Skip to content

ROADMAP.md

Latest commit

 

History

History
65 lines (51 loc) · 2.44 KB

ROADMAP.md

File metadata and controls

65 lines (51 loc) · 2.44 KB
Raw

PhaseLock Roadmap

PhaseLock is currently in Alpha. All the core features are implemented, a few of them are tested, and none of them are properly documented.

There's still a lot to do:

  • Documentation for:

    • the TypeSpec API
    • the TypeSpec emitter configuration (per-language)
    • the generated APIs (per language)
    • recommended backend architecture(s)

  • Support aliases in TypeSpec... today they cause confusing/broken generated code.

  • Support official KurrentDB clients as inputs to Engine

    • reduce the IO boilerplate needed to populate events from KurrentDB
    • serve as canonical examples for the IO handling

  • Reimplement go and python bindings to match new cross-language plan:

    • No more JS wrapper types; native types should be fully-native
    • Native queries read native types directly from storage, no JS layer
    • Generate type registry per framework for handling conversions
    • StoreGenerator redesign around the type registry
    • Eliminate "plain-json object" as central type in Engine UI.

  • Support additional languages, starting with C#, Kotlin, and Swift

  • Additional frontend demos:

    • React Native
    • iOS
    • Android
    • PhaseLock engine in a Shared Worker serving multiple tabs

  • Additional backend demos:

    • Pure TypeScript backend for library demo
    • Cloudflare CDN-based log distribution
    • Cloudflare CDN-based server-side queries
      • Will require implementing a Store MVCC capability

  • Add indices as a formal concept in the PhaseLock Engine

    • possibly a plugin interface, for define-your-own index?
    • include a btree implementation at a minimum

  • Add support for time-based triggers

    • reducers currently only wake due to events
    • currently unable to be triggered by passage of time
    • user workaround is possible (Tick events and a query for wakeups)
    • ought to be first-class in PhaseLock
    • will require careful consideration: how will it interact with catchup and replay?

  • Performance testing

    • Test various backend designs for read/write throughput and round trip latency.
    • Bottleneck analysis and optimization.
    • How do different designs scale, especially on the read side?
    • Use this data to pick recommended backend architecture(s).

  • Investigate snapshot tooling

    • Snapshots can greatly reduce cold-start times
    • Snapshots may require per-user sanitization
    • How can we offer tooling to make this easier?

  • Investigate Kurrent integrations

    • Run as part of KurrentDB?
    • Run as part of kurrent.cloud?