Found via a comment on the the excellent Rands in Repose weblog: Keith Bostic on Technical Management. Extreme summary: good management practices are even more important for virtual teams, such as you typically find in open source development.
One example is the mission statement. It is classic Dilbert fodder. Nevertheless, I think he's right. Large corporations usually do awful mission statements: gobbledygook wrapped up in irrelevance imposed from on high. Bostic is talking about mission statements that the people doing the work find useful and motivating. (I'm thinking the dreaded words "buy in") In the ideal case, a mission statement is a rule of thumb that can be applied quickly: is this something I should spend time on?
Time is the theme running through the whole talk because time is something that open source projects are typically even more short of than commercial equivalents. Volunteers cannot, by and large, give eight hour days (never mind ten hour days) week after week. Working remotely slows everything down, partly because e-mail is inefficient as a communications medium and partly because time zones add hours (sometimes days) to turnaround times.
It's not all perfect, but I wasn't too distracted by the parts where he got it wrong. Making predictions is dangerous. There was no way to know that Javascript would turn out to be mostly irrelevant. There's more than enough material here to make up for it.