--- Steve Jones <[EMAIL PROTECTED]> wrote:
> > How is an HTML FORM not a formal, technical interface > > spec? It exposes a simple data structure (a > > multi-value string dictionary), but otherwise, I > > thought simplicity was a good thing when appropriate. > > Its as formal as a word document, its not something that can be > automatically consumed. HTML FORMs are parsable (for example, we just used http://jerichohtml.sourceforge.net on a major portal project to manipulate form actions & links) and have a formal submission protocol, defined by (for example) RFC 1867 (multipart/form-data). The logical data structure is a multi-valued dictionary, key being a string, value being any mime type. > > WSDL provides very little formality. WSDL just > > provides code generation tools guidance on what global > > elements to use for operations, bodies, and faults. > > In a loosely coupled system, this frankly should be > > runtime metadata. This whole edifce, to me, strikes > > me as ridiculously brittle. > > Its better than nothing (which is where REST is) it also can't be > runtime only as it has to be versioned and managed. As others have said, unless you're versioning and managing interfaces with paper documents, there's a piece of software somewhere that's tracking and managing them. Why shouldn't that be a hypermedia system? One can access control URIs to only "publish to production" after strict change control approval -- just like a web application. > Tell me about it, the lack of WS-Contract is a big irritation to me. We agree ;-) > > I think XForms likely is a direction REST will go to > > formally describe more complex data structures in > > hypermedia. Or the Atom publishing protocol for > > handling categorized, timeseries data. > > So right now I think everyone is agreed that this is a big gap in > REST. In practice, it hasn't been that much of a gap yet. XForms are useful in that they provide a hybrid human/automated rendering & submission model. But they're not absolutely necessary if you just want to exchange XML documents -- one only needs to know the URI of the POST or PUT and a URI or prior knowledge of the schema itself. This seems informal but frankly isn't materially different in process from how a developer looks up a WSDL, generates a proxy, and uses it. The major difference is that the REST model assumes a much higher degree of loose coupling (in terms of media type, and on insisting on placing all control information into hypermedia docs, whether for computer or human consumption). Cheers Stu ____________________________________________________________________________________ Need a quick answer? Get one in minutes from people who know. Ask your question on www.Answers.yahoo.com
