Re: relative URIs for namespaces

I'm quite happy disallowing relative URLs.

Such errors should probably be in that "implementations
don't need to test for this" category, again avoiding the
need to interpret the URI.

- Dave


James Clark wrote:
> 
> David Brownell wrote:
> >
> > What would the semantics of a relative URI in an "xmlns...='...'"
> > declaration be?  That is, what's the base URI with respect to
> > which such a URI must be interpreted -- the document in which
> > it's found?  Or are relative URIs not allowed?
> 
> I would prefer not to allow relative URIs here.  I can't think why
> anyone would want to have a namespace URI relative to a document. It's a
> bit of a pain to implement (at least for some kinds of implementation).
> It requires the implementation to interpret the namespace URI rather
> than just treaingt it as an opaque string.  It also removes the
> ambiguity with xmlns="" caused by the empty string being a legal
> relative URI.  Also I suspect many of the URIs used with namespaces in
> the long term will be non-hierarchical and thus unuseable with relative
> URIs.  If we were going to allow relative URIs, I think we should
> provide a way to specify what the base is.
> 
> James

Received on Tuesday, 25 August 1998 22:44:52 UTC