As I have been pretty much house bound for the past two weeks, working on my dissertation, I ordered a few cd's from 78s (one of the best music stores in Perth) over the internet. Very good service, and the much faster than other places I tried.

Got a copy of the new Turnstyle record Geek Party. It is a good compilation.

Looks like they sold Fatso for AU$80,450. Eric Moussambani's goggles sold for a fair amount as well. There is a nice quote from him on the auction page at ebay.

working on honours dissertation. due in on monday. roy and hg's coverage of the olympics was good. fatso was a much better mascot.

jamesh certified aaronl as Dimwit

GNOME: prostituting ourselves to big-money for the chance of being a media-recognized standard since 15th August, 2000

:-)

I was looking at the coins problem. I can see a solution if you know the bad coin is heavier (or lighter -- just s/heavier/lighter). Here is a solution for that:

1. Split the coins into 3 groups of 4 coins.
2. Put two of the groups on the balance.
3. If they weigh the same, the third group contains the bad coin. Otherwise, the heavier group contains the bad coin.
4. Add two coins from one of the good groups to the bad group, and split those coins into 3 groups of 2.
5. perform the same weighing operation to find the group with the bad coin. This leaves 2 coins.
6. Weigh the last two coins. The heavier of the 2 is the bad coin.

This doesn't answer the original problem, but may give some idea of what it would look like. If you know the bad coin is heavier and can do 3 weighs, you should be able to pick the bad coin out of a group of 27 coins.

On IRC:

<raph> UTF8 really is simple enough to do in your head <raph> assuming your brain has SHL and SHR instructions <jwz> you need help, raph

The book I am tech reviewing must be almost finished.

I released a new version of dia on sunday. Didn't get everyone's patches in, but I thought it was important to get another release out because of how long ago the previous release was. Of course, after doing the release someone finds an XIM bug, I notice that the SVG CR has been put out (so the recommended namespace URI's have changed) and I didn't tell the translators about the release. So I will probably put out another release of dia soon with these fixed.

I should put out a new release of Dia soon. The next release will be the first release as an official part of the GNU project, which is pretty good. Lots of nice user interface improvements. Just one more feature before the release :)

Over the weekend, we pulled through a few CAT5 cables in the house. By putting the cable through the wall cavity there is no coax running through the doorways anymore. This should also reduce the problems of trying to find out which computer's network card has gone crazy when the network stops working.

I got my new notebook, and installed Linux on it. It is mostly working. The only real problem I have had is with resuming from a hibernate when I was in X. With X in 32-bit mode, the display gets garbled. With 16-bit mode, the whole machine locks up. Changing to another VT before hibernation avoids the problem. I wonder if XFree86 4.0 would fix the problem? It is probably caused by the Compaq bios not saving enough of the state on the graphics card (enough for the windows drivers but not enough for XF86_Mach64)

Heard from another guy who has a Sony NEWS workstation who gave some interesting info about the problem we were having with our machine (it doesn't boot). We had left the machine off for a while, and then it stopped booting. I had thought that the BIOS was dead or something, but it sounds like a NEWS box won't boot if the battery is dead also. Now I just have to find where the battery is in the system. There is no information about these machines on any of the sony web sites (they seem to be denying all knowledge of the machines). Now all I need to do is find where the battery is in the system -- they are built very solidly.

I saw the xmen movie on thursday (it was released a day earlier here, compared to the us). It was pretty good, and the ending looked like they are expecting a sequel or something.

I finally got round to ordering the my laptop, so it should be in on monday. It is a Compaq Armarda M300 (with the 1024x768 screen). With the introduction of the GST, computer hardware suddenly dropped by about 10% (and software went up 10%). It is annoying that the wholesale prices on most hardware are about the same as US street prices though.

I haven't written anything here for a while. I went to a friend's 21st birthday on friday, which was pretty good. The day after we went to a climbing gym. It was the first time I had gone this year, so I was a bit out of shape.

I started working on the gtk+-2.0 based pygtk. I wrote a simple module gobject that encapsulates the base GObject stuff. This also encapsulates the evil reference counting stuff away from the rest of the pygtk stuff. I have done wrappers for most of the Gdk types that have been converted to GObjects, and generated the wrappers for GtkWindow and GtkLabel to test out a few features. Interfacing with the python-2.0 unicode strings didn't require any extra code, as using the "s" format character for PyArg_ParseTuple causes the unicode string to be encoded in UTF8, which is what GTK wants. So you can do GtkLabel(u'\u03A0\u03B1\u03BD\u8A9E') and have it do exactly what you would expect. Lets see you do that in Tkinter :)

Did a lot of work on the extension class based pygtk. It now does most of what the non extension class based one did. With a simple skeleton generator, adding support for new libraries is very easy. I added libglade and gtkglarea support in just a few hours (gtkglarea was easier, as it required less override code). I also tightened up the type checking a bit, so you get a python exception if you try to pass a non widget to gtk_container_add for instance (rather than a Gtk-WARNING and the function failing without telling the python program). For those interested, you can get it here.

One more exam tomorrow (computer vision), then mid year break.

I think someone has been distributing binaries of dia with gnome-print support (which is broken). This has resulted in a lot of complaints and bug reports about the print support not working correctly. I hope it isn't the Helix packages with gnome-print support enabled.

It seems like no one who submit bug reports bother to take the time to check if the bug has already been reported. It is very frustrating to get lots of reports about the same bug. It is not as bad as reports about bugs in old versions of a package that have been fixed in the latest version or "it crashed" reports with no info on what the problem was. A little extra effort on the part of the person submitting the report can be very helpful to the developers.

Did my first exam today. It was differential topology. As there are only two people doing the course, I either got top or bottom mark :) Two more exams to go and then the break.

We switched ISPs on friday. I am sure that our previous ISP was saturating their network with the internet phone call service they started up a while back. At times, performance was really bad. Things are half working with the new ISP. We can reach just about everywhere except most perth ISPs. I will have to look into this.

Finished upgrading quoll today. It is now a Pentium 200MMX, which means it is now not the slowest computer in the office (not that you would notice -- it can do a lot more than the NT boxes). I switched over the mailing lists to mailman, which looks very nice. It feels nice having newer software on that machine.

I haven't posted anything here for a while. I put out a new version of dia last week. It went pretty well, but there are a few major bugs in the release (missing header include, bug in ellipses in custom shapes). I may make a new release soon when I stabilise the new code I added. With the current CVS code, we have XIM support, tearoff menus that actually work, a less buggy polygon shape, and the start of a beziergon shape. Once the beziergon is in better shape, I will probably release it as dia-0.86, and hope I don't have any more showstopper bugs. Steffen has been working on a DXF import filter, so I will be adding more properties to the standard objects to make it easier to create them through the properties interface.

Went to see the movie Looking for Alibrandi last night with some friends. It was a pretty good movie, and I recommend seeing it.

Put out a new release of dia, and a few problems were found fairly quickly. One weird one was finding that the libart in released gnome-libs was different from the one in CVS. It was different enough so that the AM_PATH_LIBART macro in CVS failed on released gnome-libs. There were also a few other minor problems I will fix soon.

Been talking with Andrei about getting gtk bindings for php working. He will probably be using some of the ideas from my python bindings.

Got the first few chapters for a book I will be tech reviewing. They are doing everything with MS office tools, which is a bit of a pain. Abiword is nowhere complete enough to do the job, as it doesn't do the revision control stuff needed. Gnumeric is probably good enough for the spread sheet work. This is the part I hate most about the MS monopoly -- they have a monopoly on how you can access the information. Dealing with Havoc's book was a lot easier, as he convinced them to let him use a text format (XML). I guess I better put a copy of msword on the old 486 then.

Haven't written anything here for a while. I put out another snapshot of the ExtensionClass based pygtk. It is available at ftp://ftp.daa.com.au/pub/james/python/pygtk-0.7.0-unstable-dont-use.tar.gz. Features this release has over the previous one are keyword argument support, better GtkCTree support and some other changes so that the examples now work with minimal modification. I am getting close to the main gtk module being feature complete, and since most code is autogenerated, the switch to gtk+-1.4 should be fairly painless and hopefully quick.

More work on new canvas. Started work on the update and render code for DiaCanvasItem and DiaCanvasGroup classes. Some of these parts of the current GnomeCanvas code feel like spagetti code, so hopefully this new canvas should be more maintainable. The event code still needs to be hooked up. Soon I may actually be able to start testing things a bit.

I went to the show put on for Science Week by Dr Karl Kruszelnicki and Adam Spencer last night. It was very interesting to hear about the weird things some people research. If you have a chance to see them in another state, go.

One interesting thing in the show was when they were pointing out one of the sections in the GST^H^H^HNew Tax System act:

For the purposes of making a declaration under this Subdivision, the Commissioner may:

In response to sad's remarks about the trust metric, yes it is really a respect metric. It only has the name trust metric because the original use for the algorithm was encryption key webs of trust. Advogato was an experiment using the same algorithms for a respect metric. I think raph mentioned this in some of the early advogato articles.

Looking at doing the expose/redraw code for DiaCanvasView's. I am implementing this with Federico's nice UTA manipulation routines from eog. The routines are GPL, but as Federico was talking about using them in GnomeCanvas, he probably won't mind relicencing to LGPL in the future.

Been hacking on the new dia canvas. The stuff is in the dia-newcanvas module. I am currently working on some of the DiaCanvas / DiaCanvasView interactions, which is interesting. Hopefully the canvas will be useful to other people doing multi view canvas stuff. It should also add minimal overhead to people doing single view stuff (do same sort of initialisation as for GnomeCanvas, but also create a view), and you get transparent print support.

Been creating some RPMs for the stuff needed to compile programs for the Mindstorms RCX (that is, binutils and egcs compiled with h8300-hitachi-hms as a target and legOS). I just realised I don't have enough batteries to test things out, and it is a public holiday :(