Change the name of the GrandPa notifications protocol

[tomaka]

At the moment, the GrandPa protocol is called /paritytech/grandpa/1. It's not great for two reasons:

1. It shouldn't contain the word paritytech.
2. Since the messages exchanged are specific to a certain chain, its name should contain the protocol ID.

Reason 2) will be important in the future, when nodes can be connected to multiple chains at the same time.

[tomaka]

The fix needs to involve a transition period where the protocol will have two names. As such, this depends on #6605.

[stale]

Hey, is anyone still working on this? Due to the inactivity this issue has been automatically marked as stale. It will be closed if no further activity occurs. Thank you for your contributions.

[tomaka]

Issue still relevant and important.

[acatangiu]

1. It shouldn't contain the word paritytech.

Using a new static /substrate/grandpa/1 protocol name is easy AFAICT:
just use new protocol name, and add old protocol name (/paritytech/grandpa/1) to fallback_names in sc_finality_grandpa::grandpa_peers_set_config()

2. Since the messages exchanged are specific to a certain chain, its name should contain the protocol ID

If we want to customize the grandpa protocol name to have chain-specific ID (e.g. /polkadot/grandpa/1), then we'd need the chain ID in places such as client/finality-grandpa/src/lib.rs::run_grandpa_voter().

Where is the chain ID specified, what can be used as a chain-specific ID?

[tomaka]

We have this thing called the protocolId in the chain specification. The sync and transaction protocols already use this protocolId in their name.

However we have a plan to use the genesis hash instead of the protocolId, and it probably makes more to transition Grandpa to use the genesis hash: #7746

[tomaka]

Using a new static /substrate/grandpa/1 protocol name is easy AFAICT:
just use new protocol name, and add old protocol name (/paritytech/grandpa/1) to fallback_names in sc_finality_grandpa::grandpa_peers_set_config()

Yes

[burdges]

Assuming this is not pressing, we might consider merging this transition with a switch to Rabin-Williams signatures or something else faster than ed25519. RustCrypto/RSA#118

This is based upon my understanding that grandpa spends a lot of CPU on signature verification.

[tomaka]

We can always transition again from /blablabla/grandpa/1 to /blablabla/grandpa/2 in the future.