Do SGML and XML co-exist?

Len Bullard cbullard at
Thu Nov 18 02:18:43 GMT 1999

Matthew Gertner wrote:
> Yeah? I always tell the same story about the origins of XML, so I'd be
> interested to know if it is wrong. :-O Basically, goes the story, round
> about the genesis of HTML 4.0 the HTML folks started to realize that
> even with the best of intentions and a heck of a lot of hard work they
> were never going to get even a small proportion of the tags that people
> (and influential companies) wanted into the language.

Somewhere in the oldenMailFiles, I have the request from an influential 
HTMLer to help "get it on track, they are on a suicide march".  I read 
what the WG was doing and told Behlendorf that until they actually began 
to understand the limits of markup systems, it was hopeless, so he 
should start understanding SGML and drop the WebIsHoly religion.  Later, 
I got the sad occasion to explain to a Netscape product manager for 
the browser that Andreesen's bold statements in France about the 
"propritariness of XML" were the rantings of a dieing exec.  They 
waited for six months and by then, they were dead on the vine.  

Predictable, avoidable, and completely their own fault.  Pennfield 
can rule stupidly, but MS's competitors killed themselves 
with their own ignorance and pride and press.

> Moreover, the
> undesirability of this approach, even if practicable, was starting to
> become clear (witness the length of the HTML 4.0 spec).

Every markup specialist in the world knew as soon as HTML was released 
what it was, where it would work, and what the flaws were.  Everyone 
tries GenCode at least once.   In the original form, according to 
Yuri, it was end-tagged (a la RTF) and some folks had to work with 
the inventor to make it a markup application.
> So they started to dream up something along the lines of generic markup
> using ghastly DIV and CLASS tags. This came to the notice of the SGML
> folks, who screamed bloody murder and started lobbying for SGML, with
> its time-tested approach to generic markup, to be the new language for
> the web. 

No.  It started earlier than that.  SGML hypertext systems with
were already on the market prior to 1990.  Some systems had been being 
worked since ISO8879 was a draft.  Long before XML, the SGML community 
had debated it on comp-text-sgml.  The problem was not complexity but 
the inanity of the HTML working group.  Hytime and DSSSL, yes, these 
are complex, but to be blunt, put all of the XML specs together, 
throw them on the floor, and see how loud a thump they make.

Beware facile demonstrations of simplicity.  Bosak knows it was 
a show piece and that once the work was underway, it would 
fatten out.  It didn't matter.

> The reaction of the HTML crowd was immediate: not SGML, of all
> things, it's far too scary and complicated! So some forward-thinking
> types from the SGML community came up with the idea of simplying SGML
> significantly in order to make it acceptable as a mass-market web
> language. Whence XML.

People had been simplifying the application of SGML all along.  I had 
discussions with Charles Goldfarb about that at the very first HyTime 
meeting I attended. His answers were along the line that as long as 
the system was a proper subset, it did not violate SGML.  He thought 
it strange anyone insisted on implementing all of the features of SGML 
given that some were there for particular purposes.  It was Charles 
and Yuri Rubinsky that took IADS to a NIST meeting to prove that SGML 
could indeed be used for simple, cheap, and easy to apply hypertext. 
The reasons we were writing conformance tests for the US Navy was 
that none existed so no one knew how to buy an SGML system.  If you 
asked SoftQuad or ArborText or Datalogics, the multiplicity of 
answers was staggering.  Hungry, desperate competitors make 
bad advisors for conformance, The SGML Way, it was called, and with 
that, competitors eliminated simple solutions in favor of a spectre 
they couldn't themselves enunciate in concordance.   Once the money 
was there, it didn't matter.

XML was SGML On The Web when the idea was introduced at the Vancouver 
meetings.  Some of us had heard the rumblings and were ready to 
support it.  I for one was not prepared for the coup d'etat that 
followed but I could see it coming after an argument with Jean 
Paoli in the hallway about how to go about it.  He kept insisting 
"it must be simple, it must be easy" and I replied, "no one 
disagrees, but it must be SGML."  At that moment, I realized 
what was coming and that the SGMLers whose hopes that Microsoft 
sponsorship was a new hope would get what the wanted.  Yuri had 
told me a year or so earlier, "Yes Len, I know HTML is hopeless, 
but for the first time, SoftQuad is actually making money."  
> Am I close? Assuming so, technical issues certainly were the primary
> motivating factor for XML. 

Not really.  It was something doable technically and we had the 
existence proofs.  After that, it was a political exercise in 
divesting ISO of authority and conferring it on the W3C.  There 
are some stories you don't hear, Matthew, and some of them are 
involve some very old and bitter rivalries for control.  Those 
had weakened the community considerably.  The CALS days were 
fairly flush, but even then, the WYSIWYG Wars had weakened 
the Republic, so the advent of the Empire was easy to fortell. 

> Denying this is a pretty extreme view (not
> that this surprises me, Len :-). The "features of the practice" may not
> be all that different, but the fact that to write a conformant SGML
> parser you needed (according to one statement I heard at an SGML
> conference as XML was coming to the fore) "to have the resources of the
> U.S. government or to be James Clark"

Point was, we already had a conformant parser.  The issues revolved 
around the anemia of using the web transport mechanisms and stateless 
protocols.  SGML requires the use of the DTD to do some magick. 
Sensible designers knew that wasn't necessary.

> surely had something to do with
> the chorus of demands for simplication in exchange for ubiquity. Not
> everything is a money-driven conspiracy.

Not everything is.  IADS was designed and coded by a political 
science major and a photographer.  You see, sometimes, it doesn't 
pay to be a computer scientist contending with the More Meta Than 
Thou crowd.  Essentially, even when you have a full toolkit, 
you may only need a screwdriver.  

It doesn't matter now, but since defending XML against the 
ravages of possible upstarts such as SML sounds a lot 
like Stalin defending against Trotsky, it helps to have 
a few dialogues for historians to mulch later.


xml-dev: A list for W3C XML Developers. To post, mailto:xml-dev at
Archived as: and on CD-ROM/ISBN 981-02-3594-1
To unsubscribe, mailto:majordomo at the following message;
unsubscribe xml-dev
To subscribe to the digests, mailto:majordomo at the following message;
subscribe xml-dev-digest
List coordinator, Henry Rzepa (mailto:rzepa at

More information about the Xml-dev mailing list