DAS nodes (tracking issue)

[liamsi]

DAS node types

Now that the MVP DAS light-client is out of the door, we should focus on getting the other node types specified, decide on a few networking and p2p-related questions and actually implement all other node types.

Experiments:

This section needs further discussion and clarification

• Further clarify performance questions around dht and ipfs usage.
  • IPFS performance improvments and experiments #387,
  • Investigate initial timeouts on DHT / DAS #377

Lazyledger-core / actual DAS fullnode:

More on Experiments:

• Implement a script/tool to spin up a primitive test net with DHT bootstrap nodes #376
• Run an experiment using a DHT bootstrap node with a DAS light client #382

Required ADRs, decisions, and components:

• ADR / design doc around different options for block propagation: ADR: Block Data Propagation #389 (related discussion: Block Data Gossiping #384, and previous explorative work adr001 under the assumption that validators also do DAS)
• Block-Sync
• Fast-Sync (almost no changes necessary)
• Everything else around DAS nodes (UX etc), and different kinds of storage nodes:
  • ADR: DAS node-types aka partial nodes #388
  • decide on libp2p vs tendermint p2p for the non-consensus nodes
• Decide on blockchain: propagate headers & DAHeaders #191
• Decide on: rpc: request block data through IPLD #189
• Decide on: Store: only store header + DA header #182

Implementation:

• Make consensus messages DA ready #181 (we won't need to change consensus messages unless it's necessary for changing block propagation)

Chores:

• Issue templates to improve communication: Update issue templates #219

Nice to haves:

• Move and refactor the PutBlock method #196

[musalbas]

Superseded by #491

Should this be closed?

[liamsi]

Yes, let me go through the few open issues and repriotize them, or move them over to #491, or, close them (if no longer relevant).

Not sure how to handle the spec-related ones at the top and the more organizational / work-flow related ones (see the unchecked boxes). Ideally, they should also be prioritized and be part of some tracking issue even if they do not block the dev-net (@adlerjohn @musalbas).

[liamsi]

OK, moved them over to #491 as well.