[BXXPwg] proposal for 1:N interactions

james woodyatt jhw@wetware.com
Wed, 30 Aug 2000 12:00:55 -0700


You write:

>i think that we're not going to solve all problems for all communication
>models using bxxp. if you're really using the former model, you should
>probably be using something that sits on top of udp or rtp or t/tcp or ...
>
>i think that the design points in the beep framework are biased toward
>reliability and that means that things get acknowledged. even in the 1:0
>case, a reliable transport is still going to have to send back acks. it's
>not much of a leap of faith to view the latter model as consistent with
>that...

Thanks for the clarification about the architecture.  This whole 1:N 
interactions thing may take a little work for me to see how it fits.

I had been focusing on the design points in the BXXP framework 
clearly aimed at establishing communications between two *authorized* 
peers, potentially with transport layer security.

I had forgotten (*sigh*) that BXXP guarantees reliable delivery of 
all its frames, and that this is quite a different thing from the 
guarantee that TCP offers in reliably delivering all its octets. 
This is why there is a need to send an acknowledgement in 1:0 
interactions "for end-to-end correctness"-- when the transport 
reports an error in reliably delivering octets, they're needed to 
sort out which MSG frames were and were not correctly delivered 
before the error occurred.

I'm awake again.


-- 
j h woodyatt <jhw@wetware.com>
http://www.wetware.com/jhw