Hello, More and more groups are using network descriptions, and want to identify specific resources, e.g. a specific node, link, or instance of a (light)path. Defining an instance identifier format is also in scope of the NML-WG. Yesterday we discussed this in the meeting, with several proposals (see the slides on GridForge). In the end in the meeting there was consensus on using the following: urn:ogf:network:<dns name>:<opaque part> So for example: "urn:ogf:network:example.net:xyz123". The opaque part of the identifier is determined by the originating domain, and it should be completely opaque. Other domains should not interpret that part, other than seeing it as a unique identifier for a resource. This makes it possible to define equality as a simple string comparison between identifiers. If you have any questions or comments on this proposal, please let us know. Regards, Jeroen.
Jeroen van der Ham wrote:
In the end in the meeting there was consensus on using the following:
urn:ogf:network:<dns name>:<opaque part>
If you have any questions or comments on this proposal, please let us know.
The use of URNs comes with some requirements, as defined in these RFCs (in order of importance): http://tools.ietf.org/html/rfc3986 - URI: Generic Syntax http://tools.ietf.org/html/rfc2141 - URN: Generic Syntax http://tools.ietf.org/html/rfc2611 - Clarifications and Recommendations http://tools.ietf.org/html/rfc3305 - URN Namespace Definition Mechanisms For example, any URN "is required to remain globally unique and persistent even when the resource ceases to exist or becomes unavailable" (rfc3986). So identifiers may never be re-used, period. * Is there a maximum length for a name? * What normalization is required for lexical equivalence? (just UTF-8, no BOM, NFC? lower case/upper case?) * what are allowed characters (may the opaque part contain ":" or the NULL character? Or Line Feed?) * whose dns name is used? From the source domain, or from whoever generates the identifier? In particular, what if domain A does not (yet) generate global identifiers, but domain B has a link terminating in that domain. May domain B just generate some sort of identifier just to be able to refer to it? I think most of the answers are in the RFCs, but it is good to be explicit, especially about the normalization, as that is mostly left open for users of a URN. I'm willing to work with Lars, who has volunteered to write the section on identifiers. Regards, Freek Dijkstra
participants (2)
-
Freek Dijkstra -
Jeroen van der Ham