Re: URN namespaces for use in 'magnet' URIs

[Peter Saint-Andre <[e-mail]>]

On 7/25/11 12:16 PM, Mykyta Yevstifeyev wrote:
> 25.07.2011 15:51, Peter Saint-Andre wrote:
>> On 7/24/11 11:05 PM, Mykyta Yevstifeyev wrote:
>>> 24.07.2011 18:43, Peter Saint-Andre wrote:
>>>> On 7/22/11 5:46 AM, Mykyta Yevstifeyev wrote:
>>>>> Hello all,
>>>>>
>>>>> The 'magnet' URIs, which are quite popular with P2P networks, have an
>>>>> element named "exact topic" which contains the URN, forming an URI
>>>>> like:
>>>>>
>>>>>> magnet:?xt=<URN>&<other parameters>
>>>>> The URNs used in 'magnet' URIs must be a "hash URN", none of which are
>>>>> currently officially registered.  The hash URN is smth. like
>>>>> <urn:sha1:c3499c2729730a7f807efb8676a92dcb6f8a3f8f>   using SHA-1; see
>>>>> Wikipedia page on 'magnet' URIs -
>>>>> http://en.wikipedia.org/wiki/Magnet_URI_scheme#URN.2C_containing_hash_.28xt.29
>>>>>
>>>>>
>>>>> - for more examples of hash URNs.
>>>>>
>>>>> As I've mentioned, none of used hash URNs use the registered
>>>>> namespaces.  Correspondingly, my question - is there a sense in
>>>>> documenting such ones?
>>>> Why not use UUIDs?
>>>>
>>>> http://tools.ietf.org/html/rfc4122
>>> P2P networks use DHT, Distributed Hash Table, which identifies the file
>>> using hashes.  See http://en.wikipedia.org/wiki/Distributed_hash_table
>> I've heard of several people who are interested in defining a URN
>> namespace for hashes. Such a namespace would be similar to the UUID
>> namespace. However, RFC 4122 is not really a formal namespace in the
>> sense of all other URN namespaces. URNs are supposed to be formally
>> issued or generated by an authoritative entity that has responsibility
>> over the namespace.
>>
>> RFC 3406 says:
>>
>>     Assumption #2:
>>
>>        The space of URN namespaces is managed.
>>
>> The UUID namespace is not managed. A hash namespace would not be
>> managed. Although there is value in having a URI that enables people to
>> use UUIDs or hashes, I don't think they should be URNs. (Yes, the UUID
>> namespace already exists, but I don't think RFC 4122 would be published
>> now if we were considering it again.)
> Yes, as UUIDs, as hash URIs may not be managed; however, my point is
> that the latter are currently widely used.  I was asking whether they
> are OK to be documented; considering they may not suit the requirements
> for URNs documenting them won't be fine as well.  So the optimal
> approach is to let hash URNs be used further but retain them undocumented.

What I'm saying is that a URI would be fine, but that URNs would be
wrong, because the hash URI "space" would be unmanaged.

If folks want to continue this discussion, I think it might belong on a
different list.

Peter

-- 
Peter Saint-Andre
https://stpeter.im/