[tor-relays] Unwarranted discrimination of relays with dynamic IP

[Rana]

-----Original Message-----
From: tor-relays [mailto:tor-relays-bounces at lists.torproject.org] On Behalf Of Matt Traudt
Sent: Sunday, December 04, 2016 6:20 PM
To: tor-relays at lists.torproject.org
Subject: Re: [tor-relays] Unwarranted discrimination of relays with dynamic IP



On 12/04/2016 10:39 AM, Rana wrote:
>> For as little as $10.00 US there are VPS' with static ip's..
> 
> Attn: Kurt Besig
> 
> Well I kind o' like my Raspberry Pi that cost me $40 including box and power supply and SD card and door to door delivery, with far more horsepower and memory than needed for running Tor relay, and my free and absolutely stable 1.5mbps that I want to donate to Tor courtesy of my ISP, and my transparent Tor proxy and my hidden service  and my wireless access point that lurk on the same Pi. 
> 
> This is not a good reason to punish my relay. Makes ZERO sense to me and to who knows how many people like me whose relays are flushed down the drain by the current DirAuth algorithms.
> 
> I can think of many an Iranian or Turkish or Chinese or Russian dissident who could use 1.5 mbps bandwidth to communicate with the free world.
> 
> 
> 

Perhaps all that other stuff you have running on the Pi is hurting your ability to max out your connection.

In any case, as I mentioned on your Reddit post a week or so ago, just because you have X available bandwidth, doesn't mean Tor will be able to use all X. I have some relays on 10 Gbps links. Even if they were only 1 Gbps links, the max traffic I'm seeing right now is about 65 Mbps. Atlas says I'm "advertising" (been measured at) ~140 Mbps.

https://atlas.torproject.org/#search/x76slvferal

So I'm pushing roughly half that atlas says I could be, and I'm pushing nowhere near the amount my hosting provider says my links are capable of.

I've heard (but haven't verified) that clients rarely use non-Stable non-Fast relays. So if you are struggling to maintain those flags, then that would be why you're having trouble getting up to 1.5 Mbps usage.

Here is how Stable is determined according to dir-spec

https://gitweb.torproject.org/torspec.git/tree/dir-spec.txt#n2267

Finally, I'd like to reiterate teor

> * a changed IP usually means a changed network with different
>   characteristics,
> * if the relay IP address changes, there's no guarantee it will be 
> just as reachable or stable at the new IP,
> * stolen keys become much less valuable,
> * duplicate keys / failover strategies are discouraged.

It sounds like your IP is _too_ dynamic for best supporting the network.

Thank you for running a relay and please do not be discouraged by numbers.

Matt

__________________________________

Thank you Matt but some of your assumptions concerning my relay do not hold water. 

Yes, I do have a Stable flag. 

No, my hidden service and my Tor proxy and My wireless access point are NOT hindering the operation of my relay, since I disabled them 3 weeks ago to make sure they do not interfere (and they could not possibly interfere when they were not disabled, their bandwidth, memory and CPU consumption were practically zero).

No, my "advertised" (misnomer in Atlas of course, should say "measured", caused much confusion on my side) bandwidth is NOT a small fraction of my real advertised bandwdith, it is about 50% of my advertised bandwidth.

No, my actual bandwidth is not just a 2-3 of times less than that measured and reported in Atlas, like in your case. In my case it is 160 [HUNDRED AND SIXTY] times less. Here is how I calculated it: my Atlas "advertised" bandwidth is 100 KB/s (=800 kbit/s). Every 6 hours my relay sends about 14 MB (as reported in heartbeats in the log). Therefore my actual average bandwidth utilization is 5 kbit/s.

No, changed IP usually does NOT mean changed network. It usually means dynamic IP which has nothing to do with changes in the network or its performance, or stolen keys.

In short, if Tor Project does not want relays with dynamic IP, it should say so and I would stop wasting my time. Otherwise, Tor should fix what's broken. There are 7000 relays total. Do you know how many Raspberry Pis are out there? Many, many times more, many of them run by privacy enthusiasts with dynamic IP. Tor is flushing them all down the drain but STATES that it wants relays with dynamic IP, too (I saw it somewhere on official Tor Project pages).