cryptogen
—it reaches out it reaches out it reaches out it reaches out—one hundred and thirteen times a second, nothing answers and it reaches out.
— cibola burn
temporal oscillations expecting each other at repeating intervals. using memory to ask the right question at the right time. passing notes in-between classes.
goals / challenge
- develop a SCALABLE mobile adhoc network that can be used over a wide range of wireless bands and high-bandwidth transports including sneakernet
- deploy a (test)net on SALVAGED hardware
- connect a world-wide community of solarpunks on an independent low-infrastructure wireless inter-network
- define a transport abstraction that makes sense for overlay networking and physical broadcast links
- refine transport concepts so they can handle large binary objects as well as the plain text meta channels
- deploy physical mesh networks for testing
- deploy a virtual mesh to link the test networks over the internet
design principles
trivial technology
a network stack is going to be fairly complex, but it should be easy to understand. simple components that work together in a way that is easy to reason about.
long-term thinking
how do we roll out an independent network without creating a lot of waste?
how much of the existing hardware can be reused and modified to support information centric networking? how low can we make the minimum required specs? encryption has to be optional, but without signatures we can only work with self-certifying, static data. perhaps a simplified scheme can be devied that is suitable for transmitting short messages over human-mediated links without the overhead of cryptography.
walkaway net traffic can be tunnled over arbitrary transports including fully DIY wireless transport using homebrew SDR or your everyday sneakernet. the idea is to make anything that can hold or transmit information into a condiut for the network. this does require making some trades when it comes to worst-case latency.
privacy
there is some evidence that information-centric networks can make anonymous routing cheaper. these types of networks are potentially more private than IP to start with as well.
- modified global passive adversary attack (reduced cost via mesh network localization)
narrow waist
this diagram shows how ip and named data networking organize their various layers around a simple common layer. this is the common protocol that ties the network together, any changes at this layer get more expensive as the network grows AND represent a breaking change.
secure scuttlebutt (SecSB) uses the same EBT strategy for gossip as described above. one interesting shared detail about NDN and SecSB is that the file format of the logs are actually the common piece that connects the high level applications of the network to the actual dissemination and storage protocols. one important consequence of this is that data replication is not coupled to the gossip network or any other repliction strategy. there are some proposals around SecSB that intend to take advantage of this, but walkaway net will put this feature at the forefront.
the narrow waist must address some important concerns:
- how are messages interpreted?
- how are messages and their components validated?
- how can a sender be verified?
- how does the network evolve as the relative security of its primitives changes?
plain-text formats
in addition to the software design, the file formats we choose and/or design should be plain text and easy to parse. not only for the sake of writing software, but to enable humans to participate in the network.
specification
see also
posts
:cat ~/tag/cryptogen
replies
related
⁂
⁂