RE: A proposed solution

At 12:03 PM 6/15/00 -0700, Larry Masinter wrote:
>Larry:
>> >I think the group should reconsider compatibility with RFC
>> 2557's algorithm for comparison.
>
>Simon:
>
>> I think that if we're going to be referencing RFC 2557,
>
>I didn't suggest 'referencing RFC 2557', I suggested 'compatibility with
>RFC 2557'.

Okay, 'borrowing from RFC 2557'.

>> It seems clear at this point that RFC 2396 by itself doesn't provide a
>> clean solution to the fairly basic needs of namespaces.
>
>It wasn't intended to.

That's very clear.  While I think URIs do very well for a large number of
resource identification tasks, they aren't a good fit with XML namespaces.
That's not RFC 2396's fault - it's just that URIs aren't the right answer
to every question.

>> The lack of a
>> Uniform Comparison Algorithm is deeply unfortunate.
>
>I learned in Math 101 that an 'equivalence relation' was any relation
>that was reflexive (a ~ a), symmetric (if a ~ b then b ~ a) and transitive
> (if (a ~ b) and (b ~ c) then (a ~ c)) but that a set might have any number
>of equivalence relations.  The set of all strings that are consistent with
>the URI specification can have many different equivalence relations; we're
>just trying to decide which one to use for namespace name equivalence.

Unfortunately, we can't even get past choosing which string to compare,
never mind the possible issues with 'true equivalence' in any particular
URI scheme.

Even the best understood part of the equation - comparing two absolute URIs
- doesn't square with many users' understandings of URLs, especially
regarding domain names, that case doesn't matter.  Explaining that simple
bit over and over doesn't suggest a clean fit to a naming process.

>The problem is not that there is no 'Uniform Comparison Algorithm', it's
>that each application of URIs needs to decide, based on its needs, which
>one to use.

The fact that URI processing involves considerably more than simple string
comparison has opened up all of the possibilities which we lucky
navel-gazers have wasted a month upon.  While a 'Uniform Comparison
Algorithm' might not have suited the varied needs of the URI community,
even providing a reasonably simple and complete default option would have
helped the XML community significantly.

Again, a needs mismatch.  There are enough of these to convince me - over
and over again - that the choice of URIs for namespace identifiers was
misguided at best.

>> Tracking multiple RFCs to describe how supposedly simple and (hopefully)
>> frequent comparisons should be made sounds like towing a semi-trailer with
>> a Saturn, once again.  (Through quicksand, lately.)
>
>The algorithm itself is simple; the fact that the definitions and comparison
>algorithm is not compactly written is fixable. Are you arguing that the
>algorithm is too complex, or just that you don't like the way it is
>described?

By fixable, are you proposing reopening RFC 2396 to pour glue over the gaps
the XML community is finding in its application of URIs?

Or do you mean that it's up to the XML community to reopen Namespaces in
XML and inflict the costs of any such changes to 'definitions and
comparison algorithm' upon itself, with no expectation that such changes
will conform to the expectations of the URI community?

I'm arguing that the algorithms are incomplete, vaguely described, and
inappropriate to what should have been a fairly simple task.

Simon St.Laurent
XML Elements of Style / XML: A Primer, 2nd Ed.
http://www.simonstl.com - XML essays and books

Received on Thursday, 15 June 2000 15:23:46 UTC