To: connolly@hal.com
In-Reply-To: "Daniel W. Connolly"'s message of Thu, 21 Jul 1994 08:45:24 -0700 <9407211545.AA12129@ulua.hal.com>
Subject: Re: URL revision
From: Larry Masinter <masinter@parc.xerox.com>
Message-Id: <94Jul21.095136pdt.2760@golden.parc.xerox.com>
Date: Thu, 21 Jul 1994 09:51:26 PDT
(This note just addresses the # issue, and not the proposed alternate
prefix schemes URL!, URL=, URL(xx.xx)= ...)
I think originally "#" was 'universally reserved'; there was a lot of
objection to that (I think, on first principle); it was intolerable to
make it reserved-according-to-scheme because of the desire of being
able to, for example, resolve
unknownscheme:blah.blah.blah#location
by handing off "unknownscheme:blah.blah.blah" to a proxy and then
dealing with '#location' when it came back.
Now that I think it through, though, the 'fragment identifier' has to
be part of this draft and part of the URL, since you want the fragment
identifier to be inside the delimiting characters:
<URL:unknownscheme:blah.blah.blah#fragmentid>
The only resolution to this conundrum that I can think of is to put
fragment identifiers back into the general URL syntax, and make #
universally reserved.
In general, the attempts to chop off some of the pieces of WWW to make
a URL syntax that was not specifically tied to WWW may have been too
aggressive. Other applications other than Mosaic and web clients might
very well have uses for fragment identifiers; removing them cannot be
done in a way that doesn't somehow disallow them (if # isn't reserved,
then URLs will appear that use # as an ordinary character; if it *is*
reserved, then we must either reserve it for some mysterious purpose
or else proceed to explain fragment identifiers.)