Older blog entries for amk (starting at number 24)

For interested readers: the PyCon conference has now posted a draft of the schedule for March 26-28. (No abstracts, though.) Paul Graham is giving the introductory keynote; there's a strong slate of numeric talks (as usual); I'll have another chance to try to come to grips with Twisted; and there's a scattering of promising-looking unclassifiable talks such as Prevayler, Satine, and Smoke.

PyCon is also taking a radical approach that's new for the Python conferences. Instead of scheduling every single minute, leaving only small breaks for coffee and lunch, there's a lot of empty space on the schedule. The intention is to let attendees self-organize, finding empty rooms and time slots for impromptu discussions or presentations. Someone named Bob Payne has proposed using OpenSpace methods to plan the free space, resulting in a conference that's more closely adjusted to the needs of the participants. It's a provocative new approach that I suspect will be an immense success if it works at all.

Oh, and did I mention it'll be much cheaper? ($150 if you register before Feb. 28, instead of the $1100 for previous conferences.) And that it'll be in downtown DC and not off in the suburban wastelands?

Cardinal/daniels: I don't have a problem with swearing in a mostly-private forum such as a changelog or check-in message. Having spent days debugging a problem that turned out to be someone else's fault, I certainly know the urge to vent one's spleen. However, I do have a problem with the message "The fuck-it-all release": it contains no useful information explaining why it's the "fuck-it-all" release? Looking at the changelog, it's not clear which is the problematic set of changes: changing the init sequence, dropping the non-blocking I/O, or changing the package priority.

Diaristic stuff: I'm finally almost done with the seemingly endless remote microscope work, meaning I can move onto something else, such as trying to build Web services for the Matisse project. I'm wavering between just using XML-RPC and attempting a REST-based design; maybe Quixote can be made a convenient framework for implementing REST-based systems.

deekayen: a popular Web site carries the seeds of its own destruction. Everyone wants to be Yahoo or EBay, but the problem is that bandwidth and hardware make that really expensive. I don't think the solution is to find ways of sustaining a few high-traffic centralized sites, because it doesn't seem to be currently possible to make this work. Instead we should have very many low-traffic sites. I just wrote a conceit about this, arguing that personal weblogs are more sustainable and, as a bonus, more interesting.

I just posted my account of Necronomicon 2001 to my Web site; H.P. Lovecraft fans will be interested. If you're a Lovecraft fan, you so want to go to this convention. Trust me. Put it on your schedule for 2003. (Less interesting day-to-day things are in my regular diary, of course.)

I've written a guide to helping develop Python, which argues that Python is a good project to apprentice on and then discusses how its development is organized. Comments on the document would be greatly appreciated. Does it give you a reasonable picture of how the process works? Are there any other topics that need to be covered?

dmerrill, regarding your job hunt: would you be interested in working with us at the MEMS Exchange? Please get in touch with me, and we can discuss this further via e-mail or phone.

(I couldn't find an e-mail address for you on your Web page, so I'm resorting to this rather public method.)

4 Oct 2000 (updated 4 Oct 2000 at 14:44 UTC) »

At work, finished setting up the new Cornell computer (I think) and started figuring out how to set things up so everyone can run a ZEO server on their development machines, instead of directly using FileStorage. As an amusement, I reformatted my /data partition to use Reiserfs instead of ext2, since I'd like to get some experience with it. My /data partition holds various large source trees that are mostly external, and that I just CVS update, compile, and perhaps install: Mozilla, KDE, Linux, and the Python 2.0 CVS tree. As an experiment, I started up a KDE compilation with "make -j 2" and then turned the computer off, since Reiserfs is supposed to handle such crashes better than ext2 does. The results weren't encouraging; the machine rebooted OK, and the kernel logged a "Replaying 5 transactions" message when the partition was mounted, but then some files, such as the "configure" script, "config.cache", and "Makefile" were replaced with binary junk, perhaps from one of the object files being produced at the time of the crash. Maybe there's something I don't understand about setting up Reiserfs, perhaps some startup script or fsck invocation needed to reconcile matters.

splork: The un-SWIGged BerkeleyDB module is here. I'm still not very confident in it because I don't have a comprehensive test suite for it. Jim Fulton also pointed out a few missing API functions that need to be added, so I hope to hack on the module again before too long.

My previous Advogato diary entry was on August 31; it's much easier for me to maintain my personal diary pages, since I can let an entry slip for a few days and still get the date right, so readers interested in my diary should follow those pages.

Made a preliminary attempt at packaging up the Z Object Database, in order to make it possible for people to install it easily and then write software that requires the ZODB. Digital Creations seems to have no interest in making the software readily available outside of Zope, so I polished up my setup.py scripts, added MANIFEST files, and wrote some instructions.

Book review: Mike Nelson's Movie Megacheese, Michael J. Nelson
A collection of Mike's columns from Home Theater magazine. More entertaining than the Roger Ebert book I recently read, and a few absolutely side-splitting sentences are scattered through it: "Much of Twister was done in the digital domain, with 1s representing incompetence and 0s representing crap." (A sentence whose memory gave me serious giggles in a meeting today when someone mentioned a binary flag.)

Yesterday: Wrote a section on augmented assignment. Added a BerkeleyDB test suite from Stefane Fermigier; I really need to make a new release of the code. No actual discussion at book club, but luckily Martha brought a nifty game called Fluxx and we played two games of it; I'm going to have to get a copy.

Today: Released BerkeleyDB v2.9.1. Downloaded the latest version of LaTeX2HTML in order to make a new HTML version of "What's New"; the obnoxious fiddling required to produce good HTML from LaTeX never stops annoying me. The sooner TeX/LaTeX are dead and buried, the better; I think that will happen around 4 years from now.

Wrote most of the python-dev summary on the Metro in the morning (thanks to my new laptop), and posted it in the evening. Proposed writing a PEP for improving the CGI-related modules for Python. Got started on the job of hand-writing a natural Python interface for BerkeleyDB. (It's amazing how much cruft SWIG adds under the hood!)

15 older entries...

New Advogato Features

New HTML Parser: The long-awaited libxml2 based HTML parser code is live. It needs further work but already handles most markup better than the original parser.

Keep up with the latest Advogato features by reading the Advogato status blog.

If you're a C programmer with some spare time, take a look at the mod_virgule project page and help us with one of the tasks on the ToDo list!