[tor-relays] new relays

[Roger Dingledine]

On Tue, Aug 27, 2013 at 11:12:01PM +0200, Tor Exit wrote:
> Why is it so bad if a Tor exit operator tries to match the use of
>their node with their own moral beliefs?

I really would like to support this if I could.

Specifically, I'd love a way for exit relay operators to only allow
people to do things *via their exit relay* that they're comfortable with.

The trouble is, I only want to do it if we can have a way for Tor clients
to automatically learn what each exit will allow, so they can pick an
exit that will allow their connection.

We have that working with exit policies right now: each relay advertises
what IP blocks and ports it will allow, and then clients learn all the
exit policies and automatically choose an exit that will support their
stream. See Andy's post for details:
https://lists.torproject.org/pipermail/tor-relays/2013-August/002560.html

The trouble with more fine-grained approaches, where you look at the
content of the communication rather than the address of it, is that
the Tor client doesn't know the entirety of the communication when it's
selecting the path to use. This seems like an inherent contradiction,
especially since the client will need to know, ahead of time, everything
the *destination* (e.g. website) will send too.

(Ok, that's just the technical trouble. There are also legal troubles
with filtering some things you consider bad while not filtering everything
that anybody could consider bad. See the EFF Tor legal faq.)

--Roger