OASIS Mailing List ArchivesView the OASIS mailing list archive below
or browse/search using MarkMail.

 


Help: OASIS Mailing Lists Help | MarkMail Help

sca-bindings message

[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]


Subject: My AI to raise an assembly issue regarding wsdl interface compatibility


Last week I took an AI to raise an assembly issue regarding interface 
compatibility between rpc/lit and doc/lit portTypes. I don't think there 
is a need to raise an issue (details below), but this was a lively 
discussion last week, so want to make sure that members agree and that I 
haven't missed something.

The argument for raising the issue lies in the fact that in WSDL 1.1, 
(with appropriate bindings) for every rpc/lit portType there exists a 
doc/lit portType that results in the same messages on the wire. Are such 
portType that generate the same messages on the wire, compatible?

Assembly currently says that compatibility is based on operation names, 
input/output types. The key here is what is meant by "type". WSDL 1.1. 
allows the "types" to be XML GED or XML type declarations. Further 
complicating this is that WSDL 1.1 (is mostly) about describing messages 
on the wire, grouping them etc. It does say much about semantics. WSDL 
1.1 doesn't separate the "abstract" and "concrete" parts cleanly. SCA 
has, for better or worse, decided to rely on WSDL 1.1 portType as the 
interface contract, which can be translated to other interface types 
such as Java and can be used with various bindings.

The definition of compatibility in assembly needs to be fixed for lot of 
other reasons (eg: make it apply not just to wires but also to 
promotions etc, define compatible superset and compatible subset). But 
if I were to interpret "type" as meant by the current definition to mean 
the same XML element or XML type then that essentially says that 
portTypes defined for rpc/lit and doc/lit are *not* compatible. I think 
that is a reasonable thing to say. Yes it restricts rpc/lit and doc/lit 
being mixed during promotion/wiring etc, but it is a reasonable 
restriction. Without that the rules will get complicated (for no 
apparent benefit) and I'm not sure how it will work across interface 
types and with non-web services bindings.

Comments?

-Anish
--


[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]