xml-dev - RE: standards development and community process

["Bullard, Claude L (Len)" <clbullar@ingr.com>]

Ok, but let's review how that was done so everyone 
understands.   What works once in different circumstances 
may or may not work again.  Sorry about the length but 
having been there for both, I guess I have some input:

Some history and comments:

1.  VRML started out on a mail list that Mark 
Pesce and Gavin Bell moderated and Brian Behelendorf 
administered.  The orientation of the initial group 
made all the difference.  They believed they could 
do it and they had a plan.

2.  The discussion was very focused.  If one began 
to stray into threads not related to the spec, Mark 
or Gavin could be very pointed in stopping that. 
And I do mean, pointed.  Tough skin required.

3.  No one was interested.  Let me explain.  Outside 
of SGI and a few small companies, no bigger concerns 
gave a hoot.  They watched to see if it turned into 
anything successful.  XML was different.  We knew it 
would work.  It became a sales job but as Bray has 
noted before, that wasn't that hard.  Companies knew it 
would work.  SGML proved it.  It just needed tightening 
and focus.  VRML 1.0 was Open Inventor --.  So like 
XML, simplifying an existing design and taking account 
for the platform (THE WWW) was the job.

4.  The authors stayed quiet (see 2).  The developers 
were semi-open about the code they were developing. Let's 
say they shared their embarrassment openly.  SGI 
kick started things with a parser.  Sharing and conformance 
were key.  After an implementation was available, the 
authors became full participants and their feedback 
was and is invaluable.  Someone has to use the thing 
but again, context.

5.  This was ONE application language.  No meta-ideas 
were introduced until VRML 2.0 (now ISO VRML97).  This 
helped the focus, so one might better compare it to 
SVG or HTML.  Narrow focus with no regard to larger 
issues helps and hinders, but to stay on schedule, 
it is vital.

6. Once we got into VRML97, there was a big list for 
discussions such as these and working lists for tightly 
focused projects.  Self-governance depends on discipline 
with respect to activity and context.  It can be difficult 
to pull together as the WGs spawn like fish and need to 
swim upstream over new obstacles to get back together. 
That integration work is the hardest work of all and 
folks like Connoly get my respect for doing it.

7.  It has been suggested that one reason the W3C did not 
embrace VRML was that it was not their spec (NIH) and 
because powerful members of the W3C (vendors) were driven 
back when they tried to muscle in.  There is something to 
what Simon is saying and don't think it doesn't happen; it 
is just a case-by-case thing and all is forgiven after the 
final drafts are approved.  For everyone's sake, practice 
Coyote and Sheepdog politics:  anything goes until the bell 
rings, then go drink beer together.  People are more important 
than technology.  Build bridges afterhours.

8.  Mistakes in design were made and some compromises 
compromised market share.  Designing a one size fits all 
usually results in coverage but not fit.  VRML is often 
cited as a failure but to date, it remains the only web 
standard for real-time 3D graphics.  It needed to be 
componentized because its emergence dovetailed the explosion 
in web browser size and that knocked it out of the download 
slipstream.  That has changed.  There are some really very 
nice VRML97 players out there now with component designs. 
Check out www.parallelgraphics.com for Cortona.  Or check 
out the Shout design, or the Blaxxun design.  I like cortona 
because it supports MP3 (so sue me, i am a musician).

ISO participation did not start until the second version 
(VRML97).  At that point, it was decided that an official 
home for the spec was needed and some of us successfully 
plead for ISO.  As I said then, leaving gold in the 
town square without guards is bad security, and the specs 
need maintenance.    There had already been a deal in the works 
between ISO and SGI over Open Inventor.  Rikk Carrey approached 
them about using VRML instead.  In the spirit of the emerging 
web, it carried.  Note that these issues were discussed and 
experienced people on the list made the points made here about 
a process.  Boy did we ever fight over process, but in the 
end, it came down to enabling the greatest number of voices 
to be heard as long as there were leaders capable of moving 
the discussion to closure around a consensus.  We also voted 
online.  Weird, but Jeffersonian democracy does work when 
the issues are clear and limited in scope.

Note that the ISO process is somewhat separated 
from the list process.  The Web3D Consortium handles the lists 
and its members keep track.  ISO works in concert with them 
and its representatives are members of the lists.  Nothing 
is hidden and it gets raucous.  What ISO does is make sure 
it keeps moving to a schedule and the editing meets spec.  
Note, the Technical leader actually has to take the punishment 
for that.  

We keep it open because that is how we choose to work and we pay for that 
with inefficiencies.  On the other hand, we don't tell 
people to bugger off and shut up unless they really really 
get on nerves, even then, no one removes them from the list. 
The challenge of maintaining a community is one we meet 
because it is the greatest accomplishment of all, far greater 
than the spec itself.  Tech comes and goes;  people count. 

Without the community, the spec is paper.  With the community, 
the spec brings coherence to our competition to outbuild 
each other.  It is to be relished, savored, and appreciated. 

Big powers have tried to muscle VRML.  It doesn't work.  
There is a period prior to choosing the VRML97 design 
which is interesting reading.  The design chosen was 
actually voted on by the list members at large.  This 
same tact has been useful at other times.  People learned 
how to negotiate to consensus, and if they embarassed themselves, 
they were no worse off than Wyle E Coyote at the bottom 
of a canyon.  They accept the potential to race the bird.

However, "as the twig is bent, so grows the tree."  VRML 
was started by individuals for whom community was the most 
important quality for a specification to become a standard. 
Without the consent of the governed, no governing authority 
can govern.   Because community, the prime ethos, has been 
such a driving force, and because anytime this ethic has 
been challenged, the member community cohered behind it, it 
has been easy to meet the challenge.   This has had mixed 
results.  

It took a year to get VRML1.0 out.  It took about 
two years to get VRML2.0 out.  VRMLnextGen is coming together, 
but the fight over the object model is fierce: the question 
being, is the DOM the right model for performance-centric apps? 
XML has been the most contentious issue we have had to deal 
with because it opens the community to the need to share 
by common means with a greater community, and yes, there 
is a lesson in that for those who think the W3C can be like 
the Web3D Consortium.

The problem with ISO is they want a standard.  The problem with 
designers is that they want fast, competitive software.  The 
challenge is to meet that in the middle with something everyone 
still cares about and is still marketable.  Gritty work.  ISO 
had to promise the spec could be freely downloaded.  Since it 
is a Web3D spec before it becomes and ISO standard, that was 
an easy arrangement to make.  It stays online and Dick Puk 
at ISO understands the community rules and abides by them.

Also, note that during the time when "no one cares", very small design 
teams ably lead pulled it together.   Behind the list, behind 
the ISO partnership, this is still the case.  Without the commitment 
of able individuals who fully understand the cost of commitment, this 
means does not work.  Even HTMLers have to admit that unless Andreesen 
and whoever else had not forced hands, we might have a different 
world today.  As I said, sometimes someone has to go BerserkerOnTheBridge 
and sometimes, a whole division has to knock them off the bridge. 
It isn't always a courtly game and those with courtly reputations 
preserve them often by withdrawing from the conflict and letting 
knights take the hard blows so they can then re-emerge, spread their 
courtly wisdom, and bow out unsullied and with the appearance 
of accomplishment.  Caveat Emptor.

Last summer, we redebated the VRML process issues as certain powers 
were unhappy with certain decisions.  We came to a good compromise 
that ensured progress without giving into a central polity that 
could ignore the lists at will.  Some folks are still unhappy 
about the results and wanted to overturn the votes.  It came 
down to Neal Trevett, President, standing up and saying, "nope, 
they voted, we heel".  And on we go.  At any time, that could 
all change.  Attention is required.

It is easy to create when the eyes of the world are elsewhere.
Real time improvisation is tough.  It takes practice and grounding 
in theory with real chops.  Experience counts.  Style counts.  
Above all, butts in seats count.  No butts = no applause = no $.

Len 
http://www.mp3.com/LenBullard

Ekam sat.h, Vipraah bahudhaa vadanti.
Daamyata. Datta. Dayadhvam.h


-----Original Message-----
From: KenNorth [mailto:KenNorth@email.msn.com]

Would the community process be the vehicle for accomplishing what VRML
developers accomplished -- low barriers to entry and a freely-downloadable
ISO spec?