Re: Pseudorandom Flow Labels

[Fernando Gont <[e-mail]>]

Hi, Shane,

On 05/04/2011 04:22 p.m., Shane Amante wrote:
> With respect to your comments on, both at the mic at the 6MAN WG and
> on the list: draft-ietf-6man-flow-3697bis-02 
> draft-ietf-6man-flow-ecmp-01 draft-ietf-6man-flow-update-04
> 
> You seem to take issue with a recommendation for creating/selecting a
> flow-label that is "pseudo-random".  Can you clarify the reason(s)
> you take issue with creating/selecting a flow-label that is
> "pseudo-random"?  (IMO, the current 'SHOULD' in the draft provides
> enough leeway that if implementations have good reason(s) to not
> select pseudo-random values for flow-labels they can choose to do so,
> but would hopefully have good reasons for that).
> 
> The reason I ask is as follows: 1)  RFC 6056: Recommendations for
> Transport Protocol Randomization, in particular Appendix A.  In
> short, a lot of [deployed] implementations are already computing a
> pseudo-random value for a transport protocol port (within the
> ephemeral port range), today, so IMHO the implementation "burden" of
> choosing a pseudo-random value must be quite low and, in theory,
> there should be good re-use of that code/logic for selecting
> flow-labels. 

And even if they had to implement this support "from scratch", the code
would be very small -- see e.g. the recent patch for port randomization
that was posted to freebsd-net.



> 2)  If we expect that if a intermediate router or switch
> is using *just* the 3-tuple of {src_ip, dst_ip + flow_label} as
> input-keys to compute a load-balancing hash algorithm, then the more
[...]

+1 for pseudo-random Flow Labels :-)

Thanks,
-- 
Fernando Gont
e-mail: [e-mail] || [e-mail]
PGP Fingerprint: 7809 84F5 322E 45C7 F1C9 3945 96EE A9EF D076 FFF1