Wednesday, April 29, 1998 at 11:58:02 AM Pacific

Paul Snively on DTDs

From Paul F. Snively, chewy@mcione.com, in response to Calling All DTDs!:

Lisa Rein is right on

DTD's may seem "limiting" when you've got a highly dynamic ODB, you can store a string at an address one moment and an integer the next, and there's no denying the power of that. But even within that context, you know that there are moments in time when the content, *at some arbitrary level of granularity*, has to have a particular structure. The most obvious example is that an address within root.websites contains the correct syntax to run through the rendering pipeline to produce a website that looks the way you want. There's no a prior reason why you couldn't describe the entire site in XML. In fact, the big question would be how much metadata about the site you'd want to include in the XML ("This is Paul's site about the aesthetics of Umberto Eco's `The Name of the Rose,'" for example).

To do that, though, you'd need a DTD.

As I've written about before, it gets even more interesting when you're trying to do more than just talk about how to lay stuff out. Layout's important. Knowing what links "really link to" is important. Talking about structured data on the wire--XML-RPC--is important. But there's still more to explore. So I'd like to ask you a favor. Rather than reiterating, now that it seems more pertinent to the current DTD discussion, I wonder if you'd be so kind as to post a link to my prior piece on semantics, which can be found at:


I'd very much like to see a discussion about how to attach semantics--which we might as well call scripts!--to XML. How do we go about ensuring congruence? If I use the XML parser in Frontier to attach semantics, in the form of UserTalk scripts, to some XML, how do I know that when that XML gets shipped to a friend's machine and parsed with a Java-based XML parser, that the Java code invoked by that parser will do the "same thing" (whatever that means) as the UserTalk scripts did?

I think this is why it seems that "big players" are trying to "centralize control" of XML: because there is a need for common semantics lest XML turn into a tower of babel, with people inventing new tags but not getting acceptance of those tags *and what they're supposed to mean* in any meaningful way.

Time for a Frontier-XML-Semantics mailing list?

This page was last built on 4/29/98; 12:01:28 PM by Betty Webmaster. webmaster@scripting.com