Re: Are *relative* URIs as namespace nemes considered harmful?

> If I have a low level layer that doesn't make a distinction between two
> namespace names even though the namespace names identify different
> resources, it will be difficult to build on top of this a higher level
> layer that uses the namespace name directly to access the resource,
> because it will be ambiguous what the resource associated with a
> namespace name is.

Any higher level that builds on top of the namespace REC with the
intention of locating a resource had better subset the URI references
that it allows. Your higher level isn't going to get much useful from a
relative URI that I accept, but it isn't going to get much useful from
mailto:davidc (or even mailto:davidc@nag.co.uk).


> In general, a higher level layer can easily identify things that lower
> level layers distinguish, but it's awkward for a higher level layer to
> distinguish things that a lower level layer identifies.

That I accept, but I wouldn't object if rdf or anything else built
up on top of namespaces just rejected my documents using mailto:
namespaces, and said I should use http or ftp or something.
I wouldn't object if it rejected relative URI either.

This is a small price to pay compared with the undocumentable chaos that
will ensue if we really settle for

<aaa xmlns="schema">.....

working with some stylesheet while it is on a web server, but then
failing to work with the same stylesheet if you save the thing to disk
and try to run it locally, because the names of all the elements
changed.


I think having higher level protocols that need globally unique
namespace names insisting that documents have globaly unique URI in
their namespace declarations is perfectly reasonable.

David

Received on Tuesday, 23 May 2000 09:30:15 UTC