docs: add section on nodes and actors

[LHerskind]

Added a section on nodes and actors.

[benesjan]

Nice writeup 👍

[PhilWindle]

I don't think we would bundle a light node on a user's device. At the top you mention that all nodes would connect to the p2p network. We wouldn't have user's devices do this. The PXE on their device would need to call a remote node.

[LHerskind]

Light client would only need to get the header, which for our case could use the P2P similar to what happens for most other blockchains or you could abuse L1 as your p2p layer for getting the header. The light client is there to provide you with stronger guarantees around the header than blind trust.

Bundling it is a way of getting users into using more secure structures. And allows us to have the logic for getting a hold of your notes at the node. In the case that you connect to your own full node, you can simply get it to return values from its local db based on whatever it checked with tags etc. And if light-node the lightnode can call out to services that might exists and validate the returned values against its headers for anchoring. That way the PXE don't need to deal with figuring out how to get the data beyond "just ask the node", and node can then implement what it desires. Similarly to what you can do for things like eth, the PXE or wallet don't care what type of node it is connected to, just that the node provides the expected answers.

[iAmMichaelConnor]

I think Phil's original point is valid, though, right? I think there are two categories here: a user who only runs a PXE and has to query everything about the blockchain from a 3rd-party server; and a user who runs a PXE & light node on their device. I do think both types of user will exist.

[benesjan]

@iAmMichaelConnor the desirable end-game of blockchains is that everyone runs a validating light clients and I feel like given that we only need to fetch the headers and that we will have a much lower block frequency then Ethereum then it actually might be feasible to aim for having users use a validating light client (of Aztec) always, no?

[iAmMichaelConnor]

It would be nice if everyone ran a node. But pragmatically, I don't think that'll happen, so recongnising this class of user seems sensible to me.
If I'm using aztec on my phone, do I want to drain my battery running a light node, or do I want to run the bare minimum software, and make requests to a node only when I need to? I might even be running a full node at home, which I can make requests to. But a resource constrained device doesn't want to stay in sync with the blockchain, I don't think?

[benesjan]

Makes sense to include it but it would be nice if it turns out that the e.g. the batter drain is so low (which feels like it might be the case 🤔) that we can just make it the default.

[PhilWindle]

I suppose I'm not entirely sure what we envisage a 'light' node doing. In my mind, the minimum amount of functionality for something to be called a 'node' is:

1. Providing world state (hash paths etc).
2. Proving block headers.
3. Providing access to the tx pool's p2p network for transaction submission.

You would never run the above on a mobile device. The p2p chatter and hashing would sap your battery and consume all of your bandwidth.

If what's required is merely a way to independently verify block headers so that I know the node I am connecting to isn't lying to me can that be achieved by configuring the PXE (which is on a user's mobile) to connect to a trusted L1 node and query that information directly?

[benesjan]

In general light nodes are nodes which have the necessary information to verify that they are not being lied to but don't store any other info. Both on Ethereum and in our case it would be just fetching and hashing the headers. By hashing the headers and hashing the archive tree users get the archive root which they then can verify against the Rollup contract on L1.

[LHerskind]

Updated document to hopefully answer this more ahead of time.

[PhilWindle]

A couple of comments

[sklppy88]

Pretty clear overall ! Maybe some super nit-picky things to look at but regardless non-critical

accident

[rahul-kothari]

LGTM but pls settle on the discussion w phil re light nodes + pxe

[iAmMichaelConnor]

I think Phil's original point is valid, though, right? I think there are two categories here: a user who only runs a PXE and has to query everything about the blockchain from a 3rd-party server; and a user who runs a PXE & light node on their device. I do think both types of user will exist.

[LHerskind]

Current dependencies on/for this PR:

• master
  • PR docs: add section on nodes and actors #3975 👈

This stack of pull requests is managed by Graphite.

[benesjan]

Looks good to me. Just some english fixes

[LHerskind]

Added a few answers to some questions, will update document tomorrow.

[LHerskind]

Altered the format a little to try keeping one sentence per line for easier suggestions and smaller diffs.

[PhilWindle]

I'm still unsure of the purpose of a 'light' node. With regards to the p2p network, I wasn't planning on broadcasting the block headers, nodes will just read them from L1 which is the source of truth. The p2p network only needs to propagate transactions.

[LHerskind]

Light nodes are to not blindly trust something like infura.
The section is not worrying around the specific type of blockchain, but general.

But since the l1 contract we have is a validating light node, you can use it as you p2p connection.

Slow re-review.