Re: stress testing

[Roger Dingledine <arma@mit.edu>]

On Thu, Jul 11, 2002 at 05:16:55AM -0400, Roger Dingledine wrote:
> Might be nice to simulate not just a user, but also another OR. That
> way we can hammer it using cells, too many cells, broken cells, random
> cells, cells of an unexpected type for a given state, etc. (Ok, maybe
> 'simulate' isn't the right word there anymore. But yeah.)

And if rather than focusing on the network side, we can instead simply
make this part of the stress testing framework be *part of the process*,
rather than a separate program, then we'll be on our way to having unit
tests of part of the onion router. :) Basically, since I've separated
the read_to_buf and process_buf functions, we can simply prepare the
buf appropriately, let process_buf do its thing, and then check to see
if it did the right thing.

Is this too far away from your original goal? Perhaps it should be a
second project?

--Roger