Very similar to Freenet. A few interesting different choices. This is just my first impressions from a talk; I haven't investigated in detail, please reply if you have!
1. A proper DHT allows them to solve the dynamic content / updatable content problem. The cluster of nodes closest to a key maintain permanent connections to each other, so they can run consensus algorithms. IMHO the best way to do this on Freenet is for updatable keys (TUKs?) to be only updatable on one node, but then duplicate across several locations in such a way that we can heal them all. 2. Economics allows them to implement permanent storage. Moving content around when nodes go down (or in Freenet when they swap locations) must be very costly. In any case I don't think they have an economy right now. Updatable content in principle allows them to build one, but I don't think basic mutable content is working at present. The rest is very similar. They "telescope" requests, i.e. relay both the request and the data, in much the same way as we do. They have CHKs, I think updatable keys are similar to SSKs, and KSKs (which could be spam-proofed by economics in future). Some things are much easier due to it being a proper DHT. They use client-side data encryption for plausible deniability (and privacy for non-public documents). They use chunking. Telescoping is supposed to provide some privacy, which is even less credible for them than it is for Freenet; routing is entirely predictable! They do not solve the Sybil problem. Internal payment systems cannot solve the Sybil problem because the bad guys can always run nodes at a slight loss - and the more of the network they control, the cheaper this gets, since they can cheat internally as well as exploit economies of scale. Especially as they can then undermine the internal currency system - creating a financial incentive to control as much of the network as possible. I don't think the Sybil problem can be solved. I didn't stay for all the Q&A, so I don't know whether they have an OMG-I-could-have-encrypted-child-abuse-images-on-my-PC problem yet. They are more open to embedding in websites, javascript, etc, which we generally reject for reasons of security; this is simply a choice - privacy or dynamic content, you can't (easily) have both. It is not clear what the plan is to solve the wider economic problem: the internet is screwed because people don't pay for anything, therefore we seek alternative revenue streams, i.e. owning everyone's data. There is some discussion of micro-clients. E.g. phones can be clients but not vaults. This means they need to get credits from somewhere else, e.g. your home vault. On Freenet, mobile nodes don't make sense, but you can connect to your home node. The other case is browsers - embedding a distributed chat system in a regular website for feedback. Again, where do you get the credits from? They don't seem to have any publish/subscribe or ULPR type stuff yet, so I seriously doubt that chat systems based on this will give acceptable performance. However that is certainly feasible, and there may be some better options than what we use on Freenet if they have currency working. -------------- next part -------------- A non-text attachment was scrubbed... Name: signature.asc Type: application/pgp-signature Size: 819 bytes Desc: OpenPGP digital signature URL: <https://emu.freenetproject.org/pipermail/tech/attachments/20170205/6eca7ec4/attachment-0001.sig>
