[Nml-wg] URN urn:ogf:network

[Freek Dijkstra]

Aaron Brown wrote:

>> So first things first:
>> - Do we want to use URNs (e.g. urn:ogf:network or urn:ogf:nml) for
>> identifiers of the classes we define?
>> - If not, do we want to use URIs as identifier?
>> - If not, are there other potential identifiers to use?
>>
> For the class definitions themselves, I think it makes sense to use URIs 
> a la namespaces so we could put some documentation at the specified URL. 

I don't understand "URIs a la namespaces". I take it that you mean 
"URL", e.g. http://ogf.org/ns/network#device which -as Jeroen said- has 
the added benefit that you can also put the schema at 
http://ogf.org/ns/network.

So that is now 1.5 "vote" in favour of URN versus 2 "votes" in favour of 
URL.

My reason for chosing URN is that I think it is closer to an identifier 
than a URL. The underlying reason is that I dislike the implied 
information that a URL carries.

One one hand, I strongly prefer URI (either URN or URL) over anything 
else -- UUID, integers and other IANA-managed namespaces. URI have the 
benefit of being readable and do not suffer from exhaustion (like 
AS-numbers and IP-numbers do).

On the other hand, I also advocate a URI that is basically an opaque 
string. I don't want to "interpret" it. So I very much dislike e.g. 
"http://ogf.org/ns/network#device?type=router" to both signify a device, 
and the type of device. This means I also dislike the fact that e.g. 
"http://ogf.org/ns/network#device" implies that there is a schema (or 
more information) available at "http://ogf.org/ns/network". I simply 
dislike that an identifier carries implied information, so I have a 
slight preference for a URN over a URL, since that carries fewer implied 
information (the only implied information is that it is a standard 
defined by the OGF if we use the urn:ogf: namespace).


> For the identifiers for individual instances, I think the URNs make more 
> sense since it doesn't imply a specific method of access to get 
> information about the element.

I think this is a completely different discussion, despite that it is 
also about identifiers.
I like your argument that we -for the time being- should not imply a 
specific method of access. Therefor, to me it is an opaque string. 
Either URN, URL, UUID or some other gooblygook ;-)

I agree with Jeroen that whatever the identifier *for instances* is, it 
should be in some kind of namespace "owned" by the owner of the instance 
(e.g. the owner of a domain should give it an identifier that is part of 
his own namespace, not an identifier, part of the OGF namespace) -- I 
don't assume you were implying that, did you?


Regards,
Freek