[
Lists Home |
Date Index |
Thread Index
]
Le jeudi 29 juin 2006 à 00:15 +0100, Michael Kay a écrit :
> I'm surprised though that Xerces doesn't catch violations of this particular
> rule: it's normally very conformant.
It used to catch these violations back in 2001 when I wrote my book but
latest versions do not catch them any longer. I guess that's what is
called a regression unless the Xerces developers have got yet another
interpretation of the rec...
One of the things I find most disturbing with W3C XML Schema are the
number of possible interpretations of the rec.
Even on W3C mailing lists, W3C XML Schema editors do never answer yes or
no to a question but "my understanding is that".
The W3C XML Schema recommendations look like an obscure holly book which
requires an army of priests and scholars to interpret them and which
interpretation is changing over time.
Let it live ten more years and its implementations will be radically
different from the ones we have now :) !
By contrast, RELAX NG is clear and can be read without having to
interpret anything.
The RELAX NG specification could be used a model for many other
specifications. Its good practices are:
1) Focus on a specific domain and stubbornly refuse to add any feature
that doesn't belong to that domain.
2) Separate the syntactical sugar from the core features. A way to
achieve that is to define a "simplification process" during which all
the syntactical sugar is removed to keep a highly normalized version of
the model.
3) Define the semantics of the core features using a mathematical model
(the semantics of the syntactic sugar are defined by the simplification
process).
And to support the spec:
4) Provide an extensive test suite (this is easy since your design is
clean).
5) Provide detailed implementation notes.
I wish more specifications could follow these principles.
I have been playing with 5 RELAX NG different implementations and have
never seen interoperability issues like those of W3C XML Schema and I
don't think that the reason is that RELAX NG implementers are more
clever than W3C XML Schema implementers but just because the
specification is much easier to implement.
And I don't think anyone who have seriously studied both specifications
can disagree.
Eric
--
GPG-PGP: 2A528005
Don't you think all these XML schema languages should work together?
http://dsdl.org
------------------------------------------------------------------------
Eric van der Vlist http://xmlfr.org http://dyomedea.com
(ISO) RELAX NG ISBN:0-596-00421-4 http://oreilly.com/catalog/relax
(W3C) XML Schema ISBN:0-596-00252-1 http://oreilly.com/catalog/xmlschema
------------------------------------------------------------------------
Ceci est une partie de message=?ISO-8859-1?Q?num=E9riquement?= =?ISO-8859-1?Q?_sign=E9e?=
|
