Wow!! It has been a long time since my last diary entry.

I've left Linuxconf development team coordination in favor of the Conectiva Linux port to the S390 platform coordination (ok, I'm mostly coordinating myself now :-). Most of the work is done. I have developed an acceptable installer (in Python!) and most of the packages are ported. We had some problems with IBM OCO modules (ick!), but we are already workarounding them (we gave up on some of our kernel patches, and I patched insmod to recognize OCO modules). Anyway, more information about this later (if I don't disappear for another year.. ;-)...

In the process of porting Conectiva Linux to S390 and PPC (Harald Welve started the PPC port, and I'm keeping it up to date and working on missing stuff) we are learning some lessons. We are trying to use those lessons to build a defacto package building system using the Python language. Unfortunately, we don't have enough people here to develop it quickly, so we are trying a more realistic and evolutive approach this time. The first part of it is almost done. While devloping it, I've studied a little bit about process groups and extended python with a missing killpg() system call. I've also discovered that when python spawns a new thread, it blocks all signals. With this information in mind, I have also extended it with a new execv() syscall, that besides doing the usual work, unblocks every signal before the real call to execv(). I hope this project becomes real someday.

I've also been playing with Python optimization lately. There's a big opportunity for somebody wanting to study and implement some concepts there. I've read some documentation about Stack Machine Optimization and made some tries (basically, optimizations around the inner loop and the Big Switch, stack caching, and other flavors of this joy). Today I found a paper from Skip Montanaro documenting some of the tries I've made (reading it first would save me a lot of time, but this knowledge will be useful anyway). You should have a look at his paper if you have any interest in the topic. Oh, don't forget to get yourself a copy of Lemburg's pybench to have a general idea of what you're doing (don't trust too much on it, it's just a benchmark). I've written Skip a mail to discuss a little bit about what could be integrated into the interpreter. Let's see where we get.

Oh... I must not forget to update the people I've certified in the past to reflect what they've been doing.