xml-dev - RE: Premature standardization considered harmful

["John D. Gwinner" <JGwinner@dazsi.com>]

Len:

Hmm ... might have send this to the wrong list, I'm going to have to check
some things.  Ouch, it got put in my 'SOAP' folder when it was sent to the
XML list.  whoops.

> John D. Gwinner wrote:
> >
> > I agree, but I'd add WITH Implementation; I learned that from
> VRML.  A spec
> > can be hashed out quicker in an open process (with some semi-formal
> > procedures, IMHO), then published as a standard, but it's got to have
> > working implementation.  VRML did all of this except the last,
> and it was
> > barrier to successful user adoption (regardless of discussions about the
> > merits or lack of the spec itself)
>
> Be fair John.  It was implemented by Live3D, Cosmo, WorldView, IDS, and
> half a dozen others easily.

That's true, but not until after the spec was finalized, which is what my
point was.  To be fair, I did stretch the point a bit - there are other
issues with VRML that probably don't belong here.

I guess I'm saying that an open source reference implementation is *very*
useful before finalizing a spec.  When the rubber meets the road you
discover things that no amount of analyzation can determine.  This was
discussed with VRML but never was a completely resolved issue - and frankly,
I'm not sure it could be; even flat 2D graphics are not completely
standardized with respect to viewing on different platforms (most web design
firms standardize on a sub-set of colors for just this reason).

I don't actually think the reference implementation has to be open source,
it could be from two competing vendors; I think CORBA went through some of
this with state of business (as opposed to state of art) implementation
later influencing future spec.  (Not a CORBA expert so I don't know)

> The nit was the spec wasn't edited fast
> enough so it bogged down because the undefined areas caused serious
> content interoperability problems.  The problems of the event model
> indeterminacy
> were and are mind boggling.

Exactly - actually *doing* it was harder than *creating it*.  This is
exactly the reason I feel that someone should try it before finalizing
anything.

I think the difference here is that you're saying finalize it then edit it
to fix implementation issues; I'm saying work on it, demonstrate it at least
a bit, then finalize it.

Another example (again extreme to make my point <G>): C++ vs. Ada.  The Ada
spec was out before compilers or so I've heard; with C++, the other way
around, and the implementation process helped finalize the spec.  (I bet I
get nitpicked on this one!)

> Consider, it is five years later, it is still the only 3D spec for the
> web,

I agree, and I think VRML itself is still the only way to go - I didn't want
anyone to think that I'm not - I just stretched a point to agree with Don
originally.

		== John ==


***************************************************************************
This is xml-dev, the mailing list for XML developers.
To unsubscribe, mailto:majordomo@xml.org&BODY=unsubscribe%20xml-dev
List archives are available at http://xml.org/archives/xml-dev/threads.html
***************************************************************************