Re: [RAI] Global Service Provider ID - draft-pfautz-service-provider-identifier-urn-01

[Hadriel Kaplan <[e-mail]>]

On Sep 23, 2011, at 11:02 AM, PFAUTZ, PENN L wrote:
> The outstanding issues seemed to be
> 1. Is a new resource assignment necessary for the URN or could an existing one, Private Enterprise Numbers or ITADs be used with a URN defined to provide a fixed length parameter as requested in the I-D?
>  
> The I-D proposed a new resource, partly because the existing ones were intended for other purposes and already had many registrations. I’d also note that PENs suggest a sub-delegation approach to assigning multiple SPIDs to an entity (an entity gets one PEN assignment and can assign values hierarchically below that (e.g. AT&T has PEN 74, and would assign 74.1, 74.2, etc). Because at least some of the service providers wanted to use SPID to prefix E.164 numbers and others want to incorporate SPID in a domain name, this would present issues.

I don't personally care whether a new registry space is created for SPID or not - it's not like IANA is running out of numbering spaces. :)  And although it takes IANA a bit of work to create a new one, I don't believe it's hard.

The thing that confuses me, though, is every characteristic described in the draft is exactly provided by ITADs, and although the ITAD space was created from RFC 3219 (the TRIP protocol spec), its semantic meaning was a Service Provider ID.  So when the draft says this in section 1:
   Although preference was also expressed for
   reuse of some existing identifier, if possible, as requirements have
   been elaborated no current identifier seems appropriate.

... it's befuddling.

So I asked some people about why ITAD wasn't what they wanted to use for SPIDs, and they said exactly what you said above:
> because the existing ones were intended for other purposes and already had many registrations


The former part isn't really important - PENs for example were created by SNMP but are now used in RADIUS VSAs, DIAMETER AVPs, IPFIX IEs, and others.  It's the number's semantics that matter, not what protocols encode it.  

It's the latter part (that they already have many registrations) that I think is really why people don't want to use ITADs, and that's what worries me in a weird way... because it implies they think some new SPID registry won't be populated with junk.  It will be.  Random people register for these things, because it's free and simple to do so.  Some people will get them for labs or test equipment, some for start-ups which go belly-up, some people get them just because.  For example there's a PEN for an enterprise called KRAP (Kaplan Research and Application Prototyping).  I registered it so I could use a PEN for some prototype code I wrote.  It will never be seen in SNMP, RADIUS, etc.  That's the kind of junk that will appear even if we start a new registry.  So my concern is that people *want* a clean registry, but don't realize that making it free+open virtually guarantees it won't be.

-hadriel