- From: <Noah_Mendelsohn@lotus.com>
- Date: Sun, 1 Jul 2001 21:37:21 -0400
- To: Rich Salz <rsalz@zolera.com>
- Cc: mnot@mnot.net, xml-dist-app@w3.org
Rich: you are right, to be effective a protocol specification must state precisely and concretely what is to go on the wire. I think the point some of us are making, and that you might be missing (or may disagree with), is that the infoset can be used to capture what is common across an extensible set of SOAP bindings, each of which must separately meet that goal of concreteness. So indeed, an HTTP binding must tell you exactly, "<" and all, what is on the wire. On the other hand, a different binding to SMTP might use a different set of rules for the bits on the wire. Now, consider a binding that uses compression to achieve speedup on some limited bandwidth transport. It probably won't have "<" visible on the wire at all. We could require that all such encodings were in fact compressions of the "<" representation of the XML, but why do that? Why shouldn't that compressing binding be free to use any technique it likes of reproducing at the far end the elements and attributes of the envelope? Maybe it creates a symbol table of element names, sends that first, and then uses code numbers. That's the reason that some of us want to use the infoset: it captures what is common across all bindings, without extraneous reference to well-formed syntax. As you make very clear, each binding must tell you exactly what will be found on the wire, and will definitely be at the "<" level if well-formed XML documents are chosen. ------------------------------------------------------------------------ Noah Mendelsohn Voice: 1-617-693-4036 Lotus Development Corp. Fax: 1-617-693-8676 One Rogers Street Cambridge, MA 02142 ------------------------------------------------------------------------
Received on Sunday, 1 July 2001 21:42:44 UTC