Standards, Work Groups, and Reality Checks: A Radical Proposal.

Benjamin Franz (
Fri, 22 Sep 1995 16:21:28 -0700 (PDT)

On Fri, 22 Sep 1995 wrote on

[example of yet another way where the approved way and practice in HTML
are now diverging deleted for space]

> They will not agree with you, because to them the latter is better! So,
> in order for us to work users towards wanting the actual standards--rather
> than personal browser standards--we will have to convince them that using
> TrueHTML is the better thing, which means it _has to be better_.
> The best way to do this is to educate the users, let them know that they
> can use <p align=center> over <center>, and that it is more accepted.
> Get some of the GOOD additions to HTML actually incorporated into the
> True standard, rather than holding out and being stubborn (such as
> border=xxx on images).

This was originally going to be an extension of Brandon's remarks. After
a few hours of writing, I came to conclusion that what I was really after
wasn't extensions of his remarks, but radical change in the thoughts
of the html-wg. So if this seems long, it is. If you want to get
to 'the beef', skip to the last sentence, and decide if you want to
read the rest.

The HTML development effort is now into about stage 6 of a project
(Punishment of the Innocent). How did it get into this rather sorry
state? Easy. By ignoring the most basic rules of writing a coding
standard that must interact with *NON*-technical users:

Rule 0: Either keep the design simple and obvious enough that
non-gurus can learn to correctly use it by reading
existing code or make it impossible to be used *at all*
without understanding the standard.

Rants about validating code or reading DTDs are utter bunk for 99.9% of
users. They not only won't - most of them have never even _heard_ of
validation or DTDs and couldn't use them even if they had. Most users
of the web *ARE NOT* programmers or computer wizards today - and the most
intensive education campaign will not make them into ones.

HTML fails either to make the standard simple enough or complex enough to
be 'unbreakable'. It has taken a middle ground that is simple enough for
people to *think* they understand it from examples, while actually being
complex enough that without actually reading the official standard you are
*CERTAIN* to write bad code. Since the DTD can be *implied*, extremely
few users even realize that there *is* a DTD; and a vanishingly
small percentage could understand that DTD if their very life depended
on it.

Rule 1: Don't play King Canute to the tide.

If every major browser company is implementing a feature *not* in the
and the *majority* of users are *using* them - the WG is play King Canute
if it refuses to try and actively attempt to integrate them into the
standard. Passive 'we'll do it when Company N give us a formal
description' for month after month is not going to cut it. At *LEAST*
four browsers have implemented these extensions now.

Playing King Canute not only doesn't work - it breeds contempt for those
stupid enough to play the leading role and insures people will listen
even *less* to them in the future. If most people aren't using your
'sooper-dooper' feature and demand different ones than you want to offer
- maybe they aren't the problem.

Rule 2: Finalize standards in a timely matter.

It simply does not matter if a _perfect_ standard comes out in 2032. If
you don't have a clear and stable _good enough_ standard when the
market needs it, the market will do it its own way, and the hell with the
WG. When this happens the WG is going to run into Rule 2 sooner or later.
Most likely sooner. FIG is a perfect example of this. The WG has been
promoting FIG for imbedded objects for about two years. It only got the
content model settled in the last month or so. IMG is and will remain the
champion for quite some time. Its replacement won't be FIG - it will be
almost certainly be EMBED.

Rule 3: "Lead, Follow, or get the hell out of the way."

Nearly every recent *successful* innovation to HTML, with the
possible exception of tables (depending on how generous you are with
regards to Netscape's implementation of same), has been brought to
fruition by people _ignoring_ the the WG and its quest for the 'perfect'
standard. That is as clear an indicator of a disfunctional standards group
as I can conceive of.

If the WG wants to understand why it is steadily becoming less and less
relevant to the evolution of HTML as is is *actually* used - all it has to
do is check the rules of standards development it spent the last two years
breaking. Not the IETF rules - the human interaction ones.

The effort to cram the ball of worms that HTML has become (and
honestly always has been) into an SGML compatible can is doomed for the
foreseable future. Not because a good SGML compatible standard can't be
written, but because the SGML driven group simply isn't in the driver's
seat anymore: The HTML driver's seat is now owned by browser and page

The HTML group is justifiably perturbed by the bizarre things claiming
Content-type: text/html. They want the newcomers to *be* test/html if
they say they are, and don't understand why they refuse to stand on some
other content-type if they aren't.

Its easy:

Most browsers read and do *something* reasonable with
text/html. Big potential audience.

Most browsers just D/L test/x-html to a file, assuming the
server I am using even knows about '.mynewextention and
experimental mime type'. Small potential audience.

Right or wrong, in market driven by *economics* rather than standards
- they are *going* to serve it up as text/html.

So how does the WG get away from this morass?

Content-type: text/sgml

It is time to quite trying to cram square-peg HTML into the round-hole SGML
mold when the *real* goal is SGML itself. It is time to apply Rule 3: The
html-wg *can't* lead on HTML and *won't* follow: it is time to get the
hell out the way.

The development of HTML of *all* levels should be officially ended
with the acceptance of HTML 2.0 (whenever that happens...) and left to
evolve in whatever direction the market carries it. I suspect a lightweight
hypertext PDL will be the result. It is time to move on to true SGML. If
SGML is (as I and many others really do believe) so much superior - it is
time to prove it.

I move to dis-establish the HTML working group.

Benjamin Franz