Re: CR 18: Proposal

Another tweak to this would be to say that an anonymous [destination]
MAP means /if
the binding in use defines it for the message in question./  If the
binding doesn't define it, then the meaning may be determined some other
way, or may be undetermined, as usual.

This ties anonymous to a reasonable meaning in the case of
request-response.  It also sends a clear message to writers of bindings:
If you want to supply a default for [destination], call it
.../ImmediateDestination, otherwise call it something else.  Lest this
seem hypothetical, I'll note that two separate cuts at a one-way SOAP
MEP [1, 2] have given .../ImmediateDestination the obvious meaning (the
second was my attempt, not realizing that DaveO had already done better
months previously).

There is a slight glitch here, in that the request-response MEP defines
ImmediateDestination as a MEP-wide property.  However, the definition
given is "The identifier of the immediate destination /of an outbound
message/." (emphasis mine).  The wording above leaves open that a
request/response binding which (unlike the case of SOAP/HTTP) also
provides an explicit address for the response message can make that
address available as a property and explicitly state that it should be
used in the case of an anonymous [destination] in a response message.

If we're uncomfortable with bindings making explicit mention of WSA, we
could make such a statement about response addresses elsewhere
(somewhere) and tweak the wording above accordingly.

The main point is that leaving a specific "hook" for future binding
writers is possible and, IMHO, useful.  The secondary point is that
binding writers have already provided bindings compatible with this
hook, without even knowing it was there.


Received on Monday, 6 February 2006 17:46:59 UTC