Re: [tsvwg] [Tsvwg] lookup time (was Re: WGLC for Port Randomization starts now (April 1st))

[Fernando Gont <[e-mail]>]

Mark Allman wrote:

>> If you retry random port, chances are that you may retry the same port
>> more than once, instead of trying a different port that might be
>> available. The larger the number of ports in uses, the more likely
>> this would be. (see the pseudo-code).
> 
> Sure.  Of course.  But, we can put this in concrete terms it would
> seem.  I.e., if we use something like alg. 3 then for a given connection
> ID we will not test for some port X again before exhausting all the
> non-X ports.  In something like alg. 2 you can in fact check X more than
> once with a probability defined by the port usage at the given time and
> also not test for all possible ports.

Agreed. And in the worst case, because of checking X more than once, you
may end up not checking other port numbers that might have worked.

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