Re: [urn] URN fragments

["Martin J. Dürst" <[e-mail]>]

On 2013/02/22 12:36, Peter Saint-Andre wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> On 2/21/13 4:36 AM, Juha Hakala wrote:

>> Eventually Alfred Hoenes, myself and others concluded that the
>> best solution is to separate identification&  URN assignment from
>> fragment assignment, that is, not to include fragment in the NSS.
>> Which means that fragments can be added to the URN by anybody, any
>> time after the URN has been assigned and the document made
>> available (via URN resolution) in the Internet.
>
> Hi Juha, this model sounds similar to what you described for the query
> component: the fragment can be "added to the URN". I'm not sure what
> that means. When you add a query component or a fragment identifier to
> the URN, is that now a different URN (which wasn't assigned by the
> minting organization or in accordance with the minting algorithm)? Or
> is "a URN with some additional data appended" and therefore not really
> a URN anymore? Or is actually a URI (such as an HTTP URI) whose path
> is the URN itself as just a string of characters embedded in the
> broader URI?

This is different for the query part and for the fragment part. The 
fragment part is independent of the scheme. The query part is not.

You will see that in Appendix A of RFC 3986 
(http://tools.ietf.org/html/rfc3986#appendix-A). There, URI is defined 
as including an (optional) fragment part, but absolute-URI is defined 
excluding the fragment part. Same in RFC 3987 for IRIs. So in terms of 
terminology, it might be possible to call an URN without a fragment part 
an absolute URN (in places where such a distinction is necessary).

For fragments, the conclusion that Juha explains above is how it was 
intended for all URIs from the start, and how RFC 3986 is written. It 
may have taken some time for the URN community to arrive at this point, 
but it's a very good conclusion, because it's what works best with the 
rest of the infrastructure.

Regards,   Martin.