[tor-talk] About time to make BitTorrent work over Tor,

[grarpamp]

On 8/26/13, Kostas Jakeliunas <kostas at jakeliunas.com> wrote:
> On Mon, Aug 26, 2013 at 2:46 PM, Nathan Suchy <
> theusernameiwantistaken at gmail.com> wrote:
>
>> I don't want this for piracy as I have a paid VPN account that is much
>> faster for that if I decide to pirate. I think we need BitTorrent though
>> to
>> work on Tor so Tor Users can securely share files with one another.
>
>
> AFAIK the most obvious issue with this (among more subtle side-channel
> attack / decloaking problems) is network scalability. Total relay bandwidth
> available is, while seemingly increasing in general, very limited given
> such use cases. [1] How does one scale BitTorrent on top of that?
> ...
> But perhaps there's still some discussion to be had. I'm sure this has been
> discussed myriads of times, however - maybe it's worth trying to browse
> through the mailing list archives.
>
> [1]: https://metrics.torproject.org/network.html#bandwidth

Technically, Bittorrent works fine when run completely inside Tor.
The resources needed at the client are reasonable (ie: cpu).
The bandwidth is currently reasonably available. And onions are
secure enough for people who insist on skirting copyright. That
alone makes me wonder why Tor hasn't grown any long lived
torrent clouds. Probably because the entry bar is high and the
slower speed is beyond the scope of the typical leeching
mindset.

What people should know is that Tor currently can't handle it at
scale in at least two areas...

Transferring 1MiB causes about 7MiB worth of reduction in Tor
bandwidth, plus the CPU and state for processing the circuits.
You can get that back by running relays, but the vast bulk of the
above users probably won't be willing or bothered to do that, let
alone be able to figure out how to do it right and in conjunction
with their BT app. So Tor is likely to tank from that alone.

Then assuming they did run relays, the last part is dealing with
N million users worth of relay and onion descriptors. That takes
out more bandwidth, the dirservers, and everyone's local cpu
and ram.

For those reasons, it's hard to truly push mass use of p2p over
Tor. Yet I think given the greater probability of a p2p influx as
time goes on [1], Tor really should have some prepared design
considerations on the table beyond "please don't do that",
which torrenters obviously ignore. Hopefully designs can be
found that aren't restrictive.

[1] Pick any hot p2p app, drop N million users worth on Tor. In
this thread it's BT... being driven slowly towards anonymous
systems by the anti's [today, Russian blocking], enhanced
by piratebrowser giving hints to millions of users "Hey, what's
this Tor thing? What are these onions I see there? Hmm..."