[Prev][Next][Index][Thread]

Re: Name groups in ELEMENT and ATTLIST Decls



On Tue, 25 Mar 1997 13:51:40 -0500 Martin Pike said:

>I have ploughed through the mail archive of this group looking for
>any arguments for or against the removal of name groups in the
>element type part of an ELEMENT and ATTLIST declaration.  i.e.
><!ELEMENT (a|b|c|d)  ...........
>
>I was suprised to find none, considering how widely used this
>construction is in SGML DTDs.  I have looked at many of the popular

The main argument against is simple to describe:  anything you can
say with name groups of this sort, you can say without them.  And
eliminating the construct eliminates the need to describe and explain
it in the spec, and the need for readers to understand it.

In any language, every new construct imposes some burden of
documentation and complication.  In a language designed to be as
lightweight as possible, complications of any kind are very hard to
justify.  Complications which do not provide new functionality are
and should be even harder to justify.

Those most interested in making XML-compatible DTDs maintainable had
(and continue to have) other fights on our hands, to defend the
existence of parameter entities and conditional sections.  Me, I
don't even find name groups useful even in full SGML; I think they
complicate things more than they help, and every time I consult the
HTML 3.2 DTD I curse the fact that I can't just search for '<!ELEMENT
foo' to find the relevant declaration.

So while I see your point, I for one remain convinced that the
decision on this question was the right one.

-C. M. Sperberg-McQueen