W3C home > Mailing lists > Public > w3c-sgml-wg@w3.org > February 1997

Re: Reference Implementation (Was: Re: XML catalog draft)

From: len bullard <cbullard@hiwaay.net>
Date: Sun, 09 Feb 1997 09:12:17 -0600
Message-ID: <32FDE951.5D42@hiwaay.net>
To: Paul Prescod <papresco@calum.csclub.uwaterloo.ca>
CC: w3c-sgml-wg@w3.org
Paul Prescod wrote:
> I agree that reference implementations are good, Len.
> I don't agree that they should be used in order to avoid standardizing
> some things. If we can agree on an appropriate default resolution
> mechanism, then we should just do so and put it in the spec.

It is not to avoid standardizing it.  It is to avoid standardizing 
something for which perfectly good alternatives can and will be 
proposed and from quarters outside the working group, the consortium, 
perhaps even the SGML community.  It is also to focus effort on 
an implementation that cn be widely shared quickly just as most 
of us use SGMLS and SP.  It is 
also to avoid a long and very rough debate in which the only 
proof would be running code anyway.  If the editors are to stick 
to their schedules, then a reference implementation effort may 
require a different approach than that used to manage this 
list.  If the consortium were to sanction that, then it would 
be possible to have several groups propose a design for an 
RI from which the best of breed can be chosen.

> On the other hand, if our resolution mechanism is so complex that there
> is a chance that vendors would balk at implementing it, then that's where
> a reference implementation would make sense. But maybe at that point,
> we've already made a mistake in making the mechanism complex.

They may not balk; they may not be interested.

Paul, everyone knew how to design HTML in the eighties.  Many understood
was required for HTTP.  NO ONE WAS INTERESTED.  Durand had it 
right when he said, "Berners-Lee said, let links break".
Sometimes the quick way is to just build it and field it.
It gets mindshare:  it creates market:  it makes substandard 
applications dominant.  Netscape and IE don't break HTML 
because they are *greedy blue meanies*; they break it 
because they have to in order to make progress.

This is what is about to happen:  several vendors see XML 
as an opportunity to get products they already had in development 
to market under a single banner.  There are those who think 
Panorama, HyBrowse, etc. represent viable implementations.
They may, but there are not references to test.  There are also 
different ways to build these just as there are better 
ways to build browsers than libWWW.  If we do not wish a ground 
war to break out, a single implementation such as libWWW exemplified, 
is a good way to measure these implementations.  If such an 
effort is begun soon, it will be possible to work out the 
technical difficulties of the XML spec before we get a lot 
further down the road.  Remember, SGML Open and other 
organizations have an interest here.

An RI doesn't mean, "this is the only way to do it"; it means, 
here is a library one MIGHT use and by which SOME tests can 
be conducted.  It is also a way to meet the criticism that 
we have specified things that can't or won't be adopted because 
no resolution mechanism was provided.  So, provide one, but 
leave some wiggle room for development.

Received on Sunday, 9 February 1997 10:12:16 UTC

This archive was generated by hypermail 2.4.0 : Friday, 17 January 2020 20:25:07 UTC