Re: [Cbor] I-D: draft-rundgren-cote-00

[Anders Rundgren <[e-mail]>]

On 2022-07-19 22:07, Carsten Bormann wrote:
> On 2022-07-19, at 18:20, Joe Hildebrand <[e-mail]> wrote:
>>
>> tag: URIs from RFC 4151 seem like they're a pretty good fit for this sort of thing.
> 
> Great recommendation.

If we stick to usability which is the #1 criterion for most people, it doesn't really matter what kind of identifier you use since you in 99.99% of all cases anyway depend on a dereferenceable (browsable) URL in order to gather information about the identifier [*].

That domains sometimes disappear is true but hardly something we can do much about.

Fortunately, the advent of public repositories supporting any number of sub-domains like GitHub, make URLs quite handy performing "dual service" which is the main target for COTE.

As an example https://fido-web-pay.github.io/ will in its next iteration use a browsable COTE identifier from this name space.

Cheers,
Anders

*] The CBOR WG already have URL dependency issues in IANA.cbor-tags: how long will http://cbor.schmorp.de/perl-object live? Since it doesn't open using https, it is probably not overly maintained either.

> 
>> Easy to mint, date included in case the domain changes hands, can't be resolved by mistake, and easy enough to reason about by inspection.
> 
> Which is why we actually use them in the example in core-problem-details [1].
> 
>>   Example from the RFC:
>>
>> tag:yaml.org,2002:int
> 
> Our example [1] is:
> 
>    "tag:3gpp.org,2022-03:TS29112": {...
> 
> (The example is set in a 3GPP context, and this fictional structure is relating to 3GPP TS29112; complete with warning that they really should register an identifier for this.)
> 
> Grüße, Carsten
> 
> [1]: https://www.ietf.org/archive/id/draft-ietf-core-problem-details-08.html#fig-example-custom-with-uri
>