Re: K-of-N information dispersal for Mixmaster

[Len Sassaman <rabbi@quickie.net>]

On Tue, 20 Aug 2002, David Hopwood wrote:

> Nomen Nescio wrote (on alt.privacy.anon-server and sci.crypt):
> >
> > Problem.  In the mixnet remailer system, large messages
> > are split up into a sequence of short fixed-size chunks,
> > mailed through different chains, and reassembled at the
> > exit point.  If one or more of the pieces fails to arrive,
> > reassembly fails and the entire message is lost.
> >
> > Wouldn't this be a good case for using an Information
> > Dispersal Algorithm which is redundant enough to allow full
> > recovery of the original data even in a lossy environment?
> >
> > This would make the remailer system fairly tolerant of high
> > rates of dropout, single-packet delays, local gridlock, etc,
> > with only 10%-20% overhead even for very large split messages.
> >
> > Would this be an issue to raise with Lance Cottrell, or has
> > he passed the mantle to another vic^H^H^H maintainer?
>
> I suggest discussing it on the mixminion-dev mailing list, which is
> designing the Type III remailer protocol (archive and subscription info
> at <http://archives.seul.org/mixminion/dev/>).
> See <http://www.mixminion.net/> for a draft paper on the design.
>
> My immediate take on this is that it's an end-to-end issue, and does not
> require support in intermediate nodes. But yes, it would probably be a
> good idea to support an IDA when fragmenting/reassembling messages.

Hmm. I proposed this a while back. I seem to recall that there were
objections over performance issues raised at that time.

Let me see if I can find the thread...


--Len.