design

Status: proposed (2026-10-17, Claude), from Daniel’s message of 2026-10-17. Related: SPEC-001 Sophia Store v0, Requirements SR-18, SR-21; Moonkale ADR-0028 §3, WP-68.

Context

The earlier vault presents Sophia as a multi-language compiler whose headline is cross-language interoperability proved by equivalence edges. Its roadmap starts with Metamath (M0) and a Lisp (M1), and puts Julia precompilation (M3) at the point “where the project justifies itself”.

Daniel’s statement of 2026-10-17 gives a different reason to continue. Moonkale needs a specification, and designing Moonkale against a Unison stand-in is hard. So Sophia should make a real first attempt at the database compiler. Moonkale’s ADR-0028 deferred “where Sophia’s definitions live” until real definitions exist, and named a base Sophia (WP-68) as the way to get some.

Decision

Sophia’s first goal is to be the store-backed code source that Moonkale edits:

  • definitions in a database;
  • content identity;
  • names;
  • printing and parsing;
  • a change feed;
  • queries.

The compiler, the checkers and interoperability are layers that come later, hung off the same identities. Their structure is fixed in v0 only far enough not to block them.

Consequences

  • The earlier roadmap’s M0 to M7 are no longer the plan. They remain a list of ideas (Reading of the Earlier Vault).
  • Sophia’s v0 language is chosen for what Moonkale has to show (functions, data types, relations, rules, holes, tests), not for proving the plumbing on Metamath.
  • Moonkale WP-18 (a Unison source) stays. Daniel, 2026-10-17: keep both Unison and Sophia (SR-23). Two independent store-backed sources are a stronger test of Moonkale’s schema than either one alone.