On CR016: wrpc:signature for RPC style

I got an unrecorded action to investigate CR016 [1], which I'm
hereby fulfilling.

The example provided by the submitter of CR016 failed to convince
me of the need for supporting substitution groups in the definition
of wrpc:signature.

The rational for the wrpc:signature extension [2], and more generally
for the RPC style [3], is to capture in a WSDL description the RPC
signature of the procedure/function/method from which a given WSDL
operation was generated. There is an assumption then that the signature
has independent interest and was likely defined first, and that the
intent of the WSDL author, be it a human or a tool, is to faithfully
convey that signature to the eventual consumers of the WSDL document.
Since the definition of what is allowed to appear in the signature of
a function is highly dependent on the specific programming language
being used, the working group tried to capture a common, useful notion
of signature and base the definition of wrpc:signature on it.

Now, in the case described in CR016, I don't know of any mapping for
any language that would use substitution groups that way. The mechanism
commonly used to map inheritance to XML Schema involves abstract types
and is already supported by the current definition of wrpc:signature.

It's possible that someone doing message- and schema-centric development
would run into this case, but it seems extremely unlikely that they
would see any value in specifying a wrpc:signature. Moreover, since
in this development model the signature is secondary to the schema
definition, such a use case goes beyond what wrpc:signature was
designed to cover.

A stronger case could be made for CR016 by exhibiting programming
language bindings that would result in substitution groups for
what amounts to function/method arguments. A good example of this
was the way LC75g [4] was addressed [5] by mapping the widespread
rest/vararg programming language feature to it.

I'd like then propose to close CR016 with no action.

I should also point out that, should the need arise, it is possible
via extensibility to define an extension to represent richer signatures
than those supported by wrpc:signature.

[1] http://www.w3.org/2002/ws/desc/5/cr-issues/issues.html#CR016
[2] 
http://www.w3.org/TR/2006/CR-wsdl20-adjuncts-20060106/#InterfaceOperation_RPC_Signature_Definition
[3] http://www.w3.org/TR/2006/CR-wsdl20-adjuncts-20060106/#RPCStyle
[4] http://www.w3.org/2002/ws/desc/4/lc-issues/#LC75g
[5] http://lists.w3.org/Archives/Public/www-ws-desc/2005Mar/0038.html

Thanks,
Roberto

Received on Wednesday, 15 March 2006 18:33:22 UTC