W3C home > Mailing lists > Public > www-tag@w3.org > February 2005

Re: Significant W3C Confusion over Namespace Meaning and Policy

From: Tim Bray <tbray@textuality.com>
Date: Mon, 14 Feb 2005 00:31:53 -0800
To: www-tag@w3.org, "w3. org" <www-tag@w3.org>
Message-id: <876ba0239466984ed67859efa106b7fb@textuality.com>

Wow, I take a week off in Hawai'i, what a welcome back.

A few things seem obvious:

- The C14n rule about propagating any and all xml:-namespace attributes 
down the element tree is awfully stupid and broken, and since the 
process gave us all lots of chances to review the C14n REC, we all 
share in the stupidity and broken-ness.
- Interestingly, I suspect quite a few c14n implementations may have 
missed that one.  I know that my genx implementation misses it, and 
furthermore, the application of Elliotte Rusty Harold's excellent test 
suite failed to detect the bug (or maybe I just missed it, but I don't 
think so).  Anyhow, just maybe it's not too late to go back and fix XML 
- The assertions about the immutability of namespaces, and what they 
necessarily identify (container vs collection) are seriously 
inoperative, since there are real working deployed examples out there 
that fail to abide by any of these assertions, and the software seems 
to deal with it, so the theologians here should too.
- The assertions about necessary linkage between namespaces and 
versioning policy are similarly inoperative.  There is no general 
policy in place and no chance of getting one.
- The "xmlid" solution would dodge the C14n broken-ness, but the pain 
of explaining this to successive generations of people down the road 
seems almost too great to bear.
- Just charging ahead with "xml:id" is perhaps bearable too; note that 
in the actual instance, the behavior of a canonicalizer would result in 
the first appearance of xml:id in the instance being in the original 
location... so software can probably work around this.

At the end of the day I'm basically not worried.  Xml:id is 
demonstrably a solution to a non-problem anyhow.  There is currently no 
basis in any normative specification for software receiving data served 
as */xml to expect ID attributes to be defined, and we've sailed along 
for five years or so now without this resulting in much breakage.

So at least we should try, as physicians do, first to do no harm. -Tim
Received on Monday, 14 February 2005 08:31:45 UTC

This archive was generated by hypermail 2.4.0 : Friday, 17 January 2020 22:56:07 UTC