The practical result (in terms of signal path) is exactly like you
describe (Csound calling out for composition data at k-rate). The only
problem with it is if getting (calculating) the composition data takes
more time than one k-period.
Actually, what I describe must not be considered a problem with the
csoundAPI, it is a general design issue for any software for
(sample/audio synchronized) realtime algorithmic composition.
Oeyvind

2008/4/26, Tim Mortimer <timmortimer@d2.net.au>:
> i've sent oeyvind some code - i don't think its as significant as he thinks
> it is. if there are problems with accurate scheduling using csound api in
> python then i thought the obvious alternate thing to try would be simply
> hosting a gui, & dictionary based data base for krate examination via
> pyops... & set all the output (& user added) "atomic elements" / list data
> values to ftables... i'm simply mocking up ideas in pure csound at this
> stage... "it aint pretty, but it works" - the csound motto perhaps?
>