Message-Id: <199409031829.LAA26286@nic.cerf.net>
Date: Sat, 03 Sep 1994 11:19:27 -0700
To: uri@bunyip.com
From: miked@CERF.NET (Michael A. Dolan)
Subject: Re: [John.Ockerbloom@gs1.sp.cs.cmu.edu: Re: FTP handling suggestion]
At 04:21 PM 9/2/94 PDT, Larry Masinter wrote:
>I missed that this hadn't gone out to the whole list.
>
>My take on this situation is that 'Mosaic for X version 2.4' doesn't
>conform to the URL specification we've written, and that libwww
>probably will have to be modified to conform. I'm less happy with a
>description of 'ftp' URLs that says:
>
> try the following thing, and if it doesn't work, try something else
>
>but I'm willing to listen of others think I'm wrong.
>================================================================
>Date: Tue, 30 Aug 1994 07:13:11 -0700
>From: John.Ockerbloom@gs1.sp.cs.cmu.edu
>To: masinter@parc.xerox.com
>Subject: Re: FTP handling suggestion
>Cc: spok@cs.cmu.edu
>
>> Could you be more explicit about what you tried and what didn't work?
>
>Sure. In Mosaic for X version 2.4 this works:
>
> ftp://rtfm.mit.edu/pub/usenet-by-hierarchy
>
>But none of these work:
>
> ftp://rtfm.mit.edu/pub%2Fusenet-by-hierarchy
> ftp://rtfm.mit.edu/%2Fpub%2Fusenet-by-hierarchy
> ftp://rtfm.mit.edu/%2Fpub%2Fusenet-by-hierarchy%2F
>
>It appears that this browser (a very common browser, and the main
>one used here) doesn't unescape the characters.
>
>On the other hand, our FTP system is set up so that users must
>cd directly to the directory they're interested in. In other
>words, if the target directory is /afs/cs.cmu.edu/user/spok/www,
>a CWD directly there will work, but a piecewise CWD starting
>with "afs", and so on, will fail.
>
>We also require the leading-slash, as do many other sites.
>
>So we're stuck with the situation that no FTP URL here (or at other
>sites that require a leading slash) will work with
>both the current browsers and the new browsers to come out of this
>RFC, unless the new browsers use the "fallback" method I suggested.
>
>Hope this helps explain things a little further.
>
>John Ockerbloom
There are two issues here:
1. Specifying existing implementations versus what we want it
to be in the future; and
2. URL syntax versus implementation algorithm.
I personally would like to see a
spec of existing practice (aka libwww) before moving forward, but I
apparently am alone in these feelings. At the risk of repeating
myself, I think we are headed for trouble by *not* trying to capture
what's already been done and in use today.
The problem above is only an algorithm issue. If the URL syntax permits
for a unique description of an object, then this example is just one of
several interesting variations on the algorithm. How one gets to retrieve the
object may, in practice, contain many variations on what was written in
the proposed RFC. Another good example is how to get through authorization.
What if "USER anonymous" doesn't work ? (Some sites require "ftp".) What
if certain PASS syntax is required at a site, etc, etc. We can continue
to enhance the algorithm specification (it needs a *lot* more work to
be used for implementation), or we can drop algorithms for now and get the
syntax done.
The FTP syntax is sound, but the specific issue raised by Mr. Ockerbloom is
only the tip of the iceburg on implementation and algorithms.
Mike
-----------------------------------------------
Michael A. Dolan <mailto:miked@cerf.net>
TerraByte Technology (619) 445-9070, FAX -8864