Message-Id: <9305180637.AA19640@interval.interval.com>
Date: Mon, 17 May 1993 23:37:28 -0800
To: uri@bunyip.com
From: Terry Winograd <winograd@interval.com>
Subject: Re: Internet Draft on URNs
I have a question about naming Naming Authorities. The draft says:
>The Naming Authority identifier consists of two sub-components, a 'scheme
>identifier' and an 'individual identifier'. A scheme identifier is the
>name of a protocol or organization which can guarantee the uniqueness and
>resolvability of the individual identifier. The individual identifier is
>the identifier of an organization which has assigned the opaque string
>component to the resource and can resolve the URN to a set of URLs for the
>resource. The scheme identifier and the individual identifier are separated by
>a colon. For example, typical Naming Authorities might be
>
> IANA:42117
>
>or
>
> ISBN_Publisher_ID:0_201_12
>etc.
It isn't clear to me why this is a two level system instead of a more
generic hierarchical system as in domain names. The proposed version
implies that there is one worldwide global name authority that is supposed
to guarantee the uniqueness of all scheme names and for any scheme there
must be a single organization that deals with all individual authorities.
If you assume that there will only be a few schemes and authorities this
would be possible (though it has a potential bottleneck). But in the
widely distributed world, URNs will be used at all levels. For example the
report series for my local department will provide URNs with the department
as the naming authority and the report sequence number as the opaque
string. In a world with appropriate privacy protections, each person will
be assigned a unique URN by every distinct organization and agency they
deal with, etc.
This gets to a comment of Dave Brennan's
>Should it be suggested that it's a Bad Idea to assign more than one
>URN to a particular resource?
It is unrealistic to legislate this or expect it to hold. As mentioned
above, the notion of "naming authority" will be highly distributed, so a
document may have a URN from a local department, one from the NTIS, one
from ISBN, etc. (just like I now have a social security number, a drivers
license number and an employee number, from different authorities). There
will be social pressure for people to make use of common widespread naming
authorities and to give preference to the URNs of those authorities in
citations, but in general we have to assume that there can be more than one
URN for the "same" item. Of course a single authority must guarantee
uniqueness for its names.
Also, one further comment. The URN proposal says:
>...the identifier of an organization which has assigned the opaque string
>component to the resource and can resolve the URN to a set of URLs for the
>resource.
I would expect in the general case that it would not be the naming
authority that provides the indexing service that maps its URNs onto
publically accessible URLs. This would be the job of "catalog and archive
servers" which would provide various levels of guarantees that items would
be kept around for specified lengths of time and could be retrieved on the
basis of their URNs (and on the basis of other citation information).
Given that we are thinking in terms of decades (or even centuries), it is
unreasonable to make it a condition of giving a name that the naming
authority commit to an indefinite period of continued storage and
accessability of the item.
--t