On Mar 25, 2004, at 1:09 PM, Dennis Sosnoski wrote:
> While my understanding of rpc/lit matches Marc's, my earlier comments
> on doc/lit vs rpc/lit still hold. I can see very little in the way of
> advantages for rpc/lit and many disadvantages (including
> interoperability, as Anne pointed out). I'm also not sure about Marc's
> claim that a schema definition can be synthesized for rpc/lit based on
> the WSDL. This goes back to my earlier question about the message
> components representing parameters - are the names of these components
> actually fixed by the WSDL, or only the types? My understanding is
> that only the types are fixed, which would mean that a schema could
> not be constructed from the WSDL.
>
The WS-I BP clarifies the local and namespace names of the operation
wrapper element and the elements for each parameter - see last
paragraph of BP 1.0a section 5.6.7 and requirements R2717, R2729 and
R2735.
> As I see it, rpc/lit is a way of holding on to the original method
> call model of SOAP. I think this model is too limited to be usable for
> general processing. Why support something that's admittedly a subset
> of the other approach, also supported? I know a number of people who
> train developers on web services (including myself) and have yet to
> meet one who advocates (or even discusses in any depth) the use of
> rpc/lit, instead moving students directly from rpc/enc to doc/lit.
>
> As for WS-I BP 1.0a, I do not see any requirement in the document that
> implementations support both doc/lit and rpc/lit. Instead, it defines
> the restrictions on how each of these alternatives is to operate.
> There are other places in the document where support for a particular
> feature is required, and that is clearly stated (for instance: "R4001
> A RECEIVER MUST accept messages that include the Unicode Byte Order
> Mark (BOM)."). If the BP 1.0a writers did indeed intent to require
> support for both doc/lit and rpc/lit they did an amazingly poor job of
> stating that in the document, which seems very inconsistent with the
> rest of their careful work.
>
The Basic Profile only talks of conformance in terms of services and
the artifacts they produce. It doesn't set a bar for tools or
platforms; provided a tool can be used to create a WS-I compliant
service it can have the WS-I stamp of conformance. One could argue that
a C compiler is WS-I compliant because it can be used to create WS-I
conformant services but most reasonable folks would probably see that
as stretching the point somewhat.
A WSDL that uses rpc/literal to describe a service is WS-I conformant.
In my view a tool that claims WS-I conformance ought to be able to
consume it.
Marc.
---
Marc Hadley <marc.hadley at sun.com>
Web Products, Technologies and Standards, Sun Microsystems.
- application/pkcs7-signature attachment: smime.p7s