Truthfully, the concept of allowing start-tag/end-tag pairs that are not properly nested doesn't seem like, in the universe of data concepts, a hugely important concept. I guess that I am not seeing its importance.
One of the key reasons that the ODA community fought against SGML was that they considered it essential to maintain both the logical structure of a text (the chapters, sections, and paragraphs) as well as the paginated structure; those are two different hierarchies over the same data. Perhaps they failed because they were being too ambitious -- simple things that handle most of the requirement always tend to win in our world -- but just because we can't do it doesn't mean that no-one wants it.
Early in my career I was involved with data rather than documents, and we fought hard for the idea that data is a network, not a hierarchy. People like to see it as a hierarchy because our brains want to work through things sequentially (hierarchies can be flattened more easily than networks). But Wittgenstein was right; the way we think about things is conditioned by our language.
Bill Kent's book "data and reality" is one of the best books on data modelling ever written, because he understands this. He knows that when we construct models, with assertions like "every employee is based in one location" we are ignoring the messiness of the real world, and more than that, we are trying to eliminate the messiness of the real world: we are forcing the real world to conform to our model of it. Text consists of a sequence of characters from a finite alphabet because that's the only way we know how to model it, and by modelling it that way, we force real texts to conform. Similarly, MusicXML contrains the music that composers are able to write.
Michael Kay Saxonica
|
