[BXXPwg] proposal for 1:N interactions

james woodyatt jhw@wetware.com
Wed, 30 Aug 2000 18:21:10 -0700


At 17:32 -0700 2000.08.30, Marshall Rose wrote:
>  > This conversation makes me wonder why we are proposing to change
>>  'REQ' and 'RSP' to 'MSG', 'RPY', 'ANS', 'NULL' and 'ERR'-- did I miss
>>  an important interchange at the IETF meeting?
>
>the reason why s/REQ/MSG/g and s/RSP/RPY/g is because a lot of folks were
>associating rpc-like semantics with beep. a casual reading of the framework
>makes clear that this isn't what the framework is about. nonetheless, i'd
>like to save several hours over the next year or so having to re-explain
>that. with MSG and RPY, it is unlikely that people will jump the gun to
>think rpc.

Your optimism is to be commended.

-----

Here's a question: is the following interaction intended not to be possible?

	C: MSG
	S: ANS
	S: ANS
	S: ERR
	...

In other words, once the first ANS message has been sent in reply to 
a MSG, is it not possible to report an error with an ERR message?

I note that RPY and ERR headers contain exactly the same information 
(except that they each have different header type codes).  I 
understand this is because it's 'nice' to have an indication in the 
header that a reply is an indication of an error as opposed to an 
indication of success.

It seems odd that there wouldn't be a matching mechanism for 
indicating error with the header type in the 1:N case where N>1.  Do 
errors only occur in the interval between when the MSG is sent and 
when the first reply is sent by the peer?

I could just be looking for symmetry where symmetry isn't 
appropriate.  What are  your thoughts?


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