[freehaven-dev] mix reliability vs robustness?

[Roger Dingledine <arma@mit.edu>]

|Our notion of reliability differs from robustness in that we do not aim
|for an ``all or nothing'' result for each message.  Instead, we focus
|on improving a sender's choice of MIX paths over random guessing in the
|``long run.''

Is there any way to make this more precise? It seems like I can paraphrase
the above as "robustness is a clear concept, but we're looking for
something different. it's, uh, different."

Our notion of increasing reliability is that over time as the dynamic
network progresses and changes, the number of messages that get through
increases up to some 'stable' point? and that if things get unbalanced
then the reliability goes down but a good system would rebalance things
so increase (maintain) reliability?

Whereas, the mix robustness notion is "right now, for this situation,
how sure are we that the message will get through? indeed, for *any*
situation, how sure are we?"

robustness isn't willing to let some things fall through the cracks in
order to get a better 'whole', whereas that's exactly what reliability
is all about.

Is there a better way to say this? Something that the theory folks at
the conference won't shoot down in an instant as being unformed?

--Roger