Collection of Internet/Collection of Internet.iso/infosrvr/dev/www_talk.930/[e-mail] _Wed Dec 2 03:09:11 1992.msg

Internet Message Format | 1994-01-24 | 4KB

Return-Path: <[e-mail]>
Received: from dxmint.cern.ch by  nxoc01.cern.ch  (NeXT-1.0 (From Sendmail 5.52)/NeXT-2.0)
    id AA05557; Wed, 2 Dec 92 03:09:11 MET
Received: by dxmint.cern.ch (5.65/DEC-Ultrix/4.3)
    id AA03629; Wed, 2 Dec 1992 03:22:20 +0100
Received: from pixel.convex.com by convex.convex.com (5.64/1.35)
    id AA09445; Tue, 1 Dec 92 20:22:17 -0600
Received: from localhost by pixel.convex.com (5.64/1.28)
    id AA20569; Tue, 1 Dec 92 20:22:15 -0600
Message-Id: <[e-mail]>
To: [e-mail]
Subject: HTML providers: please grab sgmls and the DTD
Date: Tue, 01 Dec 92 20:22:15 CST
From: Dan Connolly <[e-mail]>

Tim,

I noticed you've been diddling with the HTML files
on info.cern.ch -- quoted your attributes, dotted
your i's and crossed your t's, so to speak.

But the files still don't fit into SGML.

Please: grab the sgmls-0.8.tar ... wait! there's a 1.0!
on ifi.uio.no in /pub/SGML/SGMLS. I have experience
with the sgmls-0.8.tar.Z -- I know it builds and runs
like a champ. I'll have to check the 1.0 version out.

Then check a few of your files by doing

% sgmls -s html.dtd yourfile.html

If you get errors, either fix your software or
diddle with the DTD until you get something that
works. Even if the official HTML dtd is never
modified to match your data, you'll have _some_
SGML DTD that specifies what you produce.

A word of warning: there's a rule-of-thumb everybody
ought to follow in writing DTD's: only used mixed
content for repeatable-or groups. That is, if you've
got an element that can contain a mixture of PCDATA
and elements, don't impose any special order on
the elements.

For example: I wanted to do this:

<!ELEMENT HTML - - ((TITLE? & ISINDEX?), (H1|H2|DL|OL|XMP|#PCDATA)+)>

But then the HTML element would have mixed content (see
the TITLE, H1, elements along with #PCDATA). The problem
is that if you write:

<TITLE>foo</title>
<ISINDEX>
<h1>a header</h1>
some data...

The newline after </title> is treated as PCDATA, which is not
allowed before ISINDEX. The options you're left with are:

1. Watch out where you put newlines (we hate this.)

2. Use a repeatable-or group, like this:
<!ELEMENT HTML - - (TITLE | ISINDEX | H1|H2|DL|OL|XMP|#PCDATA)*>
Then all your structure is gone (TITLES are allowed anywhere...).

3. Only use element content (move the #PCDATA out
    of the HTML element) like this:
<!ELEMENT HTML - - ((TITLE? & ISINDEX?), BODY>
<!ELEMENT BODY - - (H1|H2|DL|OL|XMP|#PCDATA)+)>
This is what I did.

[Notice that these tags are minimizable,
so that existing HTML documents that don't have them are
still legal. This was perhaps a bad idea. For example, I
had to hack the BODY model group badly to get the desired
effect.]

I've wrestled quite a bit with the header/body thing, and
I can't seem to come up with a DTD that 1) makes most existing
HTML legal, and 2) imposes _any_ structure on the thing.

I'm just about to give up on the structure business. Do any
implementations have problems with <TITLE> elements in
the middle of the document? If not, I can just change the
DTD so that HTML is just "tag soup" -- anything goes anywhere.

Another idea is to move the TITLE and other header
info outside the HTML format and MIME messages for
HTTP transport, so that

<TITLE>the title</TITLE>
<h1>the header</h1>
... the text...

becomes

Subject: the title
Content-type: text/x-html

<h1>the header</h1>

The header information would be part of the protocol, and
not part of the HTML data. Local files would get their
header information from the filesystem. WAIS documents
would get their header information from the :document
structure.

Anyway... get the sgmls parser and validate the data
you provide! I'd like to have more folks than just
me bumping up against the SGML issues.

Dan