Net Buzz with Richard Wiggins | 2

…from transcriptions of ancient Sumerian scrolls to the technical documentation for stealth bombers, and from patients' clinical records to musical notation.

Yep, that's almost exactly what they've done, except that the Sumerian scroll research would actually use the same set of definitions as most other Arts & Humanities projects, as they share a common set of requirements. The bomber makers would share needs with other military hardware users. Medical users would have common needs too, and encoders of Bach would share requirements with encoders of Elvis, so it's more likely per-industry than per-topic.

That's entirely possible, although I suspect it will settle down to a few hundred in common use. But absolutely it lets you do this, in exactly the same way that other sets of standards and practices are established on a per-industry or per-topic basis. It shouldn't matter to the end user, as the display they see will hide the XML stuff, just like Web browsers hide the HTML.

What will happen in the short term is that it will be new, so user communities will want to experiment, and reassure themselves that they are distinct and unique, and couldn't possibly use other people's solutions. In the long term, once the newness has worn off, they'll discover that, hey, those guys have a tagset that does just what we want, let's share. So subsets of tags will settle down to a common core, plus some specialist sets in each user community.

So although in theory it's infinite, in practical terms there's only a limited number of ways people want to tag lists, for example: (a) lists with bullets, (b) lists with numbers or letters, (c) lists with keywords, (d) lists inside paragraphs, like this one, and some of the requirements overlap anyway. Generic ways of doing these things are called "architectures," and there's already a solid and well-established way to handle them in SGML and XML.