[OTR-dev] MAC keys to be revealed

[jbash at velvet.com]

> > . combine this with some onion routing (tor.eff.org) 
> 
> Pfff .tor. I am still not convinced. So much local state to transfer
> over something stateless. There was a reason SOCKS wasn't popular.
>
> Paul, believes more in IPsec with onion routing.

Um, three layers of IPsec SAs is about as much state as three layers
of TCP connections plus crypto context.

I would have been happier if Tor had carried raw IP datagrams, myself,
but I don't think state is the issue. There's always going to be some
state until public key is close to as fast as symmetric crypto, and
maybe even then if you have a billing system and are worried about
replay protection. I was more worried about isochrony and queue stalls
and excessive retransmission and that sort of thing.

Tor is the way it is because Roger wanted it to be able to run in user
space as non-root.

To pull it back onto the topic, I do all my IM over Tor. It works
great. Every time I chat with Ian, it's OTR over AIM over Tor.

The SOCKS support in Gaim makes it easy to do IM over Tor
transparently. I've also run xchat over Tor, complete with DCCs, and
HTTP over Tor with Mozilla and privoxy... all with unmodified software,
which was enabled by the surprisingly great popularity of the SOCKS
protocol...