Lists Home |
Date Index |
Been out of touch for a few days so late to this.
Yes. The implemented algorithm defines the semantic.
Essentially, http:// is overloaded. It
has a URL semantic. By going to URIness, we
ended up with a semantic of nameness which if taken
as a bare name, contradicts the original semantic.
Making it ambiguous changed nothing and added cost.
Keep this in mind:
o Dereferenceable means one can hand that string to the
http method and it will make the attempt.
o Dereferenced means it made the attempt and succeeded
and returned a representation or failed and returned an error.
It is ALWAYS dereferenceable. By design. By fact
of syntactical equivalence. By precedence of assigning
the semantic of resolving.
If we mean by dereferenceable "it can always be
dereferenced but what happens next isn't guaranteed"
we state the fact, but we have admitted that subsequent
designs have lowered the reliability of the WWW.
If we say "it is always dereferenceable and what happens
next depends on the author/owner/source", we haven't
changed anything except to say it is as reliable as
If we say, "sources are strongly advised
to put something there", we have gone as far as the
current architecture enables.
Among those choices, it seems better to put something there:
a document, if nothing else. Note: it seems better; it is a
good practice. Like etiquette, all may not follow
it, but those who do get along better by being clearer
about their intent.
Like it or not, MSIE is now the test. It is the
standard. I realize that inflames, but it is nonetheless
the working fact. I read RFPs for a living; they now
include clauses that state "compliant with the
Internet Explorer standard". One can moan that
these people don't understand standards, but in
fact, they understand perfectly in about the same
way as this has been understood and applied since
the mid-90s once Mosaic was presented.
The biggest colony wins. That is the ultimate result of
the *fielding* of the web.
The monkeys didn't quarrel with Darwin. The scientists did.
From: Joe English [mailto:email@example.com]
> Really? Test it.
Feh. If what you mean by "URIs are always dereferencable"
is "you can type them at MSIE and it will do something,"
then sure, that's a true statement. Not a very useful one,
But I think I finally see where Len is going with this.
There is an algorithm for resolving URLs; *that* is what
makes them useful. When RDF and XMLNS play Humpty Dumpty
and insist that a URI means what they intend it to mean,
nothing more and nothing less, contrary to established
practice and common understanding, it diminishes the value
Am I on the right track here?