Jun 232005
 

How’s nature forming interesting patterns such as the surface of sea shells, or simply telling apart the places to grow a head and a leg?
The mathematical explanation describes this by the interaction of an activator and an inhibitor which are produced in a concurrent, mutually influenced reaction with varying diffusion properties. Basically, this allowes local activation (activator) with lateral inhibition (inhibitor), i.e. an attribute emerges at a certain location and is suppressed in the near surroundings, but may be repeated in surrounding regions farther away.
The result is: structure! Structure, which emerges out of a homogeneous cell mass where all cells contain the same operational program (gene code). This page has a good explanation on the reaction details. By tweaking various properties of the fundamental two differential equations that describe this process, many aspects of pattern formation can be explained. Large gradients (inhibitor concentrations) can be used to form global coordinate systems which are suitable to control whole body formation, repeated small gradients (activator concentrations) allow for oscillation (pigment patterns of sea shells) or near optimal attribute coverage on surfaces (hairs).

Jun 202005
 

Have been hacking on the command line web blogger today and learned python on the way, so can you read this? Posting works already, q.e.d. However, NewsBruiser is giving me a hard time with not accepting editPost() requests. So in effect, i’ll not be able to update this entry later on which sucks.

Jun 202005
 

Finally reconnecting… since my server died earlier with a reproducible kernel BUG in the resierfs journaling code of a stock 2.6.10 (journal.c:2825). After poking around a bit, i decided to copy everything over (more than 100 Gigs) and move to ext3. Especially since i recently downgraded from 2.6.11.* because i ran into two other scenarios of filesystem corruptions under reiserfs (on different disks btw). So i simply don’t trust the reiserfs code anymore and am migrating all my machines to other file systems bit by bit. Sent off a bug report to the reiserfs people and ran into another bug in mc while backing up my files.

Jun 202005
 

Sat down with stefan today to evaluate GarageBand to get some ideas for improvements and neccessary additions in beast. The number and types of loops it offers out of the box (actually we also had Jam Pack 1) is simply thrilling. Cooking up a rich dance music background is a matter of minutes. We got lots of new GUI ideas out of it, most of which ended up in the beast TODO, and we got more dedication for extending the stock instrument set shipped with beast.

Jun 202005
 

Today we’ve been having a visit of Ileana Streinu at our university, where she gave a very interesting talk about intersection-free polygon unfolding in 2D space and related topics. One idea struck me as particularly interesting. A bit oversimplified, it is the reduction of available degrees of freedom in polygon expansion until only one degree of freedom is left for the expansion algorithm to operate on. An older but somewhat shorter explanation can be found in this paper.

Jun 202005
 

The intltool vs. scheme saga continues. Basically, Kevin Cozens (the gimp script-fu / tiny-fu maintainer) is willing to use proper markup for scheme strings via a (_ ) function, however applying such a change is post tiny-fu-1.0 stuff, so won’t happen too soon.

Jun 202005
 

Back from GUADEC, still a bit tired though. Thanks to all the helpers who made this happen! It was nice to see all those people again and talk to them. Sat down with wim to discuss atomic reference counting issues for GObject. We’ve figured most of the issues, but GClosure is still giving a bit of a headache, because its reference count is part of a bitfield.