Transferring data quickly via Tor

Mirimir mirimir at riseup.net
Sat Jan 21 01:04:34 PST 2017


On 01/21/2017 01:27 AM, grarpamp wrote:
> There's onionbalance to consider, and there are some tools to
> expand entry guard, intro point and rend point counts.

I used tor-instance-create to get a separate Tor instance for each
OnionCat instance. And I specified a distinct bind port and SocksPort
for each OnionCat instance.

This setup multipaths individual tcp6 streams. And I believe that
onionbalance spreads incoming tcp connections across multiple servers.
Right?

> Also currently a shame that no one has yet stepped up to continue
> IPv6 interface support with tor after tor kills onioncat with prop224.
> It would be a nice project for someone.

Yes, that is a shame. I'm no coder, unfortunately. But I have been
considering the idea of crowdfunding a replacement for OnionCat. I
wonder what it would take to motivate someone.

> So many neat things can be done with what onioncat provides.

Indeed!

> Ditto onionvpn.

Not an iPhone user, so ???

> Also freebsd made some fairly significant bandwidth X delay,
> and windowing stack improvements that might not be in linux yet.

Maybe so. But do check out MPTCP.

~50 Gbps :)
http://multipath-tcp.org/pmwiki.php?n=Main.50Gbps

> Also, don't forget, if you're going to be eating up bandwidth
> and resources on overlay networks, you should be giving
> back at least an estimate of your impact in free resources.
> Which is typically your use times the hop count of the given
> network.

Yes, for sure. I should emphasize that more. And what's cool about this
approach is that entry guards and middle relays are easy to host.

There's also potential to increase usability of slow relays, through
aggregation. Perhaps ten 500 kbps circuits together would give users 2-3
Mbps overall.




More information about the cypherpunks mailing list