Just a short status on CVS

Hi!

Since I now use the CVS version for some time, I can say one thing: it's
heavily broken. So if you value your data, you better stay with 0.7.0
;-)

Looks like the last changes I made are not really helpfull for
stability, maybe I am throwing them out again. The main problem seems to
be in the dynamic loading/unloading of tools into the user context. This
helps to keep thread numbers and resources low, but looks like it
triggers some very nasty bugs. From the symptoms it looks like it's
mostly connected to metakit - if you open and close database files very
often, your systems sooner or later breaks. This in combination with the
dynamic thread starting and stopping is an explosive combination.

I am not a happy programmer ;-)

I think the future road will be to throw out the dynamic
loading/unloading again and keep database files open all the time. This
is bad for plans with multi-user-pyds, as that would require far too
much open filehandles. But on the other hand it might be a good idea to
not push the metakit stuff too far and switch to some serious database
for multi-user-pyds.

On a side note: I started a second project (mostly due to requirements
at work for a small and simple webservice infrastructure) that
reimplements most of PyDS infrastructure from scratch - it's allways a
good idea to do things again after some time of hacking. That project
(Toolserver Framework for Python) might give a better base system for
PyDS in the long run (especially in the context of a multi-user-pyds),
as it is targeted much more at a standard webserver environment than
PyDS ever was.

Anyway, it might take some days until CVS is useable again, so you have
been warned! 

bye, Georg