OAuth 2.0 defines two endpoint, each with a set of error codes. These codes are 
not extensible and therefore do not require a registry. If you want to allow 
error code extensibility, you need to make the case for that with requirements 
and use cases. I have not seen any.

My API uses the 'position' parameter for protected resources. Am I expected to 
register it? It has nothing to do with OAuth, but if I supported Bearer token 
would be used alongside the 'oauth_token' parameter. The 'oauth_token' 
parameter is an nasty hack to make it easy for developers to access protected 
resources without using the Authorization header (based on browser limitations 
from 4 years ago). Defining a registry for this parameter is just adding insult 
to injury.

The bearer token draft can define whatever registries it want for those using 
*bearer* tokens. It has no say on anything else. Period. If you want to make 
changed to other drafts, you need to make a case and build consensus. By 
definition, your drafts cannot change any requirement in other drafts unless 
you update them (this is an IETF process rule).

Can you provide use cases, requirements, and examples for each of your new 
proposals?

EHL




From: Mike Jones [mailto:[email protected]]
Sent: Friday, February 25, 2011 2:40 PM
To: Eran Hammer-Lahav; [email protected]
Subject: RE: OAuth Errors Registry and OAuth Parameters Registry

I see a problem with collision of error codes.  I believe that the working 
group will likely concur.

I agree that it is odd for the bearer token specification to impose 
requirements upon the framework specification.  Hence my editorial request for 
you to incorporate these registry additions into the framework specification.  
Nonetheless, if you do not, the requirement remains, as the two specs are 
intended to work together as a seamless whole.

It is far from absurd to require parameter registrations for the resource 
request and resource response messages in OAuth specifications.  Phil and 
Brian's notes just today point out the value of just such registration.

Again, I'll politely repeat my request to update the framework specification, 
and other specifications that you are an editor of, to comply with these 
requirements.

                                                                -- Mike

From: Eran Hammer-Lahav [mailto:[email protected]]
Sent: Friday, February 25, 2011 2:33 PM
To: Mike Jones; [email protected]
Subject: RE: OAuth Errors Registry and OAuth Parameters Registry

I don't see a point in an error registry (and I find it odd for the Bearer 
token specification to impose any requirements of the protocol specification). 
Registries are created to prevent namespace collision. I don't see real problem 
with collision of error codes, especially not for a while.

The OAuth protocol errors are not extensible. To extend them, a specification 
must be published that updates the OAuth 2.0 RFC. Not making it extensible 
actually helps interoperability.

You can't define a "resource request" and "resource response" locations. That's 
absurd. The resource endpoint is service specific and by definition will have 
an unlimited number of parameters. How exactly do you expect this registry to 
work? Every provider registering every parameter they are going to use as part 
of their own API? It's enough that the 'oauth_token' parameter invades the 
provider's space as it is.

No changes are needed to the OAuth 2.0 protocol specification and I'd suggest 
you drop all these new additions from the bearer token draft.

EHL



From: [email protected] [mailto:[email protected]] On Behalf Of Mike 
Jones
Sent: Friday, February 25, 2011 2:18 PM
To: [email protected]
Subject: [OAUTH-WG] OAuth Errors Registry and OAuth Parameters Registry

I wanted to explicitly call out that draft -03 of the Bearer Token 
Specification establishes the OAuth Errors registry to increase 
interoperability among the related OAuth specifications.  Eran, when you 
produce draft -14 of the framework specification, please register the errors in 
the specification to comply with this new requirement.  Errors in other OAuth 
specifications should likewise be registered in updated drafts.

I also wanted to call out that draft -03 augments the OAuth Parameters registry 
by adding two additional parameter usage locations: "resource request" and 
"resource response".  Like parameters for the parameter usage locations 
authorization request, authorization response, token request, and token 
response, parameters for these usage locations must also be registered.

Eran, if you would prefer, on an editorial basis, to move these OAuth registry 
requirements into the framework spec, I would be fine with that.  If you do so 
in your draft -14, I will remove them from my draft -04.

                                                                Best wishes,
                                                                -- Mike

_______________________________________________
OAuth mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/oauth

Reply via email to