Redoing Xcompmgr

[catach@tpg.com.au]

The project i'm working on is about to suffer from some big time scope creep,
and so am looking for a new area to concentrate on. I noticed this in the todo
list, and wonder what was in mind when '"real" application' was mentioned. I was
thinking either a revamped xcompmgr concentrating on performance, or what I
posted about earlier about trapping X events, modifying them and rethrowing them
(some kind of event manager). The area of concentration for the event manager
would be in performance as well as security. I'd like to hear everyone elses
thoughts and opinions :)

<Quote from To Do List>
Note that the current xcompmgr application is a giant hack; a good project would
be to write a "real" application that did the same job but was nicely split up
and cleanly coded. However, the current one still has "issues" that will
probably carry over into a rewrite.

    * Paint shadows better. Current code recomputes the gaussian curve and the
entire convolution with the sharp shadow each time a window is mapped or
resized. Splitting the shadow into pieces (four corners, four edges and the
middle) would help an awful lot in resize performance.

    * Check window stacking tracking. I've seen xcompmgr get confused about
window stacking; some validation code could be added and then suitable bug fixes
could be figured out.

    * Figure out some primitive hints so wms and apps can tell xcompmgr about
translucency values and suchlike.
</Quote>

******************
Luke Bigum
catach@tpg.com.au
lbigum@cs.rmit.edu.au
******************