no circuit loops?

[Paul Syverson]

On Thu, Oct 23, 2003 at 05:00:50PM -0400, Roger Dingledine wrote:
> On Thu, Oct 23, 2003 at 01:06:06PM -0400, Paul Syverson wrote:
[snip]
> 
> > On the other hand, the absence of loops gives information about a
> > circuit. E.g., there is an observer on OR_1's network connections.
> > Alice connects through OR_1, OR_2, OR_3, OR_4, OR_5.  OR_4 is bad.
> 
> Remember that for performance and usability our path length is actually
> just OR_1, OR_2 by default. So the whole loop question is a bit moot. :)
> 

But that's just a default yes? And, it's also just temporary. We
still need to consider whether we get more vulnerability from
allowing loops or disallowing loops (or e.g. disallowing ABA loops
but allowing larger ones). And we should probably work that out before we
have route selection code that is new and shiny.

There was quite a bit of work on route selection that never made it
into the old papers for one reason or another (at least I don't think
it ever survived to camera ready), and enough assumptions and
parameters have changed that it would take reworking anyway. There's
probably at least one paper's worth of stuff here, for our copious
free time. ;>)

aloha,
Paul