What is the intended networking model for BEEFY in JAM?

[dimartiro]

Question

What is the intended networking model for BEEFY in JAM?

More specifically, should BEEFY commitment collection and aggregation be treated as an internal validator-side mechanism, with the resulting data exposed to light clients through RPC or other higher-level interfaces?

Or is there an intended JAMNP-S protocol/stream for validator-to-validator BEEFY commitment distribution that is just not documented yet?

Why I'm asking

In our implementation, we are wiring BEEFY commitment distribution between validators so commitments can be collected and aggregated after GRANDPA finalization.

However, in the current public JAMNP-S docs, I could not find a canonical stream kind or protocol definition for BEEFY commitment distribution.

So I want to clarify which of these is the intended direction:

• validator-internal BEEFY handling, with outputs exposed externally via RPC
• a dedicated JAMNP-S protocol for validator-to-validator BEEFY commitment distribution
• something else

[ggwpez]

This kind of discussion fits well into the matrix channels mentioned here: https://docs.jamcha.in/#matrix-channels
AFAIK the networking protocol is kept minimal for now and will be reworked to account for all other protocol things.

But how did you implement BEEFY and GRANDPA already? I think the Gray Paper does not sufficiently explain them in detail.

[dimartiro]

Thanks — happy to move this to Matrix.

Quick summary of what we have in go-jam:

GRANDPA

Implemented following the Gray Paper + the JAMNP-S spec: CE149 votes, CE150 commits, CE151, CE152.

BEEFY

The Gray Paper and the spec describe the BEEFY mechanics and the RPC surface, but not the validator-to-validator
networking. So we only implemented what is actually specified.

For distribution we kept an internal BeefyBroadcaster interface as a placeholder, so we can plug in whatever JAMNP-S
ends up specifying — or keep it validator-internal with only the aggregate exposed, if that is the intended model. That's what I want to confirm