Message-Id: <9306032228.AA00711@wilma.cs.utk.edu>
From: Keith Moore <moore@cs.utk.edu>
To: Larry Masinter <masinter@parc.xerox.com>
Subject: Re: URLs and types: concentrate on FTP
In-Reply-To: Your message of "Thu, 03 Jun 1993 13:42:55 PDT."
<93Jun3.134302pdt.2741@golden.parc.xerox.com>
Date: Thu, 03 Jun 1993 18:28:11 -0400
To: uri@bunyip.com
Subject: URLs and types: concentrate on FTP
Date: Thu, 3 Jun 1993 13:42:55 PDT
> The discussion seems to have been primarily philosophical, whether
> 'URLs' should or shouldn't contain meta-information.
>
> Let me back off from the philosphy, and make a more concrete
> suggestion:
>
> FTP URLs should have a way to specify type.
I don't buy it, because it's tantamount to saying that URLs are file names.
Sometimes they are, sometimes they are not.
As I see it, URLs exist to allow files on various kinds of file servers to
be referenced via a common notation. That's about it. So the primary
job of a URL is to provide enough information to tell you how (along
with the URL specification) how to get that file.
There are often several locations for a particular file and several ways to
get that file. URNs exist to establish a single, stable, name which can be
associated with multiple locations. There then needs to be a service which,
when given a URN, returns a list of URLs that you can use to get the file.
Type information belongs maybe in the URN, but better yet in the citation
information, which could also include pointers to other instances of that
object with different types (text vs. postscript, for example).
Search engines should return a list of 'partial' citations that contain at a
minimum, the URN and the type of an object, and probably a few other things
as well.
> Right now, we have to rely on the heuristics of file extensions.
> Unlike HTTP or Gopher+ URLs, there's not as much chance of extending
> FTP semantics to allow some out of bound method of finding.
Everyone agrees that the present situation is not ideal. But what you are
proposing is to legitimize these heuristics by requiring servers to conform
to them. This won't work because you are insisting, in effect, that
everyone adopt a particular file naming convention for their ftp servers.
In effect, we have 'optional' type information right now -- if the server
uses this file naming convention, the type information is there. If the
URL doesn't have a recognizable suffix, you can't tell the type of the
object.
Keith