Value of distinguishing MEPs

Someone (Dave O?) wondered what value there might be in explicitly 
flagging what SOAP MEP is in effect.  I think it's a very pertinent 
question, and I was getting ready to argue that we don't need it, but 
then something occurred to me.

It would be useful for an intermediary (whether in the SOAP sense or 
just a proxy) to know whether it can close the back-channel or not.  In 
the extant case of SOAP/HTTP req/rep, it knows that it can't.  In the 
async case, however, it's not easy to tell.  At the very least, the 
proxy would need to look at the MAP headers and know about anonymous 
endpoints and/or the defaulting rules for missing MAP endpoints.  This 
is leaving aside i054.

On the other hand, if the sender knows that it doesn't expect a reply 
(i.e., no endpoints are directed to the back-channel), it could express 
this by using "fire and forget" one-way.  If a reply may or may not come 
back (i.e., the various async cases we've been discussing), it uses 
in-[out].  If a reply is definitely expected, it uses in-out.

It seems clear that these three cases could all be treated under one 
umbrella, with a three-way "reply expected" flag, or perhaps a binary 
"send back at least an ACK" flag.  It's less clear to me whether such an 
umbrella should or need be considered a "SOAP MEP".  Maybe so and maybe not.

Received on Tuesday, 29 March 2005 17:58:44 UTC