Subject: Re: random ip_id must be configurable
From: David Laight <email@example.com>
Date: 09/13/2003 10:14:33
> > reducing? with ip_randomid(), the *guaranteed minimum* interval between
> > the generation of the same output value is 36000 calls.
> > even after 36000 calls, it is highly unlikely that we see the same
> > number generated from ip_randomid(). if you have concrete number
> > please show me.
> The following program (which originally used the kernel version of
> ip_randomid() but has been modified to use randomid(3)) shows that often
> enough, _consecutive_calls_ to randomid(3) return the _same_number_.
> This is using the 16-bit version of randomid(3) that (looks to) have the
> same configuration parameters as ip_randomid(). Very similar results
> were observed with ip_randomid() too.
I've done some further experiments it appears that neither the LCG nor
the exponential generator are always correctly built!
ru_seed2 breaks the exponential generator (it is really rather obvious why...)
(both ru_seed and ru_seed2 actually look pointless)
ru_b < 0 seems to be responsible for the LCG breakage.
David Laight: firstname.lastname@example.org