• From: "Rick Jelliffe" <rjelliffe@a...>
• To: xml-dev@l...
• Date: Tue, 12 Dec 2006 13:54:32 +1100 (EST)

Michael Kay said:
>> Rick Marshall Wrote: This is a long way of saying that I
>> think trying to get one tool to do 2 fundamentally different
>> things is always going to less than optimal.
>
> If I want to restrict colour to be red, blue or green, is it fundamentally
> different whether I do that using a grammar or using a predicate-based
> language?
>
> Perhaps if I write
>
> colour -> (red | blue | green)
>
> then I don't even know or care about the difference...

Exactly right: the questions of how you write (or a forced to write) and
how something is implemented are distinct.

And in most (but not all) cases it is certainly possible to convert from
one to another: I had a recent blog entry on converting grammars to
path/predicate constraints for example. It is possible to write an XSD or
RELAX NG schema and implement it using Schematron+say DTTL, for most
common kinds of content models. (And it is possible to make some kinds of
grammars from some kinds of Schematron schemas.)

However, if type attribution and validation implementation eventually
reverts to using paths/predicates (as I expect they will, to avoid double
handling) and it turns out that the kinds of constraints that grammars
express are not actually the particularly interesting or gotcha-ing
problems, then why maintain the grammar facade? That's why I think
grammars are promoted too high up the food chain: they are put as the
central modelling artifact rather than one possible data capture tool good
for some niches.

Cheers
Rick Jelliffe

• References:
  • RE: Rick Jelliffe quotable quote on the purpose of schemas
    • From: "Greg Alvord" <galvord@o...>
  • RE: Rick Jelliffe quotable quote on the purpose of schemas
    • From: "Michael Kay" <mike@s...>