[BXXPwg] Some recommendations regarding BXXP

Chris Hanson cmh@bDistributed.com
Sat, 12 Aug 2000 16:58:06 -0500

At 4:12 PM +0300 8/10/00, magdolna.gerendai@nokia.com wrote:
>2.) I don't understand, why the pipelining behaviour is, the requests and
>responses could come out of order ( the fragments as well). The serial
>number, (the channel number) and seqno unambiguously identifies  , which REQ
>or RSP the fragment belongs to. This gives the possibility of the real
>asynchronous transactions. In some use cases this may have importance.

I agree.  I think it should be up to a particular profile whether or 
not responses should be issued in order, rather than the higher-level 
application protocol.

Say you're designing an instant messaging protocol on top of BXXP.  A 
client could issue a "subscribe to group foo@bar.org" command, and 
then wait for responses to it.  One response might be "subscription 
request successful," and subsequent responses might be "message 
posted to foo@bar.org".  There's no reason why clients and servers 
shouldn't be able to handle these responses concurrently with other 
responses, regardless of the order in which they're received.

This type of functionality is similar to IMAP and ACAP's "untagged" 
responses, and could be very useful to higher-level protocol 
designers.  I know it's something people designing RPC-oriented 
protocols are very interested in.

   -- Chris

Christopher M. Hanson
bDistributed.com, Inc.