Thus spake John Case (case@xxxxxxxxxxxxxxxx): > On Wed, 1 Dec 2010, Maciej Zbierski wrote: > > >I was going through the "Coding Projects" site the other day and spotted > >that Tor is in need of a simulator for slow connections. I have written > >something similar as a part of my M.Sc., so I thought I could contribute by > >adapting my code to Tor's needs. First of all, has there been any progress > >on the subject (so that I don't double someone's work)? > > Why is this a coding project ? Why don't they just use FreeBSD with > dummynet ? I've personally used Linux's NetEM for testing my Tor Circuit Build Timeout learning code: http://www.linuxfoundation.org/collaborate/workgroups/networking/netem It worked well but it has a major drawback in that if you use netem on the same machine as your code is running, you can't do real packet loss simulations because the Linux networking stack hints to the TCP layer that the loss was artificial so that window sizes are not adjusted as they would be in reality: https://www.linuxfoundation.org/collaborate/workgroups/networking/netem#Caveats Getting a patch to disable this behavior would be a great help. Otherwise, I think we need to revise this task on our website to say that we're looking for a Tor network simulator, not a slow network simulator. Easy mistake, I guess ;) -- Mike Perry Mad Computer Scientist fscked.org evil labs
Attachment:
pgpS4mzEcLjkD.pgp
Description: PGP signature