Well, at least Friday was a pretty full day, starting with the keynote, Anthony Baxter's "One Snake Enter, Two Snakes Leave" covered the two upcoming Python releases: 2.X (a minor stability/feature release) and 3.0 ("the release which will break all your code").
I hadn't seen him give a technical talk before, only the talk he'd given on flashy talks last year at the LCA Speakers' Dinner, and I was curious about how well his style worked for a real talk. Very well, as it turns out -- he was entertaining, clear and still plenty technical. The video of the keynote is well worth checking for anyone who programs in Python and needs to know about the upcoming changes.
Next up was Ralph Giles' "Seeking is Hard", an explanation of the Ogg container format (as he recovered from running across campus to find a needed video adaptor to get his Mac to talk to the projector). I got a little lost in the discussion early on distinguishing packets from pages (someone asked what the motivation was for each, and that would have helped me too).
But the core of his presentation -- why seeking is hard (for a media format that has to encompass video as well as audio) -- was clear and interesting. Seeking means finding a file location corresponding to a specific time offset; Ralph discussed the difference between seeking to a file position directly proportional to the time (which works only in uncompressed formats no one uses any more), using a seek table (a good optimization, but they're often wrong so you can't count on them) and the real solution, putting timestamps in each page. He covered problems like keyframes (a video frame from which a set of subsequent frames are calculated, so you can't seek and then start playing right away; you have to search backward to the last keyframe) and multiple tracks (you have to seek in each track to get them all in sync before starting to play).
Quite interesting, and I understand video formats a little more than I did before (which was "not at all").
Of course, you have to laugh at the title of Matthew Garrett's talk: "Suspend to Disk: Why it doesn't work, can't work and never worked in the first place (and what to do about it)." And we kept laughing throughout the talk. Who knew that kernel swsusp was such a funny topic? But the talk was informative and detailed as well as funny ... a strong contender for best talk I saw at the conference.
After lunch, Keith Packard of Intel told of "Pain and Redemption on the Linux Desktop." At the beginning of his talk, Keith announced Intel's release of a Programmers Reference Manual for their graphics chipsets -- some 1700 pages of detail used in their current driver, all released under a Creative Commons license (no derivative works). Horray, Intel!
The meat of the talk was a discussion of problems with the current X model, and fixes for them, including lots of information about who was working on what. Sort of a "state of the server address".
[ 23:44 Jan 31, 2008 More linux/lca2008 | permalink to this entry ]