RPC WG Meeting - 2024 June 10

[bumblefudge]

• agenda recap/updates
  • requiredScopes - still aligned
  • connection metadata - still tinkering/designing
    • new CAIP?
    • Shane: sessionProperties.{scopename} is wierd devEx, since we explicitly do NOT want any sessionProps that aren't specific to one scope
  • more explicit lifecycle RPC calls --> new CAIP(s)
    • internal prelim draft here
  • getPermissions (wallet caps) versus hybrid/onchain permissions - need to be ready for multiple possible futures and diff permissioning models
    • one of the complications--in some proposed hybrid architectures, EOA could be one of many access points, or EOA could be a mandatory in-the-loop interface? radically diff assumptions/semantics for permissioning across those two types, and best not to deanon wallet by expressing that distinction to every dapp
    • medium-term!
  • wildcarded scope
    • bumble still owes a PR!
    • wallet-scoped method versus namespace-scoped methods - where to put wallet_ functions that aren't really scopable to a chain but are only safe to deploy in one namespace
    • Shane: personal_sign is a good example of a namespace-specific method that could have collisions in other namespaces
    • bumble: what about the offchain CAIP2 we just quietly added 2 months ago? here
    • jiexi: what about ::, null-chainId?
      • bumble: but URN?
    • alex: other namespaces will likely need this too, it shouldn't be a "special case" just for EVM? update CAIP-2 to consider this?
      • bumble: I can write that PR
      • alex: most urgent implementation question for us is how this works in EVM- we'll have a think on "chainId 0" approach
  • eip155 rename to eth namespace discussion - maybe need to timebox or defer? bip122 versus bitcoin also?
  • sessionId optionality
    • bumble: still easier to think through the consequences if there's a fleshed-out alternative with explicit methods as fallback/failover
• agenda new items
• homework, takeaways, next steps
  • bumble needs to figure out how splitting sessionProp into sessionProp and scopeMetadata could work spec-wise (backwards compat?) and implementation-wise
  • MM team will think through internally the possibility of eip155:0 as offchain/wallet_ scope object
    • hassan: but why not eip155:offchain? alex: shouldn't it break validation anyways? if it's the same reference across all namespaces that might be a cool property
  • bumble needs to review alex's new CAIP before next meeting
  • bumble needs to update CAIP-2 and/or write net-new CAIP about offchain authorizations per namespaces
  • bumble owes a PR to put at least a SHOULD against treating a namespace-wide CAIP217 with empty/no chains array as a wildcard
    • noticed another semantic weakness that should be tightened up while i was in there!