Re: Swamped (Was:Re: Call the question!)

>OPTION 4: TREAT NAMESPACE NAMES AS LITERALS FOR TESTING ATTRIBUTE
UNIQUENESS,
>BUT ABSOLUTIZE THEM FOR ALL OTHER  PURPOSES.

I don't think that advances us toward any of the conflicting goals. The
folks who insist that names must be absolutized won't like the first part,
the folks who reply that compare-after-absolutize makes relative names
impossible for applications to reliably recognize won't like the second
part, and I think both sides will consider this one inconsistant.

>That option is in fact consistent with both the namespace spec

Only because the namespace spec is sloppily written. (Which, I admit, is
why we're debating this topic.) It isn't consistant with the intent of the
spec, or with the intended uses.

>Is the idea of replacing URI references as namespace names by some other
form
>of unique identifier off the table?

I think the folks who are attempting to directly dereference relative
namespace names -- the same ones who can't deal with the FORBID option --
would object that this too retroactively invalidates a large body of
existing documents. And of course it breaks all the existing practice,
including existing W3C specs such as XSLT.

I like the idea of stepping out of this hole completely, but I think it's a
year too late.

______________________________________
Joe Kesselman  / IBM Research

Received on Tuesday, 23 May 2000 10:57:10 UTC