Planning a decentralized systems class

[nicola]

What should be in a "introduction to decentralized systems"?
Please chip in and help me listing content or papers down!
This needs a lot of cleaning up and reorganizing, looking forward to hear your comments!

• Defining Decentralization
  • systems: from distributed to decentralized
  • dimensions: privacy, trust, anonymity, fairness, interoperability
  • networks: p2p, federated, centralized
• Data layer
  • Encrypted data
  • Authenticated Data Structures (tree based: Merkle Trees, Verena, non-tree-based: accumulators)
  • Public verifiability (CONIKS)
  • Practice: Naming things Part 1 (CAS, SFS, Certificate Transparency)
• Communication/Content distribution layer/Network topology
  • Structured (DHTs, Chord, Kademlia, Bittorrent DHT, IPFS)
  • Unstructured (Gossip, Cyclon, Vicinity, Sub2Sub)
  • Practice: Mesh networks, Pubsub
• Ordering of events
  • Causal consistency (Lamport clocks)
  • Consensus with Strong consistency (PBFT)
  • Consensus with Eventual Consistency (Proof-of-Work)
  • Practice: Replicated State Machines (Ordering, Transactions, DPKI)
  • Practice: Naming things Part 2 (Blockstack, Namecoin)
• Computation layer
  • Delegating computation
  • Multi-party computation
  • "Smart contracts"
    • Computational sound proofs/systems (ZK, ADS, SNARKS, Garbled Circuits)
    • Incentive based proofs/systems (Anytrust, Proof-of-Work, Proof-of-Stake)
  • Practice: ?
• Semantic layer
  • Interoperability: Web, Standards, Portability
  • Linked Data (RDF, Solid, Semantic Web, IPLD)
  • Practice: Decentralized Social Network
• Decentralized systems and society
  • Game Theory
  • Regulations and Data Protection acts
  • Social decentralization
    • (identity, social networks)
    • Centralization on the Web (from blogs to facebook)
    • Reputation and Trust models
  • Applications: Tamper-evident, Censorship resistance

[melvincarvalho]

Decentralization is never absolute. It is shades of gray. The trick is to know where your points of centralization lie, and what tradeoffs that offer you in terms of restrictions and functionality.

Once you can establish what your points of centralization are, what restrictions that gives you, you can determine what you can interopate with, either unidirectionally or bidirectionally.

If this exercise is not undertaken, you are left with a system that is unlikely to be interoperable, and therefore, relies on organic growth. This leads to a high failure rate.

[nicola]

Hey @melvincarvalho I think this should be covered in the very first chapter (Defining decentralization), later on in the Semantic Layer and finally in the last chapter. How does that sound?

[narula]

@nicola I think we should consider teaching something like this over IAP!

[nicola]

@narula, this was originally meant to be last year's IAP class. Organizing this might be too late now (especially given the other work that I/we plan to do over IAP).

As a reference class for this one: http://introtodeeplearning.com