RE: Industry Standard's Net 21

Robert S. Thau (
Mon, 28 Jun 1999 12:39:27 -0400 (EDT)

Gavin Thomas Nicol writes:
> Mr. Thau
> > (Also, I suspect his attitude was somewhat colored by a justifiable
> > distaste for the goings-on in the IETF HTML working group, which had
> > become dominated by SGML advocates who didn't seem to understand or
> > care about the needs of the HTML installed base --- a divergence which
> > became so pathological that the IESG eventually pulled the plug.
> Hey. I don't remember you being particularly active in the HTML
> discussions, and I am probably one of those smelly, festering SGML
> advocates you allude to.

I followed the HTML-WG with interest, but I put my efforts elsewhere,
where I thought they might do more good. They seem to have had some
small effect; Apache's doing rather well these days. (And, on the
standards tip, the existence of a widely used, open source
implementation of HTTP has, IMHO, done quite a bit to keep the
commercial vendors honest).

> My recollections of the early Netscape participation in standards work
> is of arrogance, disdain, and NIH syndrome. If they had listened,
> stylesheets and other such things might have happened much earlier.

My recollections of HTML-WG and related discussions are of arrogance,
disdain, and NIH syndrome on all sides. You mention CSS below, for
instance, as something praiseworthy, yet because it came out of the
WWW community, and wasn't DSSSL, there was some measure of open
hostility towards it. See, for example,

and subsequent discussion. And that's in a case where there was no
basic disagreement about long-term technical goals.

Then there was the regrettable tendancy of the two sides to talk past
each other. For instance:

> I remember one meeting in Boston, fairly typical, where people were saying
> CSS was necessary to avoid tag explosion on the WWW. My response
> was "Hey, tag explosion might just be the best thing to happen to the
> WWW" meaning that generic markup is *necessary*. I got a whole
> roomful of snickers from people who are now rabidly embracing XML
> along *with* CSS.

You may have meant "generic markup is necessary", but what you *said*
was "tag explosion might just be the best thing to happen to the WWW".
What "tag explosion" meant, to the people who were involved in
actually putting up web pages at the time, was the proliferation of
idiosyncratic presentation hacks enabled by tags which were supported
only in one vendor's flavor-of-the-month browser. This was obviously
*not* "the best thing to happen to the WWW"; it was, in fact, a
serious problem. If you wanted those folks to take you more
seriously, you would have done better to avoid seeming to advocate it.
(Yes, I know what you meant; the problem was proliferation of hacks,
not proliferation of tags per se. My point is that most of the room
probably didn't. How could they?)

But there's a deeper problem here. You say "generic markup is
*necessary*", with emphasis. But it's not universally necessary ---
the web still mostly gets along without it. Many people building
complicated web sites are jumping through unnecessary hoops because of
that (the most common being translation from application-specific
markup which their servers store internally), but quite a few web
authors, with small sites or simple needs, aren't; generic markup, in
any form, is simply not necessary *for them*. Stylesheets and
purpose-built DTDs are overkill for someone who just wants to throw up
a few simple pages. Yet those authors do have a legitimate desire to
have some control over presentation. And it was their needs that
Netscape was trying to respond to.

The system that resulted, of course, did not suit the needs of large
publishers. But it couldn't have, for the simple reason that for
large publishers, purpose-built instantiation of generic markup *is*
necessary, and HTML, by its very nature, is not generic. There is
only one HTML DTD. And its role, as a lowest-common-denominator
format for delivery of widely disparate kinds of text, is very
different from the role of most SGML DTDs, because it can't try to
encode the semantics of the text to the same degree --- there's just
too much variety in the application domain (all forms of commerce,
entertainment, and instruction) to allow for that.

In particular, if you want to allow for presentation control (which
was, once again, a demand of the marketplace), you have to be willing
to allow it to be specified a little more explicitly, because you
can't trigger presentation magic off semantics which just aren't
there. Which is what Netscape found themselves doing, very badly.
But much of the SGML "priesthood", as you yourself call it in another
note, seemed to be trying to tell them that they shouldn't be doing
that, rather than trying to help them do it better.

For generic markup, you all eventually did the right thing --- blow
off the single DTD, and design XML. It's a standard for generic
document delivery which is lightweight and regular enough to be easily
deployed on the web, in those cases where it's actually worth the
trouble to come up with a purpose-built DTD. I've used it. I like
it. You did a great job with it. But HTML was never going to turn
into that --- it's only one DTD, remember --- and I don't personally
think that an HTML working group was a good place to, in effect, try
to start designing something else.

> I don't mind the history (can't be changed) but I take offense at people
> making statements like the above.

No worse than what you're saying about the folks on the other side
of the table.