4 ms·

I didn't know that. Is anyone working on an improved version? It seems like a great idea in principle.

nwh 13y ago
There's room for improvement, but there's a lot of core issues that just can't be flushed out. The concept of the POW is to stop spam, as all addresses are inevitably public, it doesn't really work though, as spammers typically have access to botnets which can spam all day long. Normal users just have to wait minutes to send a message. The scale issue is a weird one, they plan to split the network into different "streams" with different address types, which just sort of muddles the entire setup. Ultimately the limit is how much CPU and bandwidth the network can survive with while mirroring the entire content of the entire network. There's lots of problems with timing attacks that have been "resolved" with random sleeps, though nobody is really convinced of that too much.
- sirsar 13y ago
  >The concept of the POW is to stop spam Your other points are good, but this one is wrong. Bitmessage is currently bundled with a client because that makes adoption easier, but ultimately, Bitmessage is first and foremost a protocol. POW increases the cost of flooding attacks on the network. Clients like Thunderbird (it was easier for me to integrate Bitmessage with Thunderbird than my regular email provider) stop spam. The scaling is a rather hard problem, since, for anonymity, "everyone gets everything." If you have any ideas about how to scale a network like that, OR have any comparable methods for hindering traffic analysis, you should publicize them. The timing attack mitigation via sleeps /is/ a rather ineffective substitute for constant-time decryption. In summary: 1. POW is a non-issue, and part of the design at least the way you've put it. 2. Scaling is inevitable given the tradeoffs being made, unless you have a better idea, for which I will pay money 3. Timing attacks are a temporary problem, but they can certainly be "flushed out."
banachtarski 13y ago
It's not a great idea in principle at all. It should have been obvious that it wouldn't scale from the beginning by design.

nwh 13y ago

There's room for improvement, but there's a lot of core issues that just can't be flushed out. The concept of the POW is to stop spam, as all addresses are inevitably public, it doesn't really work though, as spammers typically have access to botnets which can spam all day long. Normal users just have to wait minutes to send a message. The scale issue is a weird one, they plan to split the network into different "streams" with different address types, which just sort of muddles the entire setup. Ultimately the limit is how much CPU and bandwidth the network can survive with while mirroring the entire content of the entire network. There's lots of problems with timing attacks that have been "resolved" with random sleeps, though nobody is really convinced of that too much.

sirsar 13y ago
>The concept of the POW is to stop spam Your other points are good, but this one is wrong. Bitmessage is currently bundled with a client because that makes adoption easier, but ultimately, Bitmessage is first and foremost a protocol. POW increases the cost of flooding attacks on the network. Clients like Thunderbird (it was easier for me to integrate Bitmessage with Thunderbird than my regular email provider) stop spam. The scaling is a rather hard problem, since, for anonymity, "everyone gets everything." If you have any ideas about how to scale a network like that, OR have any comparable methods for hindering traffic analysis, you should publicize them. The timing attack mitigation via sleeps /is/ a rather ineffective substitute for constant-time decryption. In summary: 1. POW is a non-issue, and part of the design at least the way you've put it. 2. Scaling is inevitable given the tradeoffs being made, unless you have a better idea, for which I will pay money 3. Timing attacks are a temporary problem, but they can certainly be "flushed out."

banachtarski 13y ago

It's not a great idea in principle at all. It should have been obvious that it wouldn't scale from the beginning by design.