jrandom at i2p.net wrote: >Are we talking about the same Tor? You realize that only a small >portion of the Tor userbase runs ORs, right? And that ORs are the >ones publishing their server descriptors in the directory, right? > > To avoid partitioning attacks on client knowledge, every client must regularly obtain lists of ORs from every trusted directory server (last time I checked there were three). This means the directory servers have an up-to-date list of clients. If you wanted to perform intersection attacks, you'd only have to eavesdrop on one of the servers. If you wanted to bring down the network, you'd only have to compromise two of the servers.
>* Anonymity. Running on top of I2P would allow Freenet users to blend > in with other I2P users. A less segmented anonymous user base > offers a greater anonymity set. > > This is a good argument. On the other hand, moving users from a more-secure network to a less-secure network doesn't do them any favours, so we need a proper analysis of each network's strengths and weaknesses before we start talking about consolidation. > If Freenet truly were better suited than > I2P as a comm layer, it would be sound software engineering to > build, deploy, and refine a comm layer and then, on top, build a > censorship resistant data store, rather than doing them all at once > in one big go. > > The OSI model is not the be-all and end-all of software engineering. Data stores don't have to sit on top of reliable streams which sit on top of unreliable datagrams. Personally I'm starting to think it makes sense to have two "layer 3" services: a packet delivery service and a content-addressable block cache. Both can be implemented by simple-minded routers and used by a wide range of higher-layer services, obeying the end-to-end principle, and a distributed cache has advantages for anonymity as we've discussed before. Cheers, Michael
