Re: Pseudorandom Flow Labels

[james woodyatt <[e-mail]>]

On Apr 5, 2011, at 14:48 , Thomas Narten wrote:
> [I wrote:]
>> 
>> I share your concern.  Would replacing "pseudo-random" with "low discrepancy" address your concerns?
> 
> Replacing the term with another would be fine. That said, the real issue is we need to define what we mean by whatever term we use.


Discrepancy is a measure of the deviation of a point set from a uniform distribution.  Or, so Wolfram MathWorld tells me.  <http://mathworld.wolfram.com/Discrepancy.html>   A low discrepancy sequence would be one that has *approximately* uniform distribution [though, not necessarily *exactly* uniform], which is what seems to be the object of the game here.

There are many excellent well-known low discrepancy sequences to choose from, and I'm not sure IETF needs to specify a standard one, but it's possible one could be found if we really tried hard.  If you want to get hideously pedantic, you could specify a maximum discrepancy; I'm sure Timothy Winters and his crew would just melt if you did that.

Note: another term for "low discrepancy" is "quasi-random" but I'm not sure we want to go there.


--
james woodyatt <[e-mail]>
member of technical staff, core os networking