That's the only way it makes sense to me, but I don't see that spelled
out in the spec, and even RFC 1808 appears to conflict itself. It says
that fragment identifiers are not part of the URL, but then includes an
example in section 5.1 where a stand-along fragment identifier is merged
with the BASE URL (although it's not clear if this is the current
document's default URL or the BASE URL present in the header -- it
shouldn't matter, except that we are indeed talking about treating
stand-alone fragment identifiers differently from partial (relative) URLs).
Since the two biggest browsers handle this situation differently right
now, I submit that we should produce language defining a sensible
behaviour, since there seems to be no common behaviour to document. But
does this wait for 3.0, or could it go into 2.1, or am I missing wording
somewhere in this morass of RFCs, and SGML and HTML specs that defines
behaviour here?