Re: [Tsvwg] table size (was Re: WGLC for Port Randomization starts now (April 1st))

[Mark Allman <[e-mail]>]

> > I think my formulation is better:
> > 
> >    I.e., note that [Allman] shows 10 entries will work fine for
> >    collision avoidance [to the extent that the data in [Allman] has legs
> >    outside those environments, of course] and increasing the table size
> >    increases the obfuscation and leave it at that.
> 
> Increasing the table might possibly decrease the collision rate, too.
> (i.e., it prevents the system from skipping port numbers
> unnecessarily).  I do not recall what were results of your research in
> this area, though... but will look at your paper again.

As I **said above**, I found that a table of 10 entries works fine for
collision **avoidance**.

You can likely dream up some scenario whereby a bigger table size does
in fact reduce the collision rate.  Or, perhaps whereby you can reduce
the collision rate to less than zero.  I prefer just looking at the
data, but realize YMMV.

Again, I am not asking you to say a table size of 10 is somehow ideal.
I am just saying that we have some data that suggests it is OK in some
environments and that plus a note that increasing the size increases the
obfuscation seems perfectly reasonable to me.  I don't see how pulling
some number out thin air is better.  I don't see how your number is
justified.  If you want to use it, great, I don't care.  But, I don't
see why we need to be recommending or suggesting this completely made up
value.

allman