Re: [tram] Sharing nonces across allocations - allowed?

[Jonathan Lennox <[e-mail]>]

On Feb 19, 2014, at 1:52 PM, Simon Perreault <[e-mail]> wrote:

> Le 2014-02-19 12:36, Jonathan Lennox a écrit :
>> Since it says that a client identifies a server by hostname or IP address, I interpreted this to mean that the client SHOULD use the nonce established in setting up one allocation to authenticate itself for other allocations on the same server.
> 
> Well, your client may try it, in an attempt to save a round-trip, but it
> still has no idea for how long a nonce is going to be valid. The server
> has absolute authority to declare a nonce stale whenever it wants, and
> the client must be ready for it.

Sure, absolutely.

> There's also this:
> 
>   For each allocation, the server SHOULD generate a new
>   random nonce when the allocation is first attempted following the
>   randomness recommendations in [RFC4086]
> 
> I suppose it could be interpreted either way, but I understand the above
> to mean that previous nonces are expired when the new one is generated.

That also makes sense to me.

>> Relatedly: what should a server do when presented with a nonce it's never before encountered?
> 
> Servers are not required to remember the nonces they generate. They can
> use HMAC magic to detect invalid/expired nonces. (FWIW, that's what our
> server, Numb, does.)
> 
> I suppose we could define a new "invalid nonce" error code in STUN-bis.
> However, there would no point to doing that unless it resulted in
> different client behaviour.

Right, I think this should be the 438; I can’t think of any reason to want client behavior to be different for expired nonces and invalid nonces.  (And, as you say, the server can’t always distinguish them.)  But right now the spec only says what to do for expired nonces.