From: rishab@dxm.ernet.in
Date: Tue, 22 Nov 94 16:03:49 PST
To: cypherpunks@toad.com
Subject: Re: DC-Net implementation
Message-ID: <gate.95m8Vc1w165w@dxm.ernet.in>
MIME-Version: 1.0
Content-Type: text/plain


acspring@knoware.nl (Andrew Spring):
> >Using a central node to coordinate the DC-net traffic requires that the
> >participants trust that central node.  If the central node is evil,
> A collation of N-1 nodes will always produce garbage; the whole set is
> needed for the message to fall out.
> Example
> ...
>         A sends 14 -  5      =  9 to Central node
>         B sends  5 - 11      = -6

Central nodes are not nice for various reasons, including the usual networking 
and security (trust) advantages of wide distribution. IAC a central node is
not necessary; for example, if each node were to output to the next:
  B sends Anum XOR Brnd (XOR msg) to C
where Brnd is B's random number, msg is B's message (if any) and Anum is the
similarly generated output of A. This is much closer to the original DCNet 
_bit_ flipping - the first XOR checks for equality while the second commits the
'lie' 

In this case whatever number A gets from D is the output of the net. There are
lots of interesting cryptographic sub-protocols to make too much trust 
unnecessary. I also believe that error-correction is best left to lower levels
of the network - there's no need for a DC Net not to assume a reliable data
channel.

-----------------------------------------------------------------------------
Rishab Aiyer Ghosh             "Clean the air! clean the sky! wash the wind!
rishab@dxm.ernet.in                   take stone from stone and wash them..."
rishab@arbornet.org
Voice/Fax/Data +91 11 6853410  
Voicemail +91 11 3760335                 H 34C Saket, New Delhi 110017, INDIA