Re: SGML objects

Steve Heaney (
Fri, 13 Aug 1993 20:52:53 +0200

> From Fri Aug 13 18:59:57 1993
> From: Dave_Raggett <>

Dave_Raggett <> writes:

> Jay says:
> > Speaking of sections, your point is well taken about HTML missing chances
> > to define structure, especially with sectioning tags. Could HTML+ define
> > an <H> tag, which is a container? Doesn't sound hard to implement, just
> > incremement and decrement a level counter and do the same thing as
> > explicit tags of that level. Browsers can choose a threshold beyond
> > which formatting doesn't change.
> HTML+ does indeed provide support for hierarchical containers via the
> "GROUP" element. This allows such hierarchies to span arbitrary numbers
> of nodes spread all over the world (thanks Tony). The element includes
> a role attribute to indicate the type of container, chapter, section, ...
> Lou's TEI DTD forces people to define headers in strict order, e.g.
> H1 H3 H2 is illegal as H3 is out of sequence. I think this is unnecessarily
> strict for our purposes.
> Dave Raggett

I hoped that the H1-Hn tags can be junked in favour of a single container
element with the nesting depth of this element being determined by the
browser, in order that the formating would reflect the depth of nesting.

I have seen it suggested that by using this scheme, in order to format a (lets
call it) meta-document which spanned several files and/or nodes, all files
would need to be retrieved, which would be an unnacceptable overhead.

I do not understand this argument (the need to retrieve all the sub-documents).
Would somebody be good enough to explain it to me because its bugging me, and
introducing a single <header> or <group> would seem to make sense.

I would have thought that because all tags are (or can be) paired that one
part of a meta-document (what I called a sub-document) could be formatted
sensibly in isolation to its other parts.