[Author Prev][Author Next][Thread Prev][Thread Next][Author Index][Thread Index]

Re: Why Tor is slow and what we're going to do about it



Interesting.  A couple comments.

1. The congestion control issues are interesting indeed.  I really wish
I had some time to work on tor.

2. I'm not sure that user education is the way; better fairness within
tor's congestion control algorithms feels more promising to me.  The TCP
community has done quite a bit of research about per-aggregate fairness of
congestion control; my favourite algorithm happens to be Stochastic Fair
Blue [1,2], which is cheap and reasonably fair.

3. One point you fail to mention is the lack of NAT traversal in tor.  I'm
sure it'd help if a tor relay could be run behind a NAT without the need to
manually punch holes.

Rather than developing our own variant of STUN, I'd recommend we simply get
tor to work over IPv6.  Then a user will be required to run Teredo, which
is a NAT-punching protocol for tunnelling IPv6 over IPv4; it's built into
Windows since XPSP2, and available for Unix as Free Software [3,4].

                                        Juliusz

[1] http://www.thefengs.com/wuchang/blue/CSE-TR-387-99.pdf
[2] http://www.pps.jussieu.fr/~jch/software/sfb/
[3] http://www.remlab.net/miredo/
[4] http://www.pps.jussieu.fr/~jch/software/ipv6-connectivity.html