[standards-jig] IBB Again
justin-jdev at affinix.com
Fri May 2 22:20:30 UTC 2003
We never finished the IBB discussion, and I'd like to get this thing resolved
already. The sooner we get this darn thing out of the way, the sooner we can
finish JEP-0052. The main issue is iq vs message.
To use message properly, we need JEP-0079 (Message Delivery Semantics) to be
approved and for server implementations to use it. Thus, to use message
means to lose compatibility with existing servers. Reliability is also a
requirement of IBB, and this can be achieved using delivery notifications
from the approved JEP-0022 (Message Events). This has the nice advantage
that 'ack'ing can be done selectively.
Now for my opinion:
IBB is supposed to be as reliable as S5B, which is based on TCP. IBB is not
intended to behave like RTP, so I don't consider the availability of
selective 'ack'ing to mean anything (unless we want to change the IBB
requirements). That considered, I think since using -79 and -22 together
effectively gives you iq, both stanza solutions are identical. If both
stanzas can do the same job, then I choose the one that will work over
existing servers: iq.
I guess these are the questions:
1) Do we want IBB to have the ability to be unreliable? If yes, then we have
no choice but to use message and -79. If no, then:
2) Do we want to IBB to operate over existing servers? If yes, then we use
iq. If no, then we use message and -79.
Let's answer these questions so we can get a resolution already!
More information about the Standards