Join GitHub today
Refactor key management #3296
This PR is turning into a major refactoring over the crypto and key management system, which seems both long overdue and somewhat inescapable.
The main issues that this PR aims to solve are:
I would also like to solve the fact that service construction is highly monolithic and polluted by consensus-module-specific code. Really it should just be a case of specifying the block production and finality systems and letting it do the rest. Code for set-up, take-down, RPC and key-store integration should be provided through traits that the consensus modules (Babe, Aura, Grandpa and eventually Rho and PoW) implement within their module's scope. There shouldn't be any consensus-module-specific logic dirtying up
This also adds:
gnunicorn left a comment
A few remarks, but nothing major.
Note: I only glanced over the macros and superficially looked at the SRML changes, as I missing the context to understand the design behind the changes - but nothing bad jumped out.