Updated my Mailman patch to use some of the newer features in SpamAssassin 2.20. This includes showing which rules got triggered for messages that get held (this is the feature that required the 2.20), and the ability to give messages from list members a bonus when calculating the message score (so that they are less likely to get held/discarded). The newer version is in Mailman's patch tracker:

http://sourceforge.net/tracker/?func=detail&aid=534577&group_id=103&atid=300103

With this patch and SA 2.20, I seem to be getting about 80% less messages to moderate, which is much more manageable. Adjusting the thresholds a little would probably improve matters further.

Looks like jdub is pushing to get my mailman patches put up on mail.gnome.org. It will be interesting to see how they hold up under heavy volume. It is coded fairly defensively, so it shouldn't cause mail lossage (unless you configure it to lose certain bits of mail), but I have no idea whether it has an acceptable CPU overhead.

Saw Rabit Proof Fence on Saturday. A very good movie and worth watching. It makes me sad to see how Aborigines have been treated.

Did a bit more mail system hacking, and wrote a filter to get mailman to talk to SpamAssassin directly:

http://sourceforge.net/tracker/?func=detail&aid=534577&group_id=103&atid=300103

This one asks spamd to score each message that is posted to a list. If the message goes above a particular score (configurable, default 10), it gets discarded. If it goes above another threshold (default 5), the message gets held for moderation. It probably needs a bit more tweaking to skip posts from listmembers (among other things). I mainly wrote this patch because the PyGTK mailing list currently gets more spam than real messages, which is a bit depressing, and a pain to moderate.

Made new development releases of pygtk and gnome-python recently. The previous ones didn't work with the gtk 2.0 release (trivial to fix up). Since the release, I have done some clean ups to its code generator. I have refactored it so that the 4 code paths used to generate wrappers for GObject, GInterface, GBoxed and GPointer type classes are now merged into one code path. This shortened the file by 500 lines, and should make it easier to add new features. The previous layout was getting pretty hard to manage.

linux.conf.au

Got asked to go on the paper review committee for next year's linux.conf.au. This should be interesting. For people living in Europe, it should cut off about 5 hours flying time compared to the eastern states, so hopefully we will get some cool European hackers submitting papers. Conversely, flights from the US will most likely be longer.

If you have never been to Perth, it is a great opportunity to come (it is a great place). If you live in Perth, it will be a great opportunity to meet many interesting people without > 4 hours flight :-).

linux.conf.au

Went to linux.conf.au in Brisbane last week, which was a lot of fun. I went a few days early to see what was going on at the debian mini conference, and got roped into giving a small demonstration of gnome 2.0 stuff at about 30 seconds notice. I am not very good at doing presentations with no preparation :(

Met up with gman, who was at the conference as part of his holiday over here. The talks were very interesting. Sounds like Samba 3.0 will rock (does win2k domain client RPCs), and it may not take long after that to become a win2k PDC.

Skyshow

Went to the Australia Day fireworks on Saturday night. I cycled into the city to see them, which was worth it (didn't get stuck in a traffic jam -- just had to dodge pedestrians on the way home). The fireworks looked so much better in the city (I went just a little past the narrows bridge). There were fireworks being let off the tops of the buildings in the city, off the narrows bridge, and off the barges in the middle of the river. At one stage, some small tough-ducks were driving past about 50 metres away and letting off some smaller fireworks. One of the embers from these smaller fireworks landed about 2 metres away from me! (on the esky belonging to someone sitting nearby).

More investigation of the slow processing speed for my document. It seems like the slowdown is somewhere in libxslt's chunking code.

With chunking turned on, xsltproc took 4 minutes to process the document, while without chunking (ie. producing one large file), it only took 1:30 minutes (less than half the processing time).

In comparison, using Jade to process the document took about 2 minutes with chunking turned on. With chunking turned off, it took 4 minutes.

I was updating my documentation generator for pygtk (the one that tries to make the C reference docs for GTK look like docs for Python). It was taking a while to process with db2html (which uses Jade to convert from SGML to HTML), so I thought I would look at using DocBook/XML and DV's xsltproc, which I had heard ran a lot quicker.

Unlike other people's experiences, the docs ended up taking over twice as long to process with xsltproc compared to jade! I suppose this was to do with the size (about 1.9MB of of XML source), and the number of cross references (the doc generation script added a lot of xrefs). On other docs I tried, xsltproc seemed noticably better.

Assuming that the initial count is correct, /me is on the GNOME Foundation Board of Directors! (as soon as the results are verified).

As board members are evenly spaced around the globe (I am in +800, Telsa is in +100, many are in -500 and George is in -800), board meetings are going to be at inconvenient times for some people. At the moment, it looks like it is George who will be getting up at 8am for meetings, and at midnight for me. As I said I would make meetings difficult to schedule in my candidate statement and George voted for me, I can't feel too sorry for him :-)

My arms are still a bit sore when I straighten them from the kneeboarding. I guess I need to get a bit more excercise.

I was setting up an Adaptec 2400A RAID card today. It came with a bootable configuration CD, which was pretty convenient. I put it in the drive, and see a LILO prompt, and it booted Linux, and started X.

There was a little program that gave access to the RAID config program, documentation, the ability to set up driver disks for various operating systems. The last option was to start an RXVT, so I chose that. Looking at the ps listing, I noticed that the GUI was actually a python program using PyGTK, which was a nice surprise. It is nice seeing my code being used in small applications like this.

Working at ActiveState was a lot of fun. It was interesting working on Komodo, and learning about the Mozilla code base.

I got to feed the useful mozilla fixes back upstream that was good. I got bug 107651 - Handle multiple file drops on mozilla (gtk build) using the text/uri-list target into the 0.9.6 release (which is not as useful as it could be, due to many mozilla dnd observers not being set up for multiple item drops).

In Melbourne for a month working for ActiveState on Komodo, which is pretty interesting. I am supposed to be fixing linux/gtk related bugs. It is quite difficult to work out where in the mozilla source code things are happening though :(.

Komodo is a very interesting project, and should mature into a good IDE (it is already usable under windows). Some great free software has come out of Komodo already (even if Komodo itself isn't), such as PyXPCOM.

From what has been happening, it sounds like Air New Zealand had been siphoning money out of Ansett by charging fuel and catering costs against the airline and possibly doing so after it knew Ansett was bankrupt. This has caused Ansett workers to call for a boycott of Air NZ, which the NZ PM doesn't like much.

To save money, the .au Government allowed the two domestic airlines (Qantas/Australian Airlines and Ansett) to build the air terminals themselves. At almost every domestic airport in australia, you will see a Qantas half and an Ansett half (usually with different architecture, etc). Separate checkins, separate bagage collection, separate arival/departure gates, etc. Now half of every domestic air terminal is left closed. Those people who happened to own shops inside an Ansett terminal can't open for business. I really hope some of the smaller airlines (such as Virgin) will finally be allowed to use Ansett's space in the domestic terminals, which will allow them to compete more evenly than before (before they often had to use other buildings round the airport or the international airport terminals).

It is really sad hearing so many people in the US out for blood (I have no way to tell how many people feel this way -- the internet+media can give a very skewed perspective on things). The terrorists killed many innocent people in the WTC, most likely because of issues they had with the US govenment and foreign policy. If the US turns around and kills innocent Palestinians or Afgahns (or where ever they happen to be based) in order to get the terrorists, that would be just as bad an act of terrorism.

I have been converting a lot of the boxed types in pygtk over to my new PyGBoxed code. So far, this has resulted in about 1000 less lines of non generated source code, which is helpful. It will also help wrap other addon widget libraries that have boxed types (provided they are registered with glib). I will have to submit a few patches for GTK to register the last few types that aren't already registered.

Recompiled devel gtk+ and its dependencies today from scratch, and gtk-demo still segfaults :-(. Tim committed my g_object_newv patch, so people should be able to build devel libglade. He also checked in the child properties stuff, which will allow me to handle that generically in libglade (once he adds a few missing APIs).

Once I sort out the gtk+ issues, I can get hacking on pygtk a bit more.

I am sure most people have heard about the flame war on the gnome-hackers list over the weekend. Things have settled down now, and there is talk of creating some procedures for introducing changes to the platform. Some people have argued that it is introducing too much bureaucracy, but I think it will work out quite well. Similar schemes have worked well for Python, TCL, Perl and even the internet. All have varying levels of formality, so we should be able to find a process that suits GNOME well.

Doing a bit more work on libglade2. It is still broken, but getting less broken as time goes on. Should get it so that the build completes to keep Sander happy :)

Since we are starting to get a number of functional free web browsers, I had the idea that it might be a good idea to create a Certificate Authority for free software projects and people and get its CA cert preloaded in browsers like Mozilla.

Put out another development pygtk snapshot. I actually released it yesterday, but my computer's clock was out by 12 hours when I made the release, but didn't notice it (something weird must have happened when bringing all the computers back up after the brownout on sunday), and the ntp server on the gateway didn't start up correctly so it didn't correct itself. I hate clock skew.

I have some ideas on how to decrease the amount of handcoded stuff in pygtk even further. The beginnings of this code is included in the latest snapshot (the GBoxed type). I haven't gotten round to converting any of the existing boxed types over to this new code or adjusted the code generator yet though.

First entry for 2001. A lot of things have happened.

I went on a holiday to Paris for a week and then Oxford for a week with the rest of my family. It was good, but a bit cold. I got to meet Mathieu while in Paris which was good.

A few days after I got back, I was back on the plane for Sydney (where I am writing this) for linux.conf.au. It starts on Wednesday, and should be a lot of fun.

Have been hacking on pygtk recently, and a small amount on glib HEAD and framebuffer gtk (which is looking really promising). I did up the first cut at allowing arbitrary GtkTreeModels to be defined in python code. It leaks badly, and it will probably be near impossible to fix correctly :(.

The glib patch was to add some convenience functions for the GSignal code, as it is so difficult to use the existing functions people are still creating GtkObjects because of the gtksignal compatibility wrappers. Still waiting on feedback from Tim about it

Haven't posted anything here in over a month. I just finished my last exam today, which was good.

It was interesting hearing about the formation of the KDE League, especially after Kurt Granroth's original comments after then announcement of the GNOME Foundation. I did an interview giving the perspective of a GNOME hacker on the League.

The Python bindings for GTK 2.0 are coming along nicely. I have wrappers for the new text and tree widgets mostly written, and they are very nice to use. Combined with python's unicode support, python makes a very nice environment for GTK programming.

Wrote some code to convert arbitrary elliptic arcs to bezier curves for the gnome-print driver in dia. This should stop people sending bug reports in about that problem. The output looks very nice, so no one should notice that it isn't real elliptic arcs. I was talking with Chema about adding support for arcs in gnome-print itself, possibly using my code as a fallback for drivers that don't support them.

I went to see The Dish at the cinema last night. It is a very good movie, and I recommend watching it. Apparently the release date is next week, so this was an advanced screening. They also had advanced screenings in Perth for Chopper back when it was released.

People have started putting in their nominations for the gnome-foundation board elections. We have enough candidates for an election to be necessary. None of the Sun guys have nominated yet. Maybe they will now that the OpenOffice source code has been released.

I handed in my dissertation today. The due date was Monday, but I got a one day extension, which was really helpful as it meant I could polish it a bit more. I feel much less stressed now that is out of the way. Maybe back to a bit more hacking now :-)