Re: More comments on the mailserver URL scheme

Larry Masinter (masinter@parc.xerox.com)
Sun, 25 Jun 1995 14:35:09 PDT


To: peterd@bunyip.com
Cc: uri@bunyip.com
In-Reply-To: peterd@bunyip.com's message of Sun, 25 Jun 1995 10:53:34 -0700 <95Jun25.105337pdt.2761@golden.parc.xerox.com>
Subject: Re: More comments on the mailserver URL scheme
From: Larry Masinter <masinter@parc.xerox.com>
Message-Id: <95Jun25.143523pdt.2761@golden.parc.xerox.com>
Date: Sun, 25 Jun 1995 14:35:09 PDT

> I think it is an arguement for the client and the server
> to agree on how long a URL will be and well-known limits
> is the easiest way to do that. Negotiating a length on
> each transaction is going to cost, and thus we probably
> want a fixed limit. I just can't help but think how much
> trouble we went through to increase the file name size in
> UNIX a few years ago...

There are many mechanisms of dissemination: email (point to point or
one to many), netnews (flooding), information access (web browsing),
network searching (through digger or infoseek or yahoo), etc.

Only a few of these scenarios are amenable to client-server
negotiation. Even so, while negotiating over content-type might make
sense in HTTP, negotiating over 'maximum length of URLs' doesn't seem
to. Perhaps we should be explicit in the revision of the URL RFC to
make clear that 'URLs are unlimited in length: any URL-accepting
program should be prepared for arbitrarily long ones, because in some
rare occasions they are used to encode arbitrarily long data
sequences.'

I'm not arguing that we must have a length limit or that we shouldn't
have one, just that we should be clear in the standard what the rules
are. Does Digger have a URL length limit? Does your favorite browser?