[boost] Re: async socket i/o [long]

From: Boris (boris_at_[hidden])
Date: 2005-04-16 10:22:06

Michel André wrote:
> [...] Another slight mishap is that the things you call acceptor and
> connector arent acceptors and connectors in the sense of the original
> pattern (http://www.cs.wustl.edu/~schmidt/PDF/Acc-Con.pdf or POSA 2)
> as I interpret it.

I read the 17 pages for a better understanding. And I think you are right 
that we use different definitions. In the paper acceptor and connector are 
factories while in the class hierarchy at 
http://www.highscore.de/boost/net/basic.png they are an endpoint of a 
communication link. While I understand that the acceptor-connector pattern 
makes sense I think we should use it in a level > 0. If you look at the 
connector for example (in the acceptor-connector pattern) it is used to 
establish a connection and then passes over the connection to a service 
handler to do the processing. If level 0 should be similar to Berkeley 
sockets (and as far as I understand this is one goal of level 0) I don't 
know why the connector shouldn't be able to do the processing after the 
connection is established. In this level processing basically means calling 
read() and write(). A service handler in the acceptor-connector pattern is 
probably something bigger than a socket? Using the acceptor-connector 
pattern in level 0 would work but would also make the design more 
complicated as it should be to convince network programmers to switch over 
to the C++ network library.

Boris