Re: A28: syntax of markup declarations?
After thinking long and hard on this one, I suggest adopting Charles'
proposal or some reasonable facsimile thereof.
I brought this issue up with a colleague of mine, gave him the background,
goals, rationale, etc., then went on to explain the advantages of using
the instance syntax for the declarations. He looked at me with a puzzled
expression and asked: "Why? What's the gain?" I went on in my role as
proponent for something like three or four hours on and off over a trip
and an afternoon or two. He still asks: "Why? I don't get it." Maybe I
failed in my role, maybe my colleague isn't as wise as his many years
of software engineering experience and keen insights would have me
believe, but in the end, he didn't get it.
I also have to admit that I find it easier to type declarations in the
current syntax than I would in an instance syntax. The simple reason is
that the formatting of the declaration gives me clues when I have made
a syntactic error, just as they do when I'm typing declarations for some
programming language. I wouldn't have that with an instance syntax.
Finally, I would like to use my SGML tools for XML. The only concession
I've made so far is on the RE/RS issue which is one that I feel I can
effectively control. I don't want to have the issue of having to explain
an differentiate, yet another markup language when I meet with some
department to discuss SGML. If it were only the external DTD, I'd say:
"Fine, let's do it." But the sad fact is that the DTD extends into
the instance, and I am real finicky about what happens there.
Easy transformation is a bogus argument. If that is really the option,
then let's go all the way, gloves off, no holds barred client/server,
otherwise let's stick with the basics.
Robert Streich email@example.com
Schlumberger voice: 1 512 331 3318
Austin Research fax: 1 512 331 3760