This is a day-by-day account of the 2000 Ottawa Linux Symposium. Originally this was a set of entries in my diary for July 2000.
Wed Jul 19 2000: My cycling time is getting better, only 15-20 minutes to go around my usual loop. Caught the 3:30 train to Ottawa, and saw stations such as Valois, Pointe Claire, Beaconsfield, and Dorion, all places I rode through hundreds of times when I was commuting to work at CS&T. (If you get the feeling that my visits home are lengthy nostalgia trips, you're right...) After arriving in Ottawa and checking in to the hotel, I went out and about, in search of something to do.
"Something" turned out to be having supper at Darcy McGee's pub, named after one of the few (perhaps the only?) Canadian politician to be assassinated, back in 1868. Just two blocks away, a one-man play called Blood on the Moon, about McGee's assassin and his trial, was running at the National Arts Centre, so I decided to go see it, though the Haunted Walk that starts from the pub was equally attractive. Pierre Brault, who wrote the play and is the only performer, did a good job of playing James Whelan, the accused assassin, as well as the parade of witnesses at the trial. The intent of the play is to argue for Whelan's innocence of the crime for which he was hanged, and the weakness of the evidence against him. I don't think the play succeeded -- it's quite apparent that the play is slanted in the right direction, biased enough to raise doubts about its historical accuracy. I should probably read a biography of McGee or Whelan, in order to get a more neutral account.
Thu Jul 20 2000: OLS, day 1:
I went out around 7:30 for
breakfast, and got it in a little deli where the server had apparently
just emerged from a coma, judging by his groggy hangdog expression.
Strolled past the War Memorial, as seen to the right.
The first day of talks was rather disappointing; I could only find two that interested me, Ted Ts'o's general overview of filesystems and Stephen Tweedie's ext3 talk. Like talks about new Linux ports, filesystem talks sound more interesting than they actually turn out to be. Tso's overview was reasonably good; he classified FSes into disk and network-based, and disk FSes were subdivided into FAT-derived, FFS-derived, and log-structured. Tweedie's talk covered the general organization of the ext3 FS, which adds journalling to ext2 through an abstract journalling API. I find the required changes worryingly complex; the buffer cache needs to preserve the ordering of blocks. This means the VM needs work to handle the increased memory pressure because it can't write out any random block. Deletion poses lots of hairy problems for log recovery and requires careful contortions.
Between those two talks, I skipped out and wandered along Sussex Drive. Initially I wanted to go see the Canadian War Museum, but it charges admission and I was unwilling to pay for only an hour or two. Instead I went to the National Gallery, since the permanent collection is free of charge, and it provided the most memorable sights of the day.
First I visited the Inuit section in the basement. Most pieces of aboriginal art leave me cold, and it was the same today... except for a few pieces that really attracted me. (Apologies for the glare from the glass cases in the following photos.)
I was awestruck by Kiawak Ashoona's Bird Creature, a beautiful and menacing sculpture:
Happily postcards of it were available in the museum store, so I can hang on to its image. Two more nearby sculptures were also noteworthy; I had no notebook, and so couldn't write down the titles or artists. Probably I'll go back tomorrow suitably equipped. The picture with the cross is apparently a metaphorical statement about the Church treating native shamans as demonically-inspired; a shaman is writing around the cross which has impaled him, and his forked tail is just visible (the white arc just to the right of the base of the cross).
Afterwards I looked through the Quebec art; dreary stuff, all boring portraits, unremarkable landscapes, or ecclesiastical kitsch. (Oh, there was one interesting portait of an Indian woman.)
The second stunner was Esther Warkov's House of Tea, an exhibit in the contemporary art collection. No photography allowed, and I couldn't find a postcard or other image of it, so I'll have to describe it. The piece consists of a coffin (or is it a steamer trunk partially packed for a voyage?), and suspended above it is a bundle of rolls of crepe paper, arranged to resemble a female torso. A few branches form a skeletal hand on one arm, leading one to assume it's a corpse. The trunk contains objects whose meaning isn't readily discernable. A cutout mask. A few papier-mâché insects. More papers. More rolls of crepe paper. As in a Greenaway film, the purpose of it all is indefinite though undeniably grim, the classification scheme in use (for this definitely is not a random collection) is mysterious, and tiny details are packed with meaning.
Fri Jul 21 2000: OLS, day 2, and Dad's 71st
birthday: A better day for talks. I attended Andi
Kleen's talk on Linux network programming, which went well despite yet
more microphone problems. Kleen disposed of the usual
select()/poll() choice in a single slide
before turning to the more advanced option asynchronous signal I/O,
which I first learnt about at last year's OLS in Zach Brown's talk.
Kleen also covered the abstract namespace for Unix domain sockets. (A
question I didn't get around to asking was: how is security handled?
Is access to sockets in the abstract namespace controlled by the usual
filesystem access rules?) The talk also discussed the netlink layer
for communicating between kernel- and user-space, IPv6 support, and
various Linux-specific interfaces for doing path MTU discovery.
Informative, overall, though it's unlikely I'll ever need to use them
for any application I'm likely to write.
The next block of time had nothing interesting, so I skipped out and headed for the Canadian War Museum. In about 2 hours, I couldn't go through the museum very carefully, but many of the exhibits are boring, medals and clothing and random weaponry which didn't interest me. I wanted to see things which gave a more visceral impression of war's hardship and pain, and found quite a few things: telegrams sent to inform families of a soldier's death, a woman's pictures of a fiancé who died in WW I, the claustrophobically small confines of a trench. Most movingly, there was a letter from John Payne, a 22-year old private in the Winnipeg Grenadiers who was a prisoner of war in Hong Kong. The night before making an escape attempt, he wrote a reassuring letter to his family saying he thought he stood "a jolly good chance" of succeeding, and in the second paragraph wrote:
So just in case I shouldn't make it you must remember that according to our beliefs I have departed for a better place (I hope) although it will grieve me to exchange the guitar for a harp, even though there is a greater percentage of gold in the latter.
I can recognize the sense of humour in that letter, such as the ironic undercutting of an otherwise serious statement; I could have written it myself if I used many more British mannerisms than I do. (Or was a character in a Biggles novel.) The accompanying panel completes the story, telling us that the next day Payne's escape attempt failed, and he was recaptured and summarily beheaded.
Back at OLS, Mike Shaver's "Open Source Practices: The Second 90%" was reasonably interesting, but didn't say anything terribly new. Perhaps the best talk so far was Andrew Tridgell's talk on the rsync algorithm, which was a good, meaty talk that covered the algorithm in close-enough detail so that writing an implementation would be straightforward, but not so much detail that the audience would get lost. The many fiddly details of rsync'ing directories were mercifully skipped. Tridgell then talked about various applications using librsync that are being developed, such as using rsync over HTTP to save bandwidth.
The last software-related event of the day for me was the DocBook BoF. I went to the BoF because eventually Fred Drake will translate the Python docs to DocBook, which means I'll have to learn it. As of last year my impression was that only the immaturity of the tools was holding things up, so I went to the BoF to see how things had changed. They haven't, really; there's no GUI tool for writing documents, the DocBook-to-whatever scripts were described as mediocre and not attentively maintained, and whatever-to-DocBook translators are non-existent and probably wouldn't produce good results even if they existed. Guess I can keep ignoring it until Fred actually makes the switch.
Took the Haunted Walk tour in the evening. While it was reasonably interesting, most of the ghost stories are of the tediously real variety, with nothing more definite or frightening than lights that turn on and off, or electrical devices that mysteriously malfunction. Also, I don't know if the tour guide was inexperienced or, more likely, if there's no well-defined script, but from a literary point of view the stories weren't told very well, and I spent much of the tour rewriting them in my head. The stories jumped around chronologically, often brought up the most frightening details right at the beginning, often had no glimmer of a motive for the haunting, and didn't really come to any sort of climax. Fun, but not as much fun as it could be with a well-written script.
Sat Jul 22 2000: OLS, day 3: Showed up at the Congress Centre a bit after 9 AM. There was a large party of Indian people milling around near the entrance, apparently for a wedding held elsewhere in the centre, and all dressed in beautifully elegant silk and gold clothing. Inexplicably there was also a bagpiper there, pacing back and forth, and he began practicing a piece as I headed down the stairs to the OLS area; it sounded great, though the apparently hung-over person manning the reception desk wasn't enthused by the sound.
Went to the virtual world talk because I thought it might be applicable to using VRML for sharing CAD displays and simulations over the network, something that we'd like to do, eventually. FreeWRL, maintained by the speaker, may be worth looking at, though the impression I got is that the code is rather amorphous. Next was a talk on power management, of interest because the management on our laptops still isn't working properly, so I wanted to know a bit more about it. A good talk about a specialized topic.
Checking my e-mail, I found Guido's posting to python-dev explaining the outcome of yesterday's Python consortium meeting; it's pretty favorable, despite some annoying things such as a longer CNRI licence (don't sweat too much; most of the added terms are irrelevant legal trivia) and the 1.6 release, which will essentially collect up all the work done at CNRI, represented by the state of the CVS tree at that time. Because moving to SourceForge caused Python development to accelerate tremendously, a lot of useful fixes and changes have been made since that snapshot of the CVS tree. My suggestion would be to simply ignore the 1.6 release, and wait an extra month or two for 2.0 release including all the nifty fixes.
Next I came within the range of the GNOME reality distortion field yet again, by seeing Chris Lahey's talk about the GNOME Canvas. I swore not to let this happen to me again, but it was the most interesting talk in that block... Lahey demoed Gnumeric, Sodipodi, and talked about the canvas's capaibilities; it looks really powerful and cool, though I wonder if I'll run into the usual problems trying to get a GNOME development environment working. David S. Miller's keynote, a historical reminiscence about his involvement with the Linux kernel, was highly amusing. Hey, he started with 0.99pl5, one patchlevel before I did!
After going to buy a few souvenirs, I met my parents and we drove back home to Hemmingford.