Jirka Kosek <[EMAIL PROTECTED]> writes:

> I see this as a big limitation. XML files should be self-descriptive, 

I do not believe that tacking a schema-association PI onto an XML
could rightly be called making it "self-descriptive". In terms of
schema association, I think an XML file could only rightly be
called self-descriptive if there were a possibility that a human
being could look at it and say, "Ah, that looks like an instance of
schema X", without any explicit PI being there to tell them the
URL for the name or location of a schema.

> but namespace is not always sufficient to identify document type if you 
> are using languages that have several profiles in the same namespace. 
> Some languages provide its internal means for it (version attribute). 
> But other languages like XHTML doesn't provide such mechanism.

Then it seems to me that the community of people using a class of
document types that have several profiles in the same namespace
should try to come up with a way to make instances of the
different profiles truly self-descriptive so that people and
machines have some way of distinguishing them one another. A way
more robust than this proposed PI.

> I agree with you that in general locating-rules are better approach for 
> schema associations. But locating-rules can't solve some problems (like 
> interchanging documents, changing their locations) in a way that would 
> be sufficiently friendly from user's point of view. And by users I don't 
> mean you or me, but XML BFUs (no ofense meant) -- they simply will not 
> understand questions about selecting schemas, they will not be able to 
> edit locating rules file even with some sophisticated UI, ... But simple 
> PI used in cases where schema is not registered inside editor would help.

It is possible to imagine editors being able to use a system for
determining schema associations that would not restrict them to
whatever arbitrary set of schema associations are shipped with the
editor, or even to whatever schema assocations an individual user
may have manually added locally.

> Having something like locating rules standardized and portable between 
> different tools would be really great and I would be for such activity. 
> But it still doesn't solve all the ad-hoc problems that can be easily 
> solved by PI.

I guess I am still not convinced that it's impossible to create a
locating-rule system that can solve the problems you have
described. It may not solve the problem as easily as the PI would,
but it would solve it better.

  --Mike

Attachment: smime.p7s
Description: S/MIME cryptographic signature

Reply via email to