[XML-DEV Mailing List Archive Home] [By Thread] [By Date] [Recent Entries] [Reply To This Message]

RE: SGML queries

• To: 'Joe English' <jenglish@f...>, xml-dev@l...
• Subject: RE: SGML queries
• From: "Bullard, Claude L (Len)" <clbullar@i...>
• Date: Tue, 10 Sep 2002 16:30:23 -0500

Yes, I forgot about those.  I don't think I ever 
believed them when I saw them, and then I tested 
it in a browser and felt a creeping panic. :-)

To me, the way that the HTML DTD used parameterization 
to grandfather the looseness of the tag stackers was 
proof that we needed XML and we needed to disallow 
some of the freedoms of SGML.   That probably seems 
contradictory to my usual positions, but I never 
promised consistency in the face of the real world 
anyway.  

<aside>Before anyone thinks this is a blast at 
the HTML world, try to read that section I cited 
for John Cowan and remember that the Handbook is 
the annotated explanation for ISO 8879.  It will 
really make you appreciate James Clark or anyone 
else who wrote a conformant SGML parser even more 
than you do as well as authors like Martin Bryan.</aside>

The ISO standard comes up for review cyclically. 
Keep that rewrite somewhere you can find it.  One 
never knows.

len


From: Joe English [mailto:jenglish@f...]

Bullard, Claude L (Len) wrote:

[ re: elements which allow both start- and end- tag omission ]

> *Bonehead elements* might be good.  Other than in
> the SGML Handbook, I've never seen these used
> in practice.

You're forgetting the HTML 1 backwards-compatibility hacks
in versions 2.0 through 4.01:

    <!ELEMENT HTML O O (%html.content;)    -- document root element -->
    <!ELEMENT HEAD O O (%head.content;) +(%head.misc;) -- document head -->
    <!ELEMENT BODY O O (%flow;)* +(INS|DEL) -- document body -->

> [...]
> "Although the basic principles of start tag
> ommission are reasonably straightforward, the
> detailed requirements and definitions are highly
> technical.  They hinge on the concepts "contextually
> optional element" and "contextually required element". p 163
>
> Ugly stuff.

You can say that again.  Ugly, ugly, ugly!

By the way, start-tag omission is the *real* reason that
content models are required to be deterministic in SGML.
If it weren't for that, the definition of "contextually
required element" wouldn't make any sense.  Other than
that, there's no good reason for the restriction.  (The
handwaving in Annex H does not constitute a good reason.)

I once came up with a compatible restatement of the
rules for start-tag inference that would fix this,
but apparently nobody was interested in revising the
ISO standard to fix it.

• Follow-Ups:
  • Re: SGML queries
    • From: Arjun Ray <aray@n...>
  • Re: SGML queries
    • From: John Cowan <jcowan@r...>

• Prev by Date: Re: SGML queries
• Next by Date: RE: SGML queries
• Previous by thread: Re: SGML queries
• Next by thread: Re: SGML queries
• Index(es):
  • Date
  • Thread