Re: [urn] Registration request for urn:eris: namespace

[[e-mail]]

My problem is that I have a continuing unease about the proposed
namespace.  Probably the best summary I can make is that the proposed
identifiers don't qualify as *names*.

Due to the large amount of entropy in the NSSs, an identifier that was
created for one unit of information can never be mistakenly used to
identify another unit of information even in another system.  But once
you are outside of a particular Eris system they are not only unusable,
but an individual identifier carries no discernible information.  About
all you can tell is that the local Eris system does not contain any
resource with that identifier, either because the identifier is from
another Eris system or because the local resource it identifies was
deleted.  This lack of any global significance strikes me as a
fundamental contradiction with the concept of a global namespace.

A concrete consequence of this is that you'd never expect to see a
urn:eris in a location where any other type of URI might appear, and
vice-versa.  Their only use is as input to processes which can only
handle Eris identifiers because they only have meaning within contexts
that are Eris universes.

One solution to this is to add some sort of "Eris system" identification
field to the URN.  E.g., instead of this (which is my interpretation of
the example in 2.7.1)

urn:eris:H77AGSYKAVTQPUHODJTQA7WZPTWGTTKLRB2GLMF5H53NEKFJ3FUQ

use

urn:eris:eris.example.com:H77AGSYKAVTQPUHODJTQA7WZPTWGTTKLRB2GLMF5H53NEKFJ3FUQ

where eris.example.com is a DNS name that somehow identifies the Eris
universe involved.  That approach allows straightforward differentiation
of references to the local Eris universe vs. references to other Eris
universes.  If some sort of protocol is assumed, it even becomes a URL.

Dale