Re: Singularity: can't happen here

From: Eliezer S. Yudkowsky (
Date: Sun Sep 09 2001 - 02:31:23 MDT

Action by the music industry against MP3 ftp sites and Websites gave us
Napster. Action against Napster gave us Gnutella and Bearshare.

So I've been wondering what kind of technology this implies. In general,
current solutions are decentralizing against legal attacks on a central
weak point, but they are not up to fully "revolutionary" quality, meaning
that you wouldn't want to use these solutions to support a "Moon Is a
Harsh Mistress" revolution. That level of security requires being able to
actually hide the fact that you're a node, and ensure that the discovery
and physical seizure of one node compromises a limited number of others
(ideally zero, but I can't figure out how to do that).

I don't have time to write a post that's as long as I'd like, but my
thoughts have been wandering into the territory of chained blind IP
obscuration - in other words, only the second-to-last recipient of a
packet knows who the last recipient is, because the address of the next
recipient in the chain is encrypted, and can be decoded only by the key of
the node that needs to forward it.

If the government attempts to make possession of a Bearshare application
into a felony, then there would be the additional problem of obscuring the
fact that you are acting as a node at all. Disguising the traffic strikes
me as a technologically more complex problem that could severely limit the
bandwidth of any solutions.

The bandwidth problem might result in a solution where it's possible to
have secure distributed P2P applications that involve moving small bits of
data or text without anyone noticing, but it's not possible to send entire
files without calling attention to yourself. The probable result as far
as "the successor to Bearshare" goes would be a network that traded the
opaque addresses of people who claimed to have certain files; the actual
files would be sent through a corresponding chain of ICQ recipients, or
some other "permitted" application, once the address trade was
established. On the other hand, running a disguised revolution, which
probably shouldn't be trading large files, would be an easier problem. I
wonder whether anyone has previously tried to develop some helpful
revolution-support software on behalf of the folks trapped in China.

It's a real pity that IPv6 isn't already an established standard; then
nobody but the final recipient would have any idea what the traffic was,
and it would simply be a matter of adding on a layer of abstraction so
that nobody but the final and second-to-final recipients knew where the
traffic was going. We need more legal peer-to-peer applications that
routinely encrypt their traffic, so that a revolutionary-quality system
can pretend to be one.

The prospect of *development* of this software becoming illegal leads me
to think that the software will be developed in some non-signatory nation,
but there's also the territory of opaque distributed development.

-- -- -- -- --
Eliezer S. Yudkowsky
Research Fellow, Singularity Institute for Artificial Intelligence

This archive was generated by hypermail 2b30 : Fri Oct 12 2001 - 14:40:26 MDT