--On Thursday, November 22, 2001 14:49 -0500 Mark Baker <distobj(_at_)acm(_dot_)org>
wrote:
I have a question about callout, though I suppose it's really a question
about the OPES model; why is it necessary to have a standardized callout
protocol?
From what I can tell, the requirement for it seemed to pop out of figure
4 in opes-model-00. I believe that if that diagram were drawn with a
box just below the callout server, called a "OPES gateway intermediary"
(or surrogate), then that would prevent the need for a standardized
protocol. The protocol between "OPES intermediary" and "OPES gateway
intermediary" would be the primary protocol of the chain of
intermediaries, and the protocol between "OPES gateway intermediary" and
the callout server wouldn't matter; whatever it happened to be, a
gateway could be developed to bridge the two protocols.
In the above, is the intent to have the OPES gateway intermediary
communicate with the OPES intermediary via software API or on-the-wire
protocol?
If the latter then surely the standardized callout protocol becomes that
channel of communication (which means the protocol needs to consider
potential transcoding issues that may occur, er, yuk ;-) ).
If software API then I think that limits the usefulness of what this
proposed working group is trying to do. Also, if you wanted to talk a
proprietary callout protocol then might it not be possible to have an
OPES-compliant proxylet just do that? (In which case the gateway
intermediary interface is the proxylet itself, speaking OPES proxylet API
on one side and your own custom callout protocol on the other.)