Re: [xml-dev] XHTML 2 Working Group won't be renewed?

["C. M. Sperberg-McQueen" <cmsmcq@blackmesatech.com>]

On 9 Jul 2009, at 16:04 , Peter Hunsberger wrote:

> On Thu, Jul 9, 2009 at 4:25 PM, Jim Tivy<jimt@bluestream.com> wrote:
>> Not a good idea for an open interoperatable data format.
>
> If part of the spec is a way to figure out what delimiters are being
> used in any given instance, then I'd argue the exact opposite.  I
> believe this was the case for SGML (anyone?)

The definition of variant syntaxes is indeed allowed
by SGML.  At least one SGML processor supports it in
full generality, or claimed to the last time I looked.
I don't believe I can name a second parser that supports
the feature in all its gory detail.  The definition of
variant syntaxes is hampered by the fact that ISO 8879
gives no particular guidance (that I can find) about
how to ensure that the resulting grammar is unambiguous
or can be parsed with bounded lookahead.

Personally, I think that giving SGML the ability to
define variant syntaxes may well have been necessary to
get ISO 8879 (the SGML spec) finished and generate the
necessary consensus around it; I suspect that some people
were willing to work on it because they expected to use
variant syntaxes a lot, and some people may have decided
not to oppose it because of the variant syntax feature.
But apart from things like the maximum length of names
I don't remember the ability to support variant syntaxes
being really important in SGML practice for most users.
(Software Exoterica's work on the Cinemania project may
be a counter-example, but I'm not sure they used a
variant syntax.)  In practice, faced with the choice of
defining a variant concrete syntax that would make
troff or DCF script (or even DCF GML) a conforming
SGML syntax (and then what -- claiming that troff or
DCF was a conforming SGML processor?), or just switching
to the reference concrete syntax, the marketplace voted
overwhelmingly with its feet and adopted the reference
concrete syntax.

Based on that experience, in this case I think Jim Tivy
is right:  variant concrete syntaxes are usually more
trouble than they are worth.  They impose a cost on all
implementors (and all users) of the language, and provide
benefit either to a very small number of users or to none.

YMMV, of course.

Michael Sperberg-McQueen


-- 
****************************************************************
* C. M. Sperberg-McQueen, Black Mesa Technologies LLC
* http://www.blackmesatech.com
* http://cmsmcq.com/mib
* http://balisage.net
****************************************************************