Martin Pool's blog

Mail me RSS Bazaar distcc Canonical is hiring! flickr

Visiting London

I'm in London for a couple of weeks.

It looks satisfyingly *like London*, as compared to some cities which are a bit bland and generic. The weather is appropriately grey and drab, with gilded spires somehow picking up and concentrating the light.

I went to the Tate Modern on the weekend. It was deeply moving, probably more so than any other art collection I have seen. Not only are the works without exception outstanding, but the organization and juxtaposition makes them even more meaningful and powerful. Bruce Naumann's installation in the Turbine Hall is a wonderful bit of sound-as-sculpture. (Doesn't the name Turbine Hall just beg for something like this?)

I'm going to eat wierd organs on Wednesday at St John.

There is also a purpose to this trip.

posted Tue 30 Nov 2004 in /travel | link

WotD: mythopœic

From Webster's Revised Unabridged Dictionary (1913) [web1913]:

Mythopoeic \Myth`o*p[oe]"ic\, a. [Gr. myqopoio`s making myths; my^qos myth + poiei^n to make.]

Making or producing myths; giving rise to mythical narratives.

The mythopœic fertility of the Greeks. --Grote.

posted Tue 30 Nov 2004 in /random/words | link

consuetudinary

n : a manual describing the customs of a particular group (especially the ceremonial practices of a monastic order) [syn: {consuetudinal}]

posted Tue 30 Nov 2004 in /random/words | link

Scoring LCA2005 conference papers

We're scoring and evaluating the LCA2005 abstract submissions now. This is a lot of work.

The proposals look generally very good.

It is very disappointing and frustrating when people put in inadequate abstracts, sometimes just a couple of sentences. Sometimes they sound interesting but without enough information they're always going to lose out to somebody who can explain what they're going to say. Doubly so because the whole point of a seminar is cogent presentation of interesting material, and if you can't do that in a web form why should I trust you with an hour of my time?

One common failing is to give a definition of the topic but not say what you're going to talk about. There should be some kind of indication that you'll do more than just tell me the definition.

It doesn't have to be cute. You can be entertaining in your presentation; you can even redraft the abstract later so that it's funny or intruiging in the program but judges looking at 140+ abstracts in a row would rather just have the facts.

Similarly for the presenter biography: I really don't care where you were born or went to school or what you do on the weekend. I might find your hobbies fascinating if we meet at LCA but it's totally irrelevant to choosing the best possible material for our conference. What I want to see is evidence that you know the material, and are experienced at public speaking in English.

The whole point is an assessment of the likely quality and appropriateness of the paper/seminar. Everything should be directed towards helping the judges decide that.

I'm tired and aware this is a bit ranty but it's also dead true.

posted Thu 18 Nov 2004 in /conf/lca2005 | link

An earnest plea to web spammers

Hi,

If you would like to run ads for asian midget facial porn sites on sourcefrog, please just write and ask me. Advertisements can be hosted for very reasonable rates. Sending so many web spam attempts, looks tacky and achieves nothing.

Yours, etc.

posted Thu 18 Nov 2004 in /issues/spam | link

LCA2005 progress

I've been busy getting the linux.conf.au 2005 papers database ready so that we can work out our program. Our target it to make a first release of the program this week.

We have about 146 submissions in total, and about 50 slots.

posted Mon 15 Nov 2004 in /conf/lca2005 | link

Mugga Lane

posted Sun 14 Nov 2004 in /photo | link

Treasure

posted Sat 13 Nov 2004 in /photo | link

Mural of girl's face, Canberra

Mural, Canberra, unknown artist

posted Sat 13 Nov 2004 in /photo | link

Valgrind wish

What's wrong with this code?

while ((connect(fd, sa, salen) == -1) && errno == EINTR) ;

if (errno && errno != EINPROGRESS) {

I wish Valgrind had an option to mark errno as undefined on entry to a system call, so that it's only valid to check it if the system call fails and sets errno. (I realize that strictly it ought to be carried over from the previous value if the syscall is correct, but code that relies on this may be broken.)

posted Fri 12 Nov 2004 in /software/valgrind | link

Simply yummy

yum made me very happy the other day at work.

Our software tree produces an rpm package when you say scons rpm, and the rules are set up so that for non-release builds the Subversion revision number gets into the RPM version number, like so: hp-foo-bar-3.0.trunk.r1376-1.ia64.rpm.

On an internal web server we have a directory for each suppported platform. The various built RPMs go into there, plus the libraries we depend upon that are not in the standard distribution. Then we simply run yum-arch /home/pub/yum/hp-foo/rhel3-ia64/latest/.

Various test and development machines reference this server in their yum.conf, so saying yum update on each of them will pull them up to the latest build. Other people who are testing our builds can also use yum, and we can have different directories for e.g. the tree head, alpha test, release candidates and so on. If we rev one of the underlying libraries it will get automatically upgraded as well.

Obviously it is not all that hard to install an rpm by hand. But we support several different configurations which need to be tested on different hardware, and we have internal beta users who will do more testing if we make it really easy for them (I hope). So I think the payoff is potentially quite nice. Being able to simply say yum update on a particular machine when you want to test it saves substantial time and avoids several possible errors compared to working out by hand which packages need to be updated and from where.

I haven't done it yet, but you can imagine a cron job on one machine that automatically builds an rpm and rebuilds the yum archive, and cron jobs on various test clients that automatically update and then test themselves.

There is an advantage to always installing from RPMS during test, which is that any packaging bugs are likely to be exercised often and caught early.

It's pretty simple but it works really well. Go yum.

posted Thu 11 Nov 2004 in /software/yum | link

Old school flava

% stty olcuc

(Try it! You'll like it.)

(The option exists to support ancient terminals (teletypes, etc) that do not support lowercase letters. It rocks that Linux still supports this.)

posted Thu 11 Nov 2004 in /software/humour | link

Antipattern

You can do this in a number of ways. IBM chose to do all of them. Why do you find that funny? — D. Taylor, Computer Science 350

posted Sun 7 Nov 2004 in /software/humour | link

Flame/smoke

[more]

posted Thu 4 Nov 2004 in /photo | link

The computer for the rest of us...

GNOME is getting to a really delightful level of elegance, of which Muine is just the latest installment.

Having used Stephane's mac a little bit, I wish Metacity had a "hide other windows" command, even if it was off by default. I know the whole point of it is not to add too many features, so maybe it'll never get in.

posted Mon 1 Nov 2004 in /software/gnome | link

Copyright (C) 1999-2007 Martin Pool.