Re: [dnsext] Source Port and QID selection for re-transmits?

[Nicholas Weaver <[e-mail]>]

On Oct 23, 2008, at 3:58 AM, Wouter Wijngaards wrote:

> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
>
> [e-mail] wrote:
>> On researching a draft I'm writing I've been unable to find any  
>> guidance
>> on whether the QID and Source Port of a re-transmit should be the  
>> same as
>> for the original request.
>> What are the WG's thoughts?
>
> Leave to implementation.  Lengthening the time the packet is  
> accepted is
> one choice.  Choosing new ID, port (and no longer accepting the old  
> one)
> is another choice.  You could caution against birthday attacks, but
> please do not prescribe implementation.

I don't think it matters much:

Case 1:  Retry uses the same entropy.

Case 2:  Retry uses a new entropy set.

For BOTH cases, there is an additional window of time outstanding  
where a spoofed packet could be received.  There is a trivially slight  
attacker advantage on case 1, as the attacker can continue to send  
from a randomly selected subset of the entropy space instead of  
starting over, but if you have 32b+ of entropy, this advantage is  
going to be trivially small, and if you have only 16b of entropy, you  
are sunk anyway.


--
to unsubscribe send a message to [e-mail] with
the word 'unsubscribe' in a single line as the message text body.
archive: <http://ops.ietf.org/lists/namedroppers/>