Re: Semantics of <Hn>

Nathan Torkington (Nathan.Torkington@vuw.ac.nz)
Thu, 12 Aug 1993 23:02:58 +1200


Klaus Harbo writes:

> Fortunately this not an issue of opinion, since SGML DTDs define the
> structure of document instance parsetrees unambigously (provided, of
> course, that DTD was designed properly).

I see I didn't make myself clear enough :(. I meant that people seem
to be under the impression that
<H1>globglob</H1>
textytexty
<H1>globglobglobglobglob</H1>
is in some sense a definition of a semantic region that lies between
the </H1> and the next <H1> tags and thus that headings are semantic
rather than presentation tags. I would classify this sort of heading
as being a presentation-mainly quasi-semantic tag. If a region has a
heading associated with it, it should be defined as a region ie
<REGION><H1>heading1</H1>
text of the region
</REGION><REGION><H1>next heading</H1>
...

This means that the <H> tags have presentation only value under the
current DTD, and therefore they should be allowed (as I said) wherever
paragraphs are, as well as between items in a list.

> Of course I realize that the quoted statement is intended to make the
> information providers lives easier, but the real solution for that
> problem is to give providers tools that will let write and translate
> their stuff to a proper, unambigous format; the solution is not
> letting them "do what they want" with regard to tagging.

Darn true! The short-term solution of saying ``oh well, crappy markup
should be tolerated so the iddle widdle authors don't get brain
strain'' leads to automated software tools (which, despite my prior
rantings to the contrary, are still vital) not being useful because
they can't extract the semantics from crappily-marked-up documents.

Nat