Re: Re: [Zcip] embeddable zeroconf/ip code

On Thursday 21 October 2004 00:28, Brad Hards wrote:
> On Tue, 19 Oct 2004 06:21 am, H. Peter Anvin wrote:
> > > Or am I missing something? =C2=A0The only real issue I can see here
> > > with PRNG is ensuring the sequences repeat no sooner than every
> > > 2^16 - 512 samples.
> >
> > Again, it's a quality of implementation issue.
>=20
> I don't think that matters much (if you get much=20
> beyond 10 you're dead anyway, =20
> because of the exponential backoff requirement),

Well MAX_COLLISIONS is indeed 10, but that just kicks
in an extra minute's delay between retries; there's no
exponential anything, just up to a few seconds' worth
of random delay during the probe and announce stages.


> just that there is a =20
> reasonable spread.
>=20
> I do have a unit test for the PRNG spread from zcip
> though - see attached. It =20
> relies on a framework called check, but the changes
> shouldn't be too bad to =20
> delete that dependency if you cared.

The code is already very close to zcip 0.4 and will
probably get a bit closer to it.  Thing is, the PRNG
is also being used for the random delays, so the
sequence of IP addresses chosen will also vary based
on how early the collisions get detected.

- Dave


-------------------------------------------------------
This SF.Net email is sponsored by:
Sybase ASE Linux Express Edition - download now for FREE
LinuxWorld Reader's Choice Award Winner for best database on Linux.
http://ads.osdn.com/?ad_id=5588&alloc_id=12065&op=click