Re: Use of ";" in relative URLs: procedural issue?

Karen R. Sollins (sollins@lcs.mit.edu)
Wed, 5 Feb 1997 12:34:22 -0500


Date: Wed, 5 Feb 1997 12:34:22 -0500
Message-Id: <199702051734.MAA13509@lysithea.lcs.mit.edu>
From: "Karen R. Sollins" <sollins@lcs.mit.edu>
To: masinter@parc.xerox.com
Cc: uri@bunyip.com
In-Reply-To: <97Feb3.191952pdt."241"@palimpsest.parc.xerox.com> (message from
Subject: Re: Use of ";" in relative URLs: procedural issue?

   CC: uri@bunyip.com
   Sender: Larry Masinter <masinter@parc.xerox.com>
   From: Larry Masinter <masinter@parc.xerox.com>
   Date: Mon, 3 Feb 1997 18:19:52 PST

   x.500 doesn't have the equivalent of "relative URLs", does it?

I think this is a separate issue from unordered vs. ordered
attributes.  X.500 only supports them ordered, although I believe it
permits wildcards or queries for elements in a path, leading to
searches.  In other words, if a name is {country=us, state=ma,
org=mit, dept=?, person=sollins) it would search all the departments
for "person=sollins".  Even this level of flexibility may not scale or
generalize well (i.e. think about what would happen if there were
several "?" in this list), and its been long enough that I'd have to
go back and check on even this level of detail.  But, originally,
before there was x.500, but only proposals, the leading candidate was
for unordered attributes of this sort.  It simply doesn't scale at
all, unless you know some algorithm that I don't.

But, since we are talking about URLs perhaps we should be talking
about X.400, not X.500 anyway, and I don't know the history there.  It
also does NOT support unordered attributes, and the syntax is also
ordered attributes as with X.500.  (But, it is worth noting that the
syntaxes are NOT identical; they are specified separately.)  Now, I am
not one to suggest that CCITT and ISO should be our role models, but
we should learn from their experiences where we can.

If you are thinking of using unordered attributes somehow only limited
to relative URLs, since their scope can also be very large (I don't
see any particular limitation on them), scaling must be an issue for
them as well.  But, as I said, perhaps I missed something here and
there are new ways to deal with the scaling problem that have just
passed me by.

			Karen