(no subject)

>I argued it at some length some days ago, but I am not sure that
>I could find it easily!  I augued that you can never hope in a scalable
>system to know from a name that something is "different".

Uhm... I don't quite understand why. The aliasing scenario you pose only
exists if aliasing is designed into the system -- as it is for URIs, but
isn't for literal strings. Which is the point I was making; URIs have
characteristics that we don't want for this application, as well as ones we
might.

>So if you make a namespace http://us.ibm.com/~keshlam/X
>I can make a http://www.w3.org/2000/Y and declare it
>to be for all purposes equivalent to your X.

You can so declare it. But unless there's some mechanism for globally
asserting that -- which there isn't, in the namespace spec as written --
your declaration has no effect. Code that looks for X isn't going to
magically recognize Y as the same namespace, and shouldn't.

The problem is: URIs aren't defined to behave that way. URIs say that you
can only say that X equals X; you can't say the X does not equal Y, only
that you don't know. And "I didn't know" is a bad basis for most decisions.

Yes, we can impose an additional rule, for namespaces only, about
"definitive URIs"l. But I don't see that this is any less an affront to
URIs than the suggestion that relative references be forbidden in this
context, or even that the strings be taken as literals. It's still a matter
of taking the URI syntax an applying a custom interpretation.

>It has to be clear that an XSLT template will pick out
>only those elements whose namespaces have exactly the
>given URI.  I think that is what people expect, and it works.

Absolutely. It works today because XSLT is operating in Literal mode.When
we declare that these are URIs and folks start carrying over expectations
from other uses of URIs, I'm not sure that the results will still be
"expected". Remember, most folks' impressions of normal URI behavior are
set by their browsers, not by the spec... and even the fairly savvy crowd
on this list has to keep going back to the book to check.

     ...And the user exclaimed, with a snarl and a taunt:
     "It's just what I asked for, but not what I want!"


If we can eliminate relative syntax, then we can tell the users "namespace
URIs are compared simply as strings" and it's something they can wrap their
heads around independent of _why_ it's the right answer.

______________________________________
Joe Kesselman  / IBM Research

Received on Thursday, 15 June 2000 16:18:43 UTC