From: mitra@path.net (Mitra)
Subject: Re: URNs and Meta-Information
Message-Id: <CF256y.pv@pandora.sf.ca.us>
Date: Sun, 17 Oct 1993 20:06:33 GMT
Rob,
I hope that URN's do define a reasonable access method. If they dont then I
have already proposed a namespace be chosen (maybe called "inet") that
would be defiend so that the publisher was a FQDN. The conditions for
self-registering a publisher id in this namespace would be to put up a
URN->URL/URC resolution device on a well-known port, conforming to a
well-known protocol. (I'm open to any definition of a well-known protocol,
and currently favour extending HTTP+ or Prospero to do the job).
That way, we would have both a way for anyone to become a publisher, and a
way for any client to retrieve documents in this namespace.
Note that if a publisher didnt want to run their own URN->UR* mapping,
then they would simply contract with someone who did so, and either use
their device or set up a CNAME that pointed to them.
Taking this further - I could see URN's of the form.
URN:inet::urn.xyz.com:::12345
xyz.com would set their local dns information so that urn.xyz.com resolves
to a list of URN->URC mappers that register their documents.
So retrieving a document would be a standard DNS query (forwarded from
urn.xyz.com to e.g. urn.internet.com and urn.path.net) followed by a lookup
to urn.internet.com port 12345 returning a list of URL's
A regular retrieval of the URL.
I'm suggesting a seperate namespace, because I'm not sure how to
integrate PeterD's needs for a more generic (and to my mind indeterminate)
lookup through indexing services. I'm certainly open to figuring out how
to do this, and then collapsing it to a simpler.
URN:urn.xyz.com::123456
I also see as an alternative, using the DNS itself to locate URN spaces,
so that URN:isbn::123456 would be looked up in the DNS as "isbn.urn" to
return a mapping to urn.bunyip.com and urn.publishersclearinghouse.com
where 123456 could be looked up. However for this to work we would need
to
a) all agree that all publisher id's could be resolved to URN->URL
mapping servers. and
b) all agree on a protocol for doing the lookup and
c) implement a seperate DNS hierarchy.
In my original proposal we dont need agreement other than from a critical
mass of self-publishers who will use publisher id's in the "inet" domain
space.
- Mitra