Net Buzz with Richard Wiggins | 3

One of the problems with metadata is that doing it in a formalized way on computers is relatively new, so we're still discovering things we didn't know before (and XML is new, so there are discoveries there too). Metadata itself isn't new: the Romans had it, and medieval legal manuscripts have more metadata than data. But there was a certain amount of conflict between the library people and the computer people to start with, I think, as each community learned from the other. That's past, but there's still a lot of work to do in finding out what the most usable format is.

XML itself is just a language to let you specify a structure for different classes of document, expressed as a set of tags, so it will handle metadata if that's what you set it up to do. Compare it with asking if C++ will index a phone book -- sure, if you write the program. Where XML scores is that it makes it easier to write an application, so writing one which caters for metadata is easier too.

You can see some of this emerging in the list of new XML applications: finance and software are two at the top, with commercial transaction apps for general business use pretty close. Scientific apps are up there (chemistry was actually the very first with CML), and then there's a huge cluster of vertical markets: real estate agent markup language, for example, and furniture/woodwork industry markup language. One of my favorites is DESSERT, a markup language for recipes announced at SGML/XML'97. Don't laugh -- the cookery book/magazine market is huge.

This is essential for use in languages other than English. It lets a German have element types like <Hduser>, but even more importantly, it lets non-Latin-alphabet languages use 16-bit or 32-bit coded character sets, so Japanese users can have element types that can't be represented in a character set assuming case. There are significant problems if you try case-folding with non-8-bit alphabets, because some scripts simply don't have the concept of "case,", and some have it but only for some symbols. So the simplest and most useful thing to do was to remove case-folding completely. This also solves the perennial battle between the SGML rule of fold-to-uppercase (IBM legacy) and the UNIX practice of fold-to-lowercase.

If users have got start-tags and end-tags which currently match case (ie <UL>...</UL>, <ul>...</ul>, <Ul>...</Ul>, or <uL>...</uL>), then there's no problem. It's when they've hand-edited it and gotten <Ul>...</uL> or <uL>...</Ul> that it causes trouble. The vast majority of HTML pages are reasonably consistent, though.

There will be "tidy-up-your-HTML-and-make-it-XML" programs too. Any first-year CS student should be able to do a program to consistently upcase or downcase the element type names in the tags of an HTML file ("making" it into XML is not much harder).