MaxOnionsPending questions

[Roger Dingledine]

On Fri, Aug 15, 2008 at 04:58:48AM -0500, Scott Bennett wrote:
>  What is a
> typical high water mark for the number of onionskins actually in a decryption
> queue at one time?  Under what circumstances?

Usually relays are well under the 100 mark. The only cases I've seen in
practice where it happens are:
a) If the Tor relay doesn't have much cpu to spare, and it's advertising
   a really high bandwidth value.
b) If some client goes berserk and starts sending hundreds of create
   cells.

We don't have a good solution for "b" yet. Perhaps if our queue gets
large, we should ignore create cells from relays that have recently
sent too many?

--Roger