- From: <noah_mendelsohn@us.ibm.com>
- Date: Tue, 3 Apr 2007 17:41:52 -0300
- To: "Dave Pawson" <dave.pawson@gmail.com>
- Cc: www-tag@w3.org, www-tag-request@w3.org
Dave Pawson writes: > [Mark Baker writes:] > > But that's > > not a particularly practical definition IME. A decent extensibility > > model can provide for *partial* understanding whereby the server can > > include extensions which it know may not be understood, and clients > > may or may not understand them. This provides for better decoupling > > in time - better independent evolution - than a model which relies on > > shared schemas. > > I read that as, it might work, if we're lucky. On this I agree with Mark. Of course, if you do this with no design or architecture, then Dave is right, it might or might not work. In practice, the way you make this work reliably is to have the architecture implement meta-rules that both sides know. In HTML, it's "ignore the tags (but not the content) when something is not specifically recognized". In SOAP, it's: "Never process something you don't completely understand; if you get a header you don't understand and it's marked mustUnderstand, you MUST fault, otherwise you have permission to ignore it." Crucially, both sender and receiver agree on these meta rules. They are examples of robust ways of implementing partial understanding. I agree with Mark: on the scale of the Web, and sometimes in smaller network, the ability to proceed in the face of partial understanding is often very valuable. Choosing a suitable meta-mechanism for the particular purpose can be crucial to the success of a technology such as HTML or SOAP. Noah -------------------------------------- Noah Mendelsohn IBM Corporation One Rogers Street Cambridge, MA 02142 1-617-693-4036 --------------------------------------
Received on Tuesday, 3 April 2007 20:42:02 UTC