Re: [Uri-review] draft-farrell-decade-ni

[Eric Johnson <[e-mail]>]

Hi Graham,

I'm not sure I see how your conclusion about treating this as a relative 
URI makes sense.

If "ni://example.com/index.html" is the obvious result of the relative 
computation, but is in fact meaningless in the context of the definition 
of the scheme, then I don't think that's harmless result. The 
expectation of a hierarchical scheme seems to me that you can apply 
standard hierarchical manipulations, and get a new valid URI, even if 
there's no "there" there at the given endpoint.

With the "ni" scheme as proposed, almost the whole universe of relative 
computations implied as allowed because of the "//" will result in URIs 
that are not actually valid according to the scheme definition. That 
seems ... harmful.

For the work I did on the "jms" URI scheme, we also had to address the 
question of an "authority" for resolving some of the URIs, and we were 
able to readily incorporate that into a non-hierarchical scheme.

-Eric.

On 5/6/12 9:37 AM, Graham Klyne wrote:
> On 04/05/2012 10:22, Eric Johnson wrote:
>> I looked at this proposal, and had three points of confusion.
>>
>> a) Why is this being done as a "hierarchical" URI scheme, when the 
>> URI is not,
>> in fact, hierarchical?
>>
>> That is, why ni://example.com/....?
>
> I think the document explains that.
>
> And I even have a possible use-case for this in an application I'm 
> working on - a hashed identifier where further information may be 
> available from the given authority.
>
>> When implementations that manipulate URIs encounter the :// after the 
>> scheme,
>> that should be a signal for them to associate meaning to "relative" 
>> URIs, as in:
>>
>> base URI: 
>> ni://example.com/sha-256;f4OxZX_x_FO5LcGBSKHWXfwtSx-j1ncoSt3SABJtkGk
>>
>> ... what does it mean to do "index.html" relative to said URI?
>
> The answer to that is clear from RFC 3986:  
> ni://example.com/index.html ... even if it's not very useful.
>
>> Can I compute a "sha-512" relative to the given URI? (Answer: no)
>
> I'm not sure I understand the question, but one might want to express 
> an already-known sha-512 with respect to the same base.
>
>> I conclude that this should not be a hierarchical scheme.
>
> I don't see any harm, and some possible (small) benefits.
>
> #g
> -- 
>
>> b) Why isn't this just a URN with a particular namespace identifier?
>>
>> As in:
>>
>> urn:nhi:example.com:sha-256:f4OxZX_x_FO5LcGBSKHWXfwtSx-j1ncoSt3SABJtkGk
>>
>> (where nhi I just made up, but stands for "named hashed id", because 
>> URNs
>> require at least three letters for namespace ids)
>>
>> That seems like a better fit.
>>
>> c) Since there are more known SHA standards, why not declare them all 
>> in this
>> initial proposal, so that sha-384, and sha-512 are already defined?
>>
>> -Eric.
>>
>> On 4/30/12 5:56 PM, Stephen Farrell wrote:
>>> Hi,
>>>
>>> We have a draft [1] that requests two new URI schemes.
>>>
>>> The core WG are likely to want to use these we think
>>> and possibly decade, but they're intended to be generally
>>> useful as well.
>>>
>>> Barry Leiba is planning to AD sponsor this and Alexey
>>> Melnikov will be shepherding so if you can cc them ase
>>> well as the authors on any questions or comments that'd
>>> be good.
>>>
>>> I hope the plan is to IETF LC this soon, once this
>>> review and the .well-known registration review are
>>> done.
>>>
>>> Thanks,
>>> Stephen.
>>>
>>> [1] http://tools.ietf.org/html/draft-farrell-decade-ni-05
>>> _______________________________________________
>>> Uri-review mailing list
>>> [e-mail]
>>> https://www.ietf.org/mailman/listinfo/uri-review
>> _______________________________________________
>> Uri-review mailing list
>> [e-mail]
>> https://www.ietf.org/mailman/listinfo/uri-review
>>
>