On Saturday I attended the Gentoo UK 2007 conference held at UCL in London. I would like to thank James and Peter for the effort they put into organising it. I gave a short talk on my work on KDE and some of the other work I do on scientific applications and libraries as part of the scientific herd. You can grab a copy of my slides here if you would like to take a read. I did just use them as a framework to talk around though...
I enjoyed the Gentoo UK conference. It had quite an informal tone and was a great chance to meet a few other Gentoo developers along with budding developers, users and other interested parties. I wasn't able to make the last couple of meetings down in London so it was nice to make this one, although I could only make it down for the day. I took quite a few photos that I have put up here. Both sides of my talk received interest from different parties. I am most excited about my current work, Avogadro/Kalzium 3D molecular editor, which combines my two passions.
Just wanted to say thanks for the talks you gave - really changed my perceptions of KDE, and I now have KDE 3.5 as my primary window manager on my Gentoo laptop and loving it!
I have always liked the way Gentoo handles KDE. As it has always been such a big beast, the split ebuilds were a brilliant idea. I hope to see these still maintained in the future.
I really liked also your talk at the aKademy 2007, especially how you managed to balance the discussion around KDE and Gentoo, giving at the same time a nice introduction to the principles of the latter.
The other area, scientific packages, has generally always been a real problem child for any distribution. Just look at the lapack/atlas/blas -scheme! The general picture I have is good: scientific packages in the official tree are pretty stable and the scientific overlay works. There is no use to be bleeding edge at this front.
At the same time, as for many of us scientific packages are the most critical ones, I have always been under the impression that these packages have had just a minor role in the distribution, perhaps due to the staffing needs of the herd or perhaps due to the very nature of these packages. It has sometimes been frustrating to end users that big decisions regarding for instance the toolchain prevail the things you need in your day-to-day work.
As you mention, many of the programs require sophisticated knowledge to test even the very basic functionalities. The overlay gives a nice playground for the Herd-specific testers, but could this idea be extended? We have official AT-testers, so why not strictly Herd-dependent testers with similar guidelines (and mild requirements)?
At least I trust more someone who is writing his/her Ph.D. thesis using a given package than someone who owns a x86/amd64/whatnot computer...
We do have herd testers for the scientific herd now (we may have had a few when I gave the talk - can't remember the timeline now). Hopefully I will get the chance to attend future conferences but I am based in the US for the next year or two.
Great to hear from you.