> Maybe this is just a stupid question, but do they really need to maintain 
> state of the entire conversation? I mean, both parties could just maintain 
> the result of a hash over the conversation so far, built in an iterative way. 
> Something like this:
> 
> state' = hash (state, new_message)
> 
> At the end of the conversation, they could compute GSS_mic over the state.
> 
> May be the overload of computing such hashes is worse than the memory needed 
> to store the entire state.

It's not this so much. The problem is hash agility: we want GSS EAP to use the 
crypto primitives provided by RFC 3961, and this does not have a primitive for 
incrementally updating a checksum. So, we could extend RFC 3961 (which, wearing 
my implementors' hat, will involve some fairly intrusive changes to the popular 
Kerberos implementations). Or we could use a fixed checksum algorithm, but 
that's brittle and makes the entire protocol vulnerable if a weakness is ever 
discovered. Or we could build in some mechanism for negotiating checksum 
algorithms, but that risks re-inventing the wheel.

Sam or Nico can explain this better than I, I'm sure.

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

Reply via email to