- From: Andrew Layman <andrewl@microsoft.com>
- Date: Thu, 18 May 2000 14:26:24 -0700
- To: xml-dist-app@w3.org
- Message-ID: <7CD674FF54FBD21181D800805F57CD540D226D75@RED-MSG-44>
While I agree with many of the points made below in this thread, the fact that they are considered aspects of SOAP betrays confusion between what is mandated by the SOAP specification versus permitted by it. The SOAP specification is not intended to be a soup-to-nuts design of every feature that applications might find useful. Rather, it is predicated on the idea that specifications should be layered and composable, and so SOAP defines just the bare essentials of an envelope and a distributed extensibility mechanism. That extensibility mechanism is particularly important, since, unlike many more rigid protocol designs, it allows features to be added to a message in a reliable way. Instead of building all features into the base protocol, SOAP gives the sender of a message a way to mark elements to indicate that they are required extensions to the protocol (see the soap-env:mustUnderstand attribute). For example, the SOAP spec does not require or forbid object references. Object references, and everything attendant to them, are compatible with SOAP but properly should appear in separate specification. 'The discussion of whether object references are a good or bad idea is not something that the SOAP specification per se takes a stand on. If however, a particular message relies on object references per a certain specification, SOAP provides the soap-env:mustUnderstand attribute so that this fact can become part of that message's contract. Similarly, a security model is not built into SOAP. But, whatever security model is chosen as valuable, it can be added to a SOAP message in a reliable way.
Received on Thursday, 18 May 2000 17:27:03 UTC