Note that the approach of keeping a body of data minimally marked up,
and then making associations from extrnal markup, is supported in SGML
using HyTime DATALOC links. In addition, for some types of material, a
combination is possible. For example, you could mark up paragraphs if
they were unambiguous (they generally aren't always...) and give each
paragraph an ID, so that poionting to spans of data within them (or
across them) would be a little easier.
You can experiment with this if you have SoftQuad Explorer or (more
economically, perhaps) SoftQuad Panorama Pro, and use annotations in
`personal webs'. When you save the web file, you'll see a set of just
such HyTime links.
This is, I think, a very good approach when your markup may be
contentious, or subject to multiple interpretations. If anyone is
interested, I am speaking at SGML 96 on the subject of representing
ambiguity in SGML, although I am deliberately trying _not_ to place it
in the TEI context, because I want to try and address the ``prescriptive
crowd'' and make them more aware of the issues.
Liam Quin, SoftQuad Inc | lq-text freely available Unix text retrieval
[log in to unmask] +1 416 544-9000 | FAQs: Metafont fonts, OPEN LOOK UI, OpenWindows
SGML: http://www.sq.com/ | We've moved; new 'phone number & postal address!
The barefoot programmer | `who is my neighbour?'