RE: Personal view

Personally, I'm coming to really like neither "forbid" nor
"literal" but rather "absolutize with strict interpretation of
base" plus "discourage mixing relative and absolute forms for
same NS name".

This is effectively "literal" except when comparing relative
URI forms with different base contexts.

Define the base for a relative URI used as a NS name:

 a) a namespace declaration establishes the base for any subsequent
    namespace declaration attributes in any embedded tags.
 b) in lieu of any surrounding namespace declarations, the 'base' for
    a namespace declaration is the pseudo-scheme 'xmlns-base:/'.
    In no case is the base used for namespace absolutization to be
    taken from the containing document.

This is different than any current recommendations, but it gives
the desired effect in most cases and for most existing documents.
It means that the interpretation of namespace names in an
XML document does not depend on the URI of the document itself,
but only on the contents of the document. It allows relative forms
that are really only used within a single context and aren't intended
to escape out, and also allows for nested namespaces if those turn
out to be desirable.

>   5.2. Resolving Relative References to Absolute Form
> of the rfc2396 should be applied to URI expected to be used as namespace
> name (i.e. before namespace name creation, not when handling them). So
> that "http://www.example.org/./a" is still a valid namespace name,
> but be clearly flagged as "bad practice".

In general, URIs that would (likely) be equivalent as far as determining
which
resource was identified but not byte-for-byte equal should be "bad
practice", although inline "." and ".." are not the worst offenders.
To be specific, there is no guarantee that "http://www.example.org/a/./b"
is equivalent to "http://www.example.org/a/b" for any web server, so there's
no reason to assume that these two namespaces would be the same (unlike
http://WWW.EXAMPLE.ORG/a/b and http://www.example.org/a/b which ARE
defined to be equivalent as far as accessing the resources are concerned.)

Received on Saturday, 17 June 2000 20:54:56 UTC