Re: Synthesis of the proposals for issue 64

Hi Jacek,

On Fri, Jul 18, 2003 at 01:46:11PM +0200, Jacek Kopecky wrote:
> Mark, 
> 
> what I meant (and probably should have said better) was that we don't
> need to go out of our way to support describing different web
> *resources* in a strongly-typed way. 

> On the other hand, specifying an interface with the four HTTP operations
> and naming it HTTP in some namespace of ours is a very simple thing to
> do and it does not need any changes to the language.

That's possible, but I'd prefer not to.  I think one should be able to
reference the semantics of any application protocol, not just HTTP.
For example, it would be good to able to describe an FTP based service
by reference RFC 959 and the "RETR" operation, for example.

Which brings up the issue of whether its possible to QName-ize an
operation name, ala "{http://www.ietf.org/rfc/rfc2616}GET".  The
published core draft says no, but this would what I suggest above
possible.  On the other hand, if that's too much of a pain(?), then we
should consider doing what you suggest.

> However, I don't think there is a lot of value in actually describing
> the HTTP interface as such because it would not really be used much -
> either a resource is a part of a greater thing in which case the
> interface is more complex and the operations are named differently than
> GET, POST etc.,

Granted, but that case wasn't a concern of mine when I raised the issue.

> or the HTTP resource wants to use WSDL to express some
> constraints on the operations, for example the schemas of accepted and
> emitted data, and then it will also have a different interface and it's
> not really necessary to say on the interface level that it's a
> restriction of the HTTP interface.

No?  Because I think that this is one of the few ways in which WSDL can
add value to Web-like & RESTful services.  I'd be happy to return WSDL
via the HTTP OPTIONS method that described the accepted XML schemata
of one of my POST processor resources.

> I'm presuming that we don't want to add some serious object-oriented
> modeling capabilities (operation overriding, for example) to our
> interfaces.

Right.

> Oh, and thanks for the praise there,

No problem.

Mark.
-- 
Mark Baker.   Ottawa, Ontario, CANADA.        http://www.markbaker.ca

Received on Monday, 21 July 2003 22:01:22 UTC