even claim to support MW wikitext. People are so naïve, even we can't support it properly:P
• Regex-based iterative search engines rocked on ten pages, but had to be ditched to support something larger.
• Some people think that RESTful URLs make sense:P

Distributed testing

On this talk, we discussed a strategy opposite to our plans of creating a one big virtualization cluster. Instead, people who can't afford it runs tests on volunteers' machines to cover different platforms/environments.