Re: Globalizing URIs

Brian Behlendorf (brian@organic.com)
Tue, 15 Aug 1995 14:41:20 -0700 (PDT)


Date: Tue, 15 Aug 1995 14:41:20 -0700 (PDT)
From: Brian Behlendorf <brian@organic.com>
To: "Karen R. Sollins" <sollins@lcs.mit.edu>
Cc: conklin@info.cren.net, uri@bunyip.com
Subject: Re: Globalizing URIs
In-Reply-To: <199508151700.NAA07155@lysithea.lcs.mit.edu>
Message-Id: <Pine.SOL.3.91.950815135102.476A-100000@eat.organic.com>

On Tue, 15 Aug 1995, Karen R. Sollins wrote:
>    From: conklin@info.cren.net (Jim Conklin):
>      I respectfully disagree.  URL's worked and were accepted and used, and
>    people remember them (yes, REMEMBER them!) precisely BECAUSE people very
>    INTENTIONALLY attach semantic names to them.  This, of course, is the point
>    Roy made, just recouched a bit.
> 
> I'm afraid I disagree with you at least partially.  Yes, URLs were
> accepted and used.  But I don't consider them terribly successful.
> First, because they do often include semantics, we expect certain URLs
> to work.  But, we get too many surprises.  For example, www.mit.edu
> was grabbed by a student computing organization early on, so now MIT
> uses web.mit.edu, much to many people's confusion.  Because it is a
> single namespace and we are using a great deal of abbreviation to make
> it compact, there is contention and disagreement over allocation of
> names.  That's the first problem.

Any place with a common namespace that is expected to handle everyone
is going to have namespace collision problems as long as the goal 
of the namespace is to provide pneumonics.  If you remove the need 
for pneumonics, then you don't have the namespace collision, but you make 
the reference much less transcribable.  MIT could distribute the URL
http:///18.72.0.31/ instead of http://web.mit.edu/ if it really wanted 
and solve that problem without URN's, but obviously they're willing to 
live with the "drawback" of using web.mit.edu.

And in my opinion, compactness is a Good Thing.  Has MIT registered 
MassachusettsInstituteofTechnology.edu?

> The second problem, which I am finding more and more often, is a
> reconfirmation that URLs are the wrong thing to be embedding in other
> documents.  Because URLs contain "location" information, they become
> invalid.  Therefore, when I go to a page and try to follow a link, or
> memorize what I believe is the "right" URL for something, but the
> destination has moved, I'm stuck.  

This is only because the tools out there to provide proper redirection are
practically nonexistant, and the people authoring the space aren't given
sufficient reason to keep the URL's constant, and the *culture* of
redirection hasn't been fully implemented.  Most link breaks are deliberate
actions - and part of that deliberate action can be to have requests for that
document return either a verbose error message "this document is no longer
available at this address due to ....." or a 301 permanent redirect that
browsers can replace entries in bookmark files with, that meta-indices like
Yahoo can replace database entries with (or remove altogether), etc.

In addition to encouraging people to use HTTP redirects to fix broken links
(there are *lots* of ways to encourage this, too - having the browser send
mail to the author of the page where the link anchor appeared for example),
it seems like a lot of work could be done in the world of
dead-link-resolution.  If http://a.host.com/ doesn't work anymore because the
machine's name no longer exists, might it make sense to query host.com's DNS
to see if there are any other WWW servers anywhere?  (for which we need an
equivalent to MX in DNS)  Might we need a standard way to ask an HTTP 
server "do you have a document resembling *this*"?  

We need to ask ourselves *why* links are breaking.  I think we would find 
webmaster laziness, sloppiness, politics, or a general misunderstanding 
as to *why* persistant URL's are a good thing, to be the chief reasons.  I 
would also say creating a new namespace because the other is not being 
used properly is not a general solution.

To the extent that the WWW is an organic entity, death is a part of life. 

Phew, I've been wanting to say that since I read "URN's considered 
harmful".  

All this said, there *is* a role for universal names, and that is for 
common objects: IETF documents, government documents, Aesop fables, VRML 
teapots, etc.  Let's focus on that, rather than having a goal of "keeping 
links from breaking".

	Brian

--=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=--
brian@organic.com  brian@hyperreal.com  http://www.[hyperreal,organic].com/