[boost] Re: Aaron's Informal Proposal for a Universal Demultiplexor for Boost

From: Tommy (tommyboost_at_[hidden])
Date: 2004-09-14 04:12:57

Hi Aaron

I've been following the threads spawned by you and Carlos with great interest!
And this proposal is interesting indeed! As far as I can tell, I like your 
approach. 

I'm not sure how, but I think I have a hunch how the demultiplexor can be used 
in conjunction with other components, such as threads, to realize 
functionalities on a higher-level such as a socket library but I relly would 
like to see some diagrams and more code and examples - please! 

You asked us to forget all about event handling, but... I'm not sure whether 
the event demultiplexing model is synchronous or asynchronous? (Compare the 
patterns Reactor (sync) and Proactor (async) as described in POSA2. Btw do you 
consider ACE's implementation of these patterns "big, clunky, monolithic 
demultiplexors"?)

I strongly agree with you that the demultiplexor library should not use threads 
as those are part of the concurrency model - separation of concerns. However, 
if needed, it should of possible to combine a synchronous event demultiplexer 
together with a thread-pool to implement the Leader-Followers pattern via a 
thread-pool reactor.

/Tommy

PS.

I'm confused regarding the use of the terms multiplexor and demultiplexor. I 
hope you mean demultiplexor when you say multiplexor - right?