- From: <Noah_Mendelsohn@lotus.com>
- Date: Fri, 21 Sep 2001 10:57:44 -0400
- To: Marc Hadley <marc.hadley@sun.com>
- Cc: Jacek Kopecky <jacek@idoox.com>, xml-dist-app@w3.org, xml-dist-app-request@w3.org
I can't speak to the reasons why they were originally left out, but there are real performance costs in dealing with things like DTDs in a web services environment. We are in many cases competing with systems that are binary rather than text based. Whether the cost is significant depends on the sort of parsing you are doing, the message rates you are trying to sustain, etc., but simple is good. Even features like namespaces, which cause you not to know the name of an element until you have put together all of its attributes (namespace decls) have their costs. Namespaces are worth it, IMO; internal subset DTD's are a complication we don't need IMO. I think PI's are relatively cheap, but I think the complicate the model, and I'm not sure we need them. Is a PI part of the header entry in which it occurs. Must it be "understood" by the processing model? All this seems like unmotivated complexity. Just because it's available in XML doesn't mean we have to encourage its use, I think (I do think there should be ways to carry complete XML documents in a SOAP document, but PI's and internal subsets are only a small piece of that problem.) ------------------------------------------------------------------------ Noah Mendelsohn Voice: 1-617-693-4036 Lotus Development Corp. Fax: 1-617-693-8676 One Rogers Street Cambridge, MA 02142 ------------------------------------------------------------------------
Received on Friday, 21 September 2001 11:06:11 UTC