At 01:57 PM 6/25/97 -0700, Lucky Green <shamrock@netcom.com> wrote:
Regarding DC nets: Modern tree structures only double the bandwidth requirement. DC nets are practical today. I would encourage CP's to work on implementations.
How do you do that? Arrange the XORs to do x0^x1, x2^x3...., then (x0^x1)^(x2^x3)... etc., ending up with user 0 doing the last XOR, and broadcasting or tree-casting the results? It cuts down on bandwidth for the average user, but does end up with a few users doing most of the work, and perhaps there's some security risk in the unbalanced workload. It's an interesting approach, and I suppose you could do a bit more work (logn instead of 2) to spread the partial results around so more people can calculate them directly (e.g. user 0 sends user 2 x0^x1, and sends user 1 x2^x3, and if you're still paranoid you can have user 0 also send user 3 x0^x1, which user 2 and user 3 can compare....) You've probably got to put more thought into getting the details right, like collision detection and backoff, but it's still doable. # Thanks; Bill # Bill Stewart, +1-415-442-2215 stewarts@ix.netcom.com # You can get PGP outside the US at ftp.ox.ac.uk/pub/crypto/pgp # (If this is a mailing list or news, please Cc: me on replies. Thanks.)