Message-Id: <9402250247.AA13289@expresso.bunyip.com>
From: Peter Deutsch <peterd@bunyip.com>
Date: Thu, 24 Feb 1994 21:47:41 -0500
In-Reply-To: Larry Masinter's message as of Feb 24, 15:57
To: Larry Masinter <masinter@parc.xerox.com>, mitra@pandora.sf.ca.us
Subject: Re: URN to URC scenario
Okay, another one and _then_ I go to bed... :-)
[ Larry wrote: ]
> My main concern about any plan to use DNS for doing URN to URC
> resolution is to make sure that URNs can remain valid and resolvable
> even after the `publisher' ceases to exist, or even if a name gets
> reassigned. That is, a document written and published by someone at
> `foo.com' should have a URN that can be resolvable even after
> `foo.com' ceases to exist in any other form, or even if `foo.com' gets
> reassigned to some other organization.
But be careful - do you consider "foo.com" the publisher
of this document, or merely the physical (or is it
virtual?) location from which the publisher did his thing?
It seems to me that it's merely the later.
If we define the publisher as a separate entity from the
location, one whose properties are independent of that
location, I think we're on much firmer ground. I certainly
don't want to tie a publishing system to DNS hostnames in
the long term for this reason. I'd rather have a specific
set of Internet naming authorities (a la ISBN) to which
each of us as potential publishers can apply for
membership, and each of which in turn holds its own right
to issue names in a namespace. Each such naming authority
would operate (or have someone operate for them) servers
if they want to be able to answer questions about location
of specific publishers and/or published entities, but this
is actually not a requirement for the generation of URNs,
only for the dereferencing step.
Now, if your naming authority does go away (and it will
happen sometimes, although hopefully less frequently than
loosing individual documents and publishers) you might not
be able to find anything under that naming authority on
the net, but at least the distributed nature of the
authority mechanism ensures that the URN stays valid (in
the sense that it wont start getting reused and referring
to something else). That's probably about the best we can
hope for in this environment and seems pretty good to me
at this point.
I don't think it realistic on the Internet to require that
a user could always find a server willing to answer
questions about particular URNs if that publisher or
naming authority goes away. It _should_ be possible to
require that a URN not suddenly start referring to someone
else's document if that does happen.
What does this mean in practice? It means that we need
something that has distributed, scalable authorities for
allocating parts of a namespace. The namespace authorities
do need to have a longer life than the individual
publishers but I don't think "eternity" is the only end
date most people will be able to live with. And we need to
_not_ tie naming too closely to systems which clearly do
not have the desired properties where we already can see
that it wont work.
> Mitra's original proposal (where the URN was of the form xxx/yyy which
> got turned into yyy.xxx.urn for the DNS lookup) had this property, at
> least if the `urn' top level domain didn't allow for the reassignment
> of names.
>
> If you use existing domains for registering URNs, the problem is that
> they DO, at least in the timeframe of the lifetime of URNs.
If we separate out architecture from implementation,
there's the question of what these things should be look
like and the question of how we set up and adminster
specific naming authorities. If the fact that the current
naming authority for DNS has practices that we'd like
changed (such as reissuing a name that is allowed to lapse
- do they _really_ do that?), then let's look at the cost
and feasibility of changing these practices. On the other
hand, that does seem very much an implementation issue and
thus separate from the discussion of what the URN format
should look like and what properties it should have and so
on.
- peterd
--
---------------------------------------------------------------------------
"The future belongs to neither the conduit or content players, but
those who control the filtering, searching and sense-making tools
we will rely on to navigate through the expanses of cyberspace."
- Paul Saffo, (_Wired_: March,1994)
---------------------------------------------------------------------------