I prefer a 1.1 version because I find it less confusing to tell
Consumers "To use my SP, you must find a library that supports OAuth
1.1" than to say "To use my SP, you must find a library that supports
OAuth Core 2009.1"

Upon further reflection, this is a minor issue. What I really care
about is having consistent vocabulary for describing what form of
OAuth an SP or a library support.  I'd be satisfied with either 1.0 or
1.1 if we clearly documented how to refer the different forms of OAuth
(perhaps in Appendix A.1?) .

Jesse

On Fri, May 1, 2009 at 2:01 PM, Eran Hammer-Lahav <[email protected]> wrote:
>
>
>
>> -----Original Message-----
>> From: [email protected] [mailto:[email protected]] On Behalf
>> Of Dossy Shiobara
>> Sent: Friday, May 01, 2009 1:47 PM
>> To: [email protected]
>> Subject: [oauth] Re: This whole version business
>>
>>
>> On 5/1/09 3:06 PM, Eran Hammer-Lahav wrote:
>> > They didn't change the protocol version (as in 'GET /something
>> > HTTP/1.1') because it added no value and would have just broke the
>> web.
>>
>> Explain how rev'ing HTTP to 1.2 would have "broke the web" ... ?
>
> Millions of client and server would no longer be able to interoperate without 
> deploying new software, servers, proxies, caches, etc. When the client and 
> server speaks a different protocol, they cannot interoperate. But they 
> *obviously* can even with the changes made to HTTP 1.1 (reality proves this). 
> If they changed the version to 1.2, old clients will no longer work with new 
> servers and the change would have added confusion. The key here is 'added no 
> value'. If you need to change the wire version for an actual technical 
> reason, do it. But since changing the wire version break stuff, you need to 
> have a reason to do it.
>
>
>> And similarly, how does changing oauth_version to 1.1 "break" OAuth?
>> Can you actually outline an actual scenario where this happens?
>
> New client speaking version 1.1 tries to talk to a 2-legged OAuth endpoint 
> which was not upgraded because *there is no reason to*. Pretty much all large 
> providers use a centralized OAuth token service with distributed API servers. 
> The server issuing tokens is *different* from the server accepting 
> OAuth-signed requests. If we leave the wire version as-is, only the 
> centralized server needs changes, not any of the API servers. This is a *big 
> fucking difference* to *actual running code*.
>
>> I thought the whole point of the proposed change to OAuth is to _close
>> a
>> security hole_.  That means, requests made to or from an implementation
>> of the previous specification are INSECURE and SHOULD NOT COMPLETE,
>> PERIOD.
>>
>> Or, have I learned a different definition of "security hole" than what
>> the OAuth community uses?
>
> There is no hole in the signature workflow. Only in how new Access Tokens are 
> being issued. We are solving it with this fix. This fix doesn't require 
> changes to the wire version because it is easy to identify by any server. So 
> yes, you are missing the point. Half the spec isn't broken, doesn't suffer 
> from a security hole, and does not need to be broken.
>
> EHL
>
> >
>

--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google Groups 
"OAuth" group.
To post to this group, send email to [email protected]
To unsubscribe from this group, send email to [email protected]
For more options, visit this group at http://groups.google.com/group/oauth?hl=en
-~----------~----~----~----~------~----~------~--~---

Reply via email to