- From: Francis Norton <francis@redrice.com>
- Date: Wed, 03 Oct 2001 10:46:19 +0100
- To: Noah_Mendelsohn@lotus.com
- CC: hutch@xampl.com, xml-dist-app@w3.org
Noah_Mendelsohn@lotus.com wrote: > > > Isn't that exactly the need that's met by the use of XML Infoset to model > messages at a local node in the latest specification? See [1]. I think > it's fair to say that W3C is moving toward infoset as opposed to the XPath > data model as the generic abstract model for XML...or more to the point, > they will come together as necessary. So, I think we've done what you're > suggesting. The Infoset does not currently provide a model for DTDs (or > schemas); though there is such an abstraction for schemas in the schemas > spec, SOAP does not use it. Yes, it's there in yesterday's spec. And infoset has to be the correct way of modelling it - the comparison to XPath is perhaps a marketing or persuasion point. > > Still, as has been pointed out, a given transport binding either will or > won't allow DTDs in the XML serialization on the wire. I can assure you, > having built implementations both ways, that requiring the receiver to > parse and apply entity definitions from a DTD will cost you something in a > high performance implementation. So, as I've said to often now, I think > we should not in general allow DTD's in abstract SOAP messages, and should > not use them in the wire formats for any transport bindings offered by the > protocol workgroup itself. > > This is an important topic, but I'm beginning to feel that we've gotten > the essential points on the table for the workgroup to consider. Should > we wind down the email thread for now? > Fine - I feel I've made my basic request (not to rule out SOAP implementations that cannot detect offending DTDs because the parser has already normalised the message) and that it has had a pretty good airing - thanks! Francis.
Received on Wednesday, 3 October 2001 05:47:30 UTC