DC-nets, up from the basic concept

[Pierre Uszynski]

A few pseudonymous questions went astray and landed in
my mail box among a few others (without much of a return
address too... the problems of anonymity)

So I'll volunteer some possible answers, and while doing
so will generate more questions :-). Others feel free to
elaborate (especially thoses who HAVE read the papers)

Still, no mention of detweil(er)ing will be made :-)
although I will mention spoofing...

1) "PHYSICAL LAYER"
===================

Q: How do you implement the idea that I can see my coin
and the coin on the left, but no others?

A: You use public key cryptography. You join a "table"
by contacting one of the persons already at the "table".
He hands you address and public key of (say) his
right neighbor, and you hand him your address and your
public key. Each time you "toss a coin" you encrypt it and
send it to his ex right neighbor (now your right
neighbor). You use the same public key protocols to
send out to everybody your "different/same" vote, and
collect and count everybody elses.

It's easy to add and remove individuals, and you can be
part of as many "tables" as you want.

2) MORE SPOOFING
================

New question: how do you make sure you are not spoofed
to death, and (mis)led to join a "table" of, say, 126
personae of the same three letter entity, and Yourself.
In which case, your messages are not very untraceable
anymore?

Not very satisfying answer: you meet in person your
left AND right neighbor at least once to exchange
address and keys... Any better answer?

3) GROUP RESPONSIBILITY AND HIGHER LAYERS
=========================================

Q: couldn't every member of the group be
implicated because they must all participate to produce
the bit of information?

A: No, when a "table" is established, what you have is
a CSMA/CD network in the classical sense (CSMA/CD = Carrier
Sense Multiple Access/ Collision Detection) just like
Ethernet is one.  Your "DC-net physical layer" is now
very complicated, and multi-layered itself, but that
only affects speed, not much else.

You could (maaaybe :-( establish the layering:

Coaxial - Ethernet - IP - TCP / Telnet - RFC822 email -
PEM - DC-net - IP - TCP / Telnet - RFC822 email - PEM

Except that now, your second layer of TCP/IP is
untraceable. By the same token (sorry :-), once you
have these protocols established for one "table", you can
internetwork many "tables" at will. Poof! Inter-DC-net!

But, boy, would the above implementations be slow :-)

In any case, even though the participants of any network
must cooperate for any message to go through (especially
true in the case of Usenet, or FIDO, for example), it is
hard to hold all the participants responsible for every
message just for being part of the infrastructure.

It would be different (even with todays networks), if
a large proportion of the messages were CLEARLY illegal.
Still, even now, it is impossible to tell whether an encrypted
message's content is illegal or not (see the previous
netnews discussion with respect to FIDO, mail privacy and
encryption), and peeking in un-encrypted email is
illegal, depending on caller contract.

So I would, until contradicted by the legal types
among us, consider a DC-net or Inter-DC-net no different
legally, than any other cooperative network out there
(Usenet, FIDO, internet...), until specifically
regulated.

Disclaimer: Of course, this is only a technical opinion,
legal decisions may or may not consider the technical
precedents :-) Reasonable technical arguments have as
much weight legally as unreasonable ones (or so it seems
at times) (meaning: of course, I'm not a lawuer)

Pierre.
pierre at shell.portal.com