[apps-discuss] Device URNs

[Jari Arkko <[e-mail]>]

Hi,

We'd like to get some feedback on a draft that we wrote a while ago, on identifying devices. The basic idea is a definition of a new URN type, e.g., "urn:dev:mac:0024befffe804ff1" that can use hardware identifiers such as MAC addresses or 1-wire identifiers. The authors have been using identifiers of this type as identifiers in data streams coming out of very simple devices that lack any configuration or even the capability to be configured. For instance, a sensor that is measuring something produces a stream of measurement messages, identifying itself with its hardware address. The information is collected to a database somewhere in the network, and that database can correlate things like "urn:dev:ow:10e2073a01080063" meaning the sensor in Jari's kitchen. There are also other possible use cases, equipment catalogs, for instance.

http://tools.ietf.org/html/draft-arkko-core-dev-urn-01

The draft has three different address types currently, but the list could be changed. For instance, we know that the third variant, cryptographic identifiers, is not strictly speaking aligned with current requirements on what URNs must satisfy, as cryptographic identifiers have only statistical, not administered uniqueness.

Thoughts? Good idea? Bad idea? Improvement suggestions? Other things* we should use instead?

Jari

*) The draft has some discussion of why we feel UUIDs are not suitable in all cases.