Re: Datatype interface

[Kohsuke KAWAGUCHI <kohsuke.kawaguchi@eng.sun.com>]

It didn't reach the mailing list, and it didn't have the right zip file
attached. So I'm resending it to the list.
------------


> Datatype.createStreamingValidator should not be allowed to throw an
> UnsupportedOperationException.  A library that does not want to support
> streaming validation efficiently can simply use your
> helpers.StreamingValidatorImpl.  It is better for the library to do this
> rather than the client, because it is simpler and easier for the library to
> do this, whereas if the client does it then they will (unless they are very
> careful) end up throwing lots of exceptions, which can be a big performance
> killer.

Done. You are right.


> > Can I assume that the design is stabilized?
> 
> Yes.  We might need to add a method to ValidationContext to return the base
> URIs, depending on what the semantics of anyURI turn out to be.

Then the next problem would be the license. Is it going to be covered by
OASIS IPR, or by the other license (e.g., MIT)?

Also, do you mind if I list your name as the primary author of the
interface set?



As for "anyURI" type,

In my understanding, "anyURI" type of W3C XML Schema Part 2 does not
have well-defined value space. So I think all the datatype library can
do is to simply pass the lexical value to the application.
I think it is harmful for datatype libraries to knead the lexical values
by doing things like:

- unescaping %HH encodings, or
- combine "baseURI" to obtain the fully-qualified URI



regards,
--
Kohsuke KAWAGUCHI                          +1 650 786 0721
Sun Microsystems                   kohsuke.kawaguchi@sun.com

relaxng.zip