Date: June 12th 2009

It's probably becoming pretty clear why I decided to stay with a simple
popup gui and wait for Blender 2.50 to do the real gui. The problem with
the current api's in Blender is that we have to compromise somewhere.

The options are:

1) pupmenu - limited input types, no radio buttons etc or advanced
layout, no events. Can be seen in RC scripts.

2) UIBlock - poor handling of screen refreshes and no use of popup menu
for the sculpt type. Can be seen in git scripts.

3) Display in current window - no control over size and position of
Window. This is how most Blender scripts do it. This is the easiest to
implement.

4) tkinter - Most flexible GUI but needs considerable programming to
implement. A tkinter application basically acts as a top level os
window, Blender won't process any events unless called directly until
the gui closes. As we are trailblazing here, there's always the
potential for hitting a show stopping problem.

So 2) has show stopper problems. That leaves a choice between two sorts
of user issues in 1) and 3) or programmer problems in 4)..

1) is a step back now, so our real choices are 3) and 4). Normally I'd
just go for 4) as any show stoppers could be fixed with a bit of c
coding. However with Blender 2.50 now the next version due out, any
patches to enable things we need in 2.49 wouldn't get to an official
release version of Blender.

Anyway, Gaia has the tough decisions to make now, I'm still optimistic
that we can capture all the necessary events to make tkinter seamless,
but I wouldn't complain about a switch to 3) either. Small user problem
< lots of programmer problems in my opinion ;)