W3C home > Mailing lists > Public > xml-names-editor@w3.org > December 2002

Re: Namespaces 1.1 Last Call -- I18N WG comments

From: Richard Tobin <richard@cogsci.ed.ac.uk>
Date: Fri, 6 Dec 2002 13:27:36 GMT
Message-Id: <200212061327.NAA17633@rhymer.cogsci.ed.ac.uk>
To: Martin Duerst <duerst@w3.org>, Richard Tobin <richard@cogsci.ed.ac.uk>, w3c-i18n-ig@w3.org
Cc: xml-names-editor@w3.org

> In particular, the position of [ and ] to indicate the extent
> of the definitions are still wrong.

I'll look for a way to improve this before we get to Rec, but I don't
think it's serious enough to worry about now.

> Also, the syntax details of the IRI definition are already
> out of sync again.

This is inevitable, but hopefully things will be more stable by the
time we get to Rec, even if the RFC hasn't been published.  It -02
still the one we should reference?  I think the definition we give
should be consistent with the one we can reference, though we could
add a statement that it is expected that the private use ranges will
be added.

> I therefore really wonder what the purpose of your syntax definition is.
> For the namespace spec itself, only comparison is what matters, or not?
> Is a something like <p:p xmlns:p='@@@:a#a#a'> namespace-well-formed or
> not? To what extent is a namespace-aware processor supposed to check
> the details of the URI/IRI syntax?

We have said that

(a) namespace names must be IRIs for well-formedness
(b) namespace processors don't have to check.

> To what extent do actual implementations
> check currently? If a check isn't needed, or isn't done by current
> implementations, what's the purpose of the syntax definition in the
> spec?

We believe that most existing namespace processors don't check.  But
we also know that some *applications* expect to be able to parse
namespace names and store them in URI structures, for example when
probing for a schema at that address.  (In fact, we have a formal
objection to the change to IRIs from someone whose application relies
on them being URIs.)

I think its reasonable that you should not get an error (other than a
retrieval error of course) from a well-formed document when you
attempt to use its namespace name as an IRI, so I believe that (a)
above is right.  We decided on (b) as a result of your comments
(on the TAG list if I remember correctly) and in keeping with existing
practice.  And if we say they must be IRIs, I think we need to give
a provisional definition in the absence of an RFC.

We can of course revisit this in the light of IRI progress at the PR
stage.

-- Richard
Received on Friday, 6 December 2002 08:27:39 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 7 December 2009 10:59:43 GMT