[Tdg] after the meeting

[Stefan Muller Arisona]

hi,
as promised i wanted to add a few thoughts after yesterday's meeting:

- to me it seems very important that you can present a few very concrete
examples where you employ mint. as stephen pointed out, there must be a
clear position, how mint relates to other / previous stuff, and i think
a good examples can help.

- i would like to know a little more about the messaging system itself
(including code), however i dont think its necessary to do this in terms
of a general meeting. who would be available? wes? also we'll have to
think of how to start bringing soundium into mint (as a test case)

finally, a few notes about or approach to decklights processing (as for
the jargon, soundium is the system as a whole, or just the client
application [java], depending on context, decklight is the underlying
processing engine [c++]...):

when design the graph thing, we followed somewhat the opposite approach
of mint, as we decided not to use messaging wherever possible, but to
view the graph as a large "data bus". as individual parts of the graph
are running asynchronously (i.e., in different threads), there would be
some special handling at asynchronous interconnections, where we allow
dedicated operators, such as maximising/minimising/averaging/queuing,
etc.  the design was based on the viewpoint, that you're often not
really interested in the queue contents, just in the front or the tail,
and thus a simple bus scheme would do. this allows for a lot of
optimisations, especially for connections which are inside a thread,
where to compiler basically can inline everything, and all boils down to
a simple assignment / read.

for those interested in more details:

- i'm at CNSI 2213, just stop by. i can also show a few demos
- further reading: http://www.corebounce.org/wiki/Main/Publications
-- the three pds (ICMC03, Grapp06, and AMCMM06 give a good overview, the
last one in particular to the things mentioned above).

best,
/stefan