Theory and practice of URIs and Namespace names

"In theory, theory and pratice are the same; in practice, they're not."

To be clear with my opinion, I am in favor of the 'literal
equality but forbid relative references'.

a)I think it's important to be careful about using "is" vs. using
"used as" when discussing the relationship between URIs and namespace
names.

The term "URI" is well defined, in theory and in pratice, as
a resource identifier. XMLs choice to use the URI syntax and
possibly some of the URI semantics for namespace names does not
mean that a namespace name "is" a URI, it means that you can
decide what you want to do, and how much of the URI semantics
you want to inherit for namespace names.

That's why we're having this design debate, and why we should
treat it as a design debate and not as some kind of finger
pointing about what the 'truth' is.

b) In writing standards, there are more degrees of freedom than
 there are in designing products. In particular, it would be
 possible to, for example, 'allow' the use of relative URI
 references, but to not recommend them, because of their
 intrinsic unreliability. There are many kinds of URIs that
 would make poor namespace names, and the specification does
 not (and should not) go to any particular length or effort
 to define their meaning.

 What does it mean to use a "data" URL as a namespace name?
 What does it mean to use a "mailto" URL as a namespace name?
 What does it mean to use a HTTP query URL as a namespace name?

c) There is no harm in declaring existing documents 'noncompliant'
with a new specification. Compliance is not an intrinsic property
of the document itself. It is a property of the object being tested
and the specification that it is being tested against. There is no harm
in declaring that all of those 'thousands' of documents that were
compliant with an old specification are not compliant with a new
specification.
You have not changed them and or the specification with which they
were compliant when you issue a new spec.

I think the new spec should recommend against using relative URIs,
because of the difficulties involved in resolving the interpretation,
computation of base. Standards should be practical as well as theoretically
sound. This is just practical.

Larry
--
http://larry.masinter.net

Received on Tuesday, 16 May 2000 19:28:19 UTC