Re: HTML should not be a file format, but an output format

Paul Prescod (
Tue, 25 Mar 1997 12:52:24 -0500 (EST)

From: Paul Prescod <>
Message-Id: <>
Subject: Re: HTML should not be a file format, but an output format
To: (Steven Champeon)
Date: Tue, 25 Mar 1997 12:52:24 -0500 (EST)
In-Reply-To: <> from "Steven Champeon" at Mar 25, 97 11:56:47 am

> At 07:30 PM 3/24/97 -0500, Paul Prescod wrote:
> >But the great majority of the information on the Web does not belong in
> >HTML as a source format. It is this situation that we seek to correct
> >when we speak of SGML "vs." HTML. It is a jargon shorthand such as that
> >employed by any industry in trying to simplify communication. It only
> >confuses a small minority. Those who are familiar with SGML know "the
> >code". Those who are not familiar with SGML typically do not know that
> >HTML is SGML and thus do not see anything stranger about talking about
> >SGML "vs." HTML than about PDF "vs." HTML.
> The "small minority" which you confuse encompasses all users of the
> Web having no familiarity with SGML. I am familiar with SGML, and I
> know "the code" as well. I am resistant to the idea of using code
> when the idea is to evangelize SGML. You have already alienated the
> audience you wish to embrace.

Not so. Those users of the Web have never heard of "DTDs" and thus would be
confused by talk about them. I consider it much more confusing to use the
media's way of looking at it where HTML is a "subset" of SGML than it is
to just distinguish between them. I wish that they would contrast SGML and
HTML rather than convincing people that SGML is HTML++.

At some point in the education of an HTMLer into the world of SGML there will
confusion. We can move it right to the front, by describing SGML as a 
meta-language and HTML as an application of that meta language every time
we discuss SGML, or we can move it to the point where they are actually 
curious about SGML and are willing to expend some mental energy to figure
out what a meta-language really is and why it would be interesting. I 
personally think that it is better to wait until they are curious than to
scare them off at the front-end. Describing SGML as a meta-language and
HTML as an application of that meta-language in casual conversation about
the two strikes me as taking a New Math approach: we could teach you about
mulitplication and addtion, but let's start by talking about commutativeness,
associativeness and the rest of set theory -- those are the real 
underlying concepts.

Either way, the difference between them recognizing the relationship between
SGML and HTML today or next week is fairly small, in my opinion. It would
not really matter if they never figured it out: the HTML that they "know"
is not much of a starting point for learning SGML. It is a DOCTYPE-less, 
ENTITY-less, mash of tags that may even allow overlapping tags in the right 

When I teach SGML I mention the relationship to HTML late in the course: the 
same for my Introduction to SGML document. The other nastiness in HTML is
how do you explain all of the formatting constructs in relation to the 
SGML Philosophy? Once again, it seems better to do it later, rather than 

 Paul Prescod