- From: Larry Masinter <LMM@acm.org>
- Date: Tue, 29 Apr 2003 23:58:16 -0700
- To: <www-tag@w3.org>
I think even after IRIs are approved, there is a problem with the multiple definitions of equivalence which can only be resolved by encouraging systems that use strict (strcmp) equivalence to never use more than one of various forms that might be equivalent using a looser definition. Of * http://www.example.org/~wilbur * http://www.example.org/%7ewilbur * http://www.example.org/%7Ewilbur Only the first should be used as a namespace name. Namespace processors may assume that they will never any of the others. ============================ Of * http://www.example.org/ros%e9 * http://www.example.org/ros%c3%a9 * http://www.example.org/ros%c3%A9 * http://www.example.org/ros%C3%a9 * http://www.example.org/ros%C3%A9 None of these should be used as a namespace name. Namespace processors may assume they will never encounter these. http://www.example.org/rosé is preferable. ================================== Among * http://www.example.org/wine * http://www.Example.org/wine * http://www.example.org/Wine the second should never be used as a namespace name. Using the first and third as distinct namespace names isn't a great idea, but it isn't as bad to disallow them. Namespace processors may assume they will never encounter the second example. ======================= In general: make the hard cases moot. I know that it was suggested that a requirement of namespaces 1.1 that they be a superset of namespaces 1.0, but avoid the hobgoblin of consistency, and "do the right thing". No one will mind.
Received on Wednesday, 30 April 2003 02:58:30 UTC