Re: A proposed solution

> > A URI space is typically defined with a set of
> > properties concerning uniqueness,
>
> It can't be reasonable to expect a parser for xml namespaces (which is
> only really a minor change on XML 1.0 non validating parser,
> tokenising element and attribute names differently, and a couple of
> extra constraints) to suddenly know the uniqueness properties of any
> URI scheme. By the above you seem to be proposing dropping the
> character-for-character comparison even for absolute URIs.
> I think that this is just way too big a change and will break
> all existing namespace systems and change the meaning of all
> documents using a namespace.

This is not what it says at all. The proposal says that in order for you
to handle relative URIs you need to know the context you are in. There
is nothing unreasonable or new in this. That has nothing to do with the
properties of the URI space that the URI may belong to. Within a
well-known context you can use relative URIs as unique identifiers
within that context without expanding them.

The proposal absolutely allows up to compare URIs
character-for-character - only requirement is that you know that you are
not crossing contexts when dealing with relative URIs.

> >  An application is also responsible for ensuring that
> > relative identifiers are not treated as unique identifiers across
> > contexts as ignorance of context can make distinct identifiers
appear
> > undifferentiated.
>
> this seems to imply that a given document is tied to some application
> or range of applications. As the rules about how namespace names are
> processed appears to be made application specific. Perhaps this fits
> with the Microsoft methodology of working with XML files, but it
> doesn't fit well with my normal model of documents having an
> independent life and being processed by any number of different
> applications at different places by different people.

Not at all - it merely states that you can't use relative URIs without
taking the context within which they are defined into account. Note that
this doesn't mean that you necessarily have to explicitly absolutize.

Henrik Frystyk Nielsen,
mailto:frystyk@microsoft.com

Received on Friday, 9 June 2000 11:30:47 UTC