[BXXPwg] proposal for 1:N interactions

james woodyatt jhw@wetware.com
Tue, 29 Aug 2000 14:26:35 -0700


"Huston Franklin" <HUSTON@novell.com> wrote:
>[others wrote:]
>>>      - 1:0 interactions still require an acknowledgement from the
>>>  remote side:
>>>          otherwise, we could run out of msgno's
>>>          and it's probably needed for end-to-end correctness
>
>>Can the author expand on "probably needed for end-to-end correctness?"
>>Unless there is some reason to believe that acknowledgements are
>>always needed, this seems like an unfortunate consequence of the
>>design choices.  Especially since msgno is only present as a
>>consistency checking device when the underlying transport is
>>sequenced.
>
>This is nice if a session aborts so that it is know if the MSG was processed.

Okay... but if it is "nice to know" that the MSG was processed, is it 
still a 1:0 interaction at that point?

It seems to me that either one or the other of these two statements is true:

    + A peer sending a message may be uninterested in whether the 
message is ever received by the other peer or discarded by the 
transport or the network.

...XOR...

    + A peer sending a message is always interested in whether the 
message is received by the other peer, in which case at least one 
response is required.

I think the former statement is true.  I have seen many application 
protocols in which peers push messages without having any need to 
know whether they are received or discarded.


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