LC review of WS-Addressing

Hi all,

I have reviewed WS-Addressing (you may see my comments at [1]) and I
have identified two issues that are somewhat related to WSDL.

First, discussed in [2], is about ReplyTo being required when a reply is
expected. Currently in WSDL 2 with its only available binding replies
come back to the requester using the same communication channel, i.e. an
HTTP request/response, which would make the ReplyTo address being the
".../anonymous" URI, which might be overkill. I suggest that ReplyTo is
made optional (with the default being ".../anonymous" which IMO doesn't
change the intent of the spec but makes many messages smaller. 8-)

I suggest that we (the WS-Desc WG) comment that ReplyTo should be
optional even when replies are expected.


Second, discussed in [3], is about action being required in all
WS-Addressing-compliant messages. I believe that the intent of the WSDL
Operation Name Mapping Requirement (ONMR) is that message bodies
identify the operation so we don't need action URI (unless that is the
extension that satisfies the ONMR) because WS-Addressing RECOMMENDS that
the action URI identify an operation input, output or fault.

So there are a few concrete points touching WSDL:

1) WS-Addressing seems to imply that interface operations (in
particular, their inputs, outputs and faults) have stable semantics
(because action identifies semantics through a WSDL operation) and I
believe the WS-Description group has said previously that operations
don't necessarily have semantics assignable to them.

2) WS-Addressing doesn't formulate the parts about action as an
extension that would satisfy ONMR, which it probably should to play
nicely with ONMR.

3) WS-Addressing action may be redundant in many scenarios due to other
means of satisfying ONMR, namely unique elements which seem prevalent.

I suggest that we decide whether operations have semantics (which would
be supported by the view that interface extension keeps compatibility,
which I heard from DaveO on the list) and if they don't, say that to the
WS-Addressing group, and further I suggest that we suggest that action
in ws-addressing be made optional and that it be formulated as an
extension satisfying the ONMR, when used (appropriately).

Hope it helps,

Jacek

[1] http://lists.w3.org/Archives/Public/public-ws-addressing-comments/2005May/author.html
[2] http://lists.w3.org/Archives/Public/public-ws-addressing-comments/2005May/0010.html
[3] http://lists.w3.org/Archives/Public/public-ws-addressing-comments/2005May/0011.html

Received on Tuesday, 3 May 2005 16:05:24 UTC