On Wed, Oct 5, 2011 at 6:08 PM, Luke Howard <[email protected]> wrote:
>> a) there are no sequencing problems, since RFC4121 can handle
>> out-of-sequence tokens;
>
> Hmm. I'm not convinced that this wouldn't have created other problems, I need 
> to think about it some more.

Well, if you can guarantee that all per-msg tokens will have been
processed by the time you're ready to handle the app's tokens...  Ah,
now I understand Sam's comment about PROT_READY.

>> b) you could create separate "contexts" for internal each use of
>> RFC4121, with different keys in each case, all derived from the EAP
>> keys.
>
> Yes, this would have worked, although using usage numbers and 3961 is a 
> little easier from an implementation perspective (you only have one derived 
> key to manage).

Yes, which is why I agree with the proposal :)

>> Now, RFC4121 is going to require RFC3961/2 anyways, so using RFC3961/2
>> directly in this part of GSS-EAP is hardly a burden -- unless you have
>> an API to the first but not the second...
>
>
> Well, there is an API for RFC4121. But there's no clear way to manufacture a 
> context purely in order to use its cryptographic services. (e.g. you could 
> import a fake context, but you'd have to understand its contents, which are 
> implementation dependent.) So I think it's 6 of one, half a dozen the other.

We ought to define a subset of RFC4121 exported sec contexts for the
purpose of using it as a per-msg token library!  :)

Nico
--
_______________________________________________
abfab mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/abfab

Reply via email to