From: mitra@pandora.sf.ca.us (Mitra)
Subject: Re: Fragments are Client side [Was: URN Requirements ]
Date: Wed, 18 May 1994 18:53:45 GMT
Message-Id: <Cq0HtM.B6z@pandora.sf.ca.us>
Daniel W. Connolly (connolly@hal.com) wrote:
: The #fragment part of a WWW address is for use by the client after
: it has retrieved the info described by the rest of the address.
: Hence I expect that these Xanadu spans are more like WWW search
: specifications, in that they select certain information to bring
: over the wire.
: Somewhere in the WWW doc, it explains that a WWW "document" is
: defined as the smallest unit of information that can be retrieved.
: So if a server is willing to give out, for example, certain paragraphs
: of a given document, it would be incorrect to write:
: HREF="xanadu://host/dir/file#para1-5"
: Rather, you would write:
: HREF="xanadu://host/dir/file?para1-5"
: or perhaps just:
: HREF="xanadu://host/dir/file/para1-5"
Right - which is why these are URL's - i.e. instructions for retrieving
a document, rather than a reference to the document.
In a URN context, I would see a reference is to a fragment of a
document. It doesn't tell you whether the client or the server does the
fragmentation. That might depend on the capabilities of the protocol
being used, the bandwidth of the link, whether the client wants to cache
the whole document etc.
It would be stupid (IMHO :-) to assign a URN to every possible fragment,
since these would be URN's assigned by someone who is NOT the publisher
of the document.
I'm also not proposing fragments are incorporated inside URN's or URL's,
the proper place is within a URC - i.e. fragmentation information might
be something passed with a URN or URL to a server.
- Mitra