RE: Grinding to a halt on Issue 27.

> I won't copy Noah's post -- suffice it to say he's entirely correct as
> far as I can speak: the state of the XML Schema REC, the intentions of
> the W3C XML Schema WG and all the implementations I'm aware of
> respectively require, expect and use, as Misha so helpfully put it,
> strcmp to compare namespace names.  There is no license anywhere in
> the REC as far as I know to do otherwise.

Of course.  Nobody is arguing otherwise, as far as I can tell.  

However, the fact that it is clearly specified does not in any way make
the dichotomy appear less capricious and arbitrary to the average end
user.  I am simply pointing out that this dichotomy leads to confused
user expectations and implementation bugs.  

If the world were as cleanly separated as the specs, and the two
use-cases for URI comparison never occurred together in the same
scenarios, there wouldn't be a problem.  But as it stands today, our
users are faced with what appears to them to be inconsistent behavior
(especially in the non-trivial scenarios which I documented earlier in
this thread), and we end up having to train them like lawyers to be able
to understand why the behavior is "actually consistent".

*We* understand the difference between "URI as an opaque string vs. URI
as an identifier".  But the vast majority of XML users will never grok
this, and they will resent any attempts by us to *force* them to
confront such legalistic gibberish.  Of course it's not "gibberish" to
people like us who actually care about the minutiae, but most people use
XML as a means to some other end and don't want to be forced to split
hairs to get their tool to work.  And the saddest part is that there is
no good reason for this dichotomy to even exist.  As far as I can tell,
it's just an accident of history and of convenience.  I can't even in
good faith explain to users "sure we compare URIs in two incompatible
ways, but we did it for your own good."  Instead I have to tell them
"sure we compare URIs in two incompatible ways, but at least the two
different ways are consistently applied according to the specs."

Received on Thursday, 1 May 2003 14:14:05 UTC