Hej Eldalië,

thanks for your answer.

Tor project runs campaigns for running more bridges [1-3] because there was/is (?) a shortage and/or countries changed their censorship policy (e. g. Russia). So I'd say the statement in the FAQ is a little bit to simple. At different times a shortage of one or the other node might occur. :)

I think the answer might also depend on what kind of servers you are able to run. E. g. a server in a cloud/with a hoster where you can spin instances up/down on a regular basis (weekly/monthly?) and change the IP in this process, might be good for bridges in case they got blocked by a censor (no idea on how long a usual bridge is usable before it gets blocked in parts of the world). If you are able to run the node for many years with the same IP addressing, maybe a non-bridge would be better, so the node might become a fallback-authority node one day.

If you don't use HiddenServiceSingleHopMode for you onion services, six tor nodes are in between the visitor and the server which results in a lot of bandwidth usage in the tor network. So some non-exits might also be handy in case onion service usage picks up.

If you take a look at https://metrics.torproject.org/bandwidth-flags.html it seems to me like there is much more advertised bandwidth than consumed bandwidth for all flags. So maybe more bridges is the answer?

But I really cannot tell because I don't have enough insights and so I just went with bridges (time to be blocked for a usual bridge would be really interesting to know if/when I should convert the bridges to regular nodes or change IP addressing).

Thanks and have a great weekend!

Fran

[1] https://blog.torproject.org/run-tor-bridges-defend-open-internet/ [2] https://blog.torproject.org/run-a-bridge-campaign/ [3] https://blog.torproject.org/wrapping-up-bridges-campaign/

On 2/4/22 10:40, Eldalië via tor-relays wrote:

Hello there.

...

And generally, what's more needed atm: bridge, non-exit, exit or snowflake?

I'm not an expert at all, but the most needed type of relay is exit ([1]) and I suppose always will be, as they are the most prone to raise legal issues ([1]) and have the highest requirements ([2]). Also, to my understanding, if there were many exit nodes but guard or middle were missing, exit nodes would be used for the first two hops as well, so if you have enough resources and you will to I'd say you should go with an exit node. As for the question about bridges I'm a newbie, so I'll leave the answer to someone with some expertise :).

Eldalië

Quoting Fran via tor-relays (2022-02-04 18:34:45)

I don't know how bridges with obfs4 vs snowflake work in hiding that they provide access to the tor network (where is the difference?). I assume that there is a difference, because if not, why bother with another approach. So providing these two different access methods would allow people to connect with any of the two methods. If the IP addresses would be blocked both wouldn't work anymore, but till then more options would be available.

If IP addresses wouldn't be limited giving both methods individual IP addresses would of course be better. I can do this easily for v6, even out of another /64 but for v4...well...

But would running bridge and snowflake on the same v4 do any "harm"? For example like bridges are easier to be detected and only running snowflake might result in a longer usability of the node.

Yes, there are many differencies. snowflake does make the traffic look like webrtc (like a video conference) and obfs4 makes the traffic look like random noise. Also the clients use different mechanisms to discover the relays.

If you run both in the same IP address and the censor has a way to discover one but not the other both of them will be blocked at once. So you are making it easier for the censor to discover them and block them. That is why we don't want people to run both in the same IP address.