Re: [urn] [apps-discuss] URNs are not URIs (another look at RFC 3986)

[Phillip Hallam-Baker <[e-mail]>]

On Wed, Apr 23, 2014 at 6:04 AM, Maurizio Lunghi
<[e-mail]> wrote:
> I am wondering if it is useful and maybe necessary to define what a 'persistent identifier' is .... as many of you said technology is mot persistent per se but policy underpinning the system can make persistent the identifiers and the service linked to them .... in the APARSEN project www.aparsen.eu in WP 22 we reached agreement with many experts about a list of criteria that a 'trusted persistent identifier system' must have and we could start from this work to define a broader consensus .... doing so in my view we  could clearly distinguish between a URN or URI and a PI ... does it sound reasonable?
>

People have been compiling wish lists for properties of names for 20+
years. The issue is not what people would want, it is what they can
get.

I can't take the 'urn' approach seriously until someone provides an
explanation of

1) How the persistence properties are to be achieved

2) Why a taxonomic distinction between URLs and URNs helps achieve the
persistence properties.


A name is by definition a signifier that bears only a conventional
relationship to the thing signified. In the case of DNS that
relationship is mediated by the DNS infrastructure.

There are only two ways to make an identifier persistent. The first
being to use an index rather than a name. SHA256(c) is a persistent
identifier for the content c. Which is what we use in the ni: scheme.

The second is to have a first come, first served registry that never
revokes previous assignments. Now one could imagine a whole new
infrastructure for establishing such a scheme. But really, wouldn't
anyone setting up such a scheme apply for a TLD?

So now imagine we have .pid which is a first come, first served
registry that never revokes an assignment or alternatively requires
the date to be inserted into the HTTP URIs.

I can give you a HTTP method identifier which is persistent using such
an infrastructure.

Now we can argue over whether the proposal I make is actually
practical. But I think it is at least as practical as all the
'persistent identifier' schemes based on a new registry.


The bottom line is that persistence is a qualified property of a
naming infrastructure, not a syntactic distinction.

If people want to argue further, I want them to put up and show HOW
their persistence scheme actually works. No more wish lists or
requirements for identifiers. I want to see how the scheme works at a
technical and political level to achieve the persistence requirements.



-- 
Website: http://hallambaker.com/