I think the gist of the last sentence is this:
Consider to different contexts where a particular syntactic construct is used. The valid uses in the two contexts might be represented by two vectors U1 and U2:
U1 ^ - - - - - | | | A1 | | | .----------> U2
Now, since the DTD is context free, it has to allow the convex hull (here repesented by the area A1) in order to allow both U1 and U2. Clearly, this superset is much larger than the desired set (containing only U1 and U2).
If the uses U1 and U2 in the different contexts can somehow be brought _closer_ to each other, we can end up with something like this:
_U1' /| / -->U2' /A'-- / -- .--
Now A' is much smaller, and we have achieved a smaller (more optimal) superset just by changing our language a bit.
/ Marcus Comstedt (ACROSS) (Hail Ilpalazzo!)
Previous text:
2002-10-18 21:31: Subject: Autodocs in the latest 7.3
I don't understand your last sentence.
The difference is perhaps that I consider such a superset-dtd to still be of some use, or at least be better than saying "no, it's not possible to write a dtd for this".
/ Martin Stjernholm, Roxen IS