Today was the second day of FOSS. I'll jot down my observations from today.
The first talk I attended was from LWN's Jonathon Corbet. He discussed recent happenings in the Linux kernel. He noted that the development cycle of the kernel was becoming more "professional", with two week cycles for adding features (and four weeks for fixing bugs). Jonathon was very well versed in the new developments. It was refreshing to sit back and here some semi- detailed updates on the kernel. I haven't updated mine since the 2.6.10 days.
Later I attending a basic introduction to TurboGears given by Swaroop. It was pretty well attended. I haven't tried to do much with TurboGears, but this presentation made it feel very similar to the Rails demos I've seen (that's a compliment). I like that the TG developers are taking advantage of existing projects (and not re-inventing the wheel). My biggest issue with TG is the Python 2.4 requirement. How many distros are shipping 2.4 these days? (Gentoo and ubuntu??)
Speaking of Ruby and Python, there appears to be a bit of interest by the conference attendees in these languages. Many people were asking about them and what distringuishes them or discussing their use of them.
After some booth work I caught the tail end of "Google: Powered by Linux". According to the speaker, Arvind Jain, Google is one of "the biggest hardware producers". He said they actually make their own machines. Look out Dell you are about to be disrupted by the Gbox! (They'll probably give them away for free as long as they can monitor the keystrokes ;)). Their machines also have a failure rate of 1 in 1000. (For every 1000 computers 1/day will fail). But they don't care since the Google File System provides them with redundancies. Next followed some discussion of their grid abstraction layer (that allows programmers to easily develop grid applications using GFS, GWS (Global Work System) and Map Reduce). Supposedly these tools make it dead simple to create very scalable apps. (Perhaps ActiveGrid can provide the common folk with similar functionality, since Google's stuff is proprietary and only used by Google).
Arvind also mentioned something that they will be introducing in the future called "Big Table". This is their homegrown distributed database. That's about all the details he gave. (No it doesn't use flat files).
Other interesting google snippets: In a two week period of Aug 2004, there were 26,000 worker deaths (so I guess you can estimate the size of their cluster(s) by that). Also, people have forgotten how to spell since 10% of queries are mispelled.
The last talks I attended were the end of Ebuild 101 and using Gentoo in Production. The takeaways here, are that there is a lot of power in portage (and ebuild scripts) that is pretty easy to access and that yes, people actually do use Gentoo in production systems. BUT, unless one knows what they are doing it is "not recommended". One the other hand if you do know what you are doing, it can make administration very simple (especially for large numbers of boxes) with no downtime. I'm very impressed by Gentoo. I think it is a shining example of open source well executed (even the documentation is superb).