> How would people feel about adding text in a future version of the
> standard to say that
>
> (1) if a BASE is present, relative references are resolved
> relative to the specified BASE
[ which is what the current standard says ]
> (2) a stand-alone fragment is not a relative reference, but is
> a reference internal to the current copy of the document
I would be much happier if the next version of HTML
allowed an ID attribute on every element, and added an IDREF
attribute to <A>:
<!ATTLIST A
...
REFID IDREF #IMPLIED
...
>
The HREF attribute should always be interpreted as a
(possibly relative) URL, even when it contains nothing
but a fragment identifier. Links within a document should
be specified with <A refid=foo> instead of <A href="#FOO">.
That removes all ambiguity.
> (3) if the complete URI formed to resolve a relative reference
> is the current document or a location within the current
> document (where the "current document" is defined to be
> what is specified by BASE or, if BASE is not present, the URI
> under which the document was retrieved, then the browser
> shall resolve it within the current copy of the document
I don't like this. What if the current document has expired?
I would like to see some clarification on the semantics
of <BASE>, though, namely:
If a base address is specified for a document, then
that document (must|need not) be retrievable
by resolving that address.
Either semantic would be acceptable, though I would prefer
the latter, i.e., the base address is *only* used to
resolve relative links. <META>, <LINK>, or an external
mechanism could be used to specify a "preferred address"
for the document if that is necessary.
--Joe English
joe@art.com