Summary
Roadmap item (README "Next": "additional lenses"). The architecture's core bet — one ordered key-value kernel, thin model lenses on top — is proven by the v1 trio (kv -> document -> relational). This exploratory issue tracks which lens, if any, earns the next slot.
Candidates (none committed)
- Time-series / append-log lens — the kernel's ordered byte keys make time-prefixed range scans natural; pairs well with the WAL-is-the-database model. Likely the strongest fit.
- Graph lens — adjacency as composite keys (
edge:<from>:<to>); traversals are prefix scans. Honest question: does it earn its comprehension budget vs documenting the pattern in a guide?
- Queue lens — ordered keys + atomic transact give exactly-once-per-consumer semantics in-process; small and teachable.
- Search/inverted-index lens — overlaps heavily with the secondary-indexes item; probably a capability there, not a lens.
Bar for admission (from the manifesto)
A new lens must (1) add a genuinely different mental model, not sugar over an existing lens; (2) stay a pure view over transact() — no kernel changes; (3) register honestly in the catalog so Studio renders it faithfully; (4) justify its lines the way document and relational did.
Deliverable
A DESIGN.md addendum ratifying at most one next lens (or explicitly ratifying "none for now"), before any implementation.
Summary
Roadmap item (README "Next": "additional lenses"). The architecture's core bet — one ordered key-value kernel, thin model lenses on top — is proven by the v1 trio (kv -> document -> relational). This exploratory issue tracks which lens, if any, earns the next slot.
Candidates (none committed)
edge:<from>:<to>); traversals are prefix scans. Honest question: does it earn its comprehension budget vs documenting the pattern in a guide?Bar for admission (from the manifesto)
A new lens must (1) add a genuinely different mental model, not sugar over an existing lens; (2) stay a pure view over
transact()— no kernel changes; (3) register honestly in the catalog so Studio renders it faithfully; (4) justify its lines the way document and relational did.Deliverable
A DESIGN.md addendum ratifying at most one next lens (or explicitly ratifying "none for now"), before any implementation.