The current architecture tightly couples the C ABI exports (extern "C") within the sdks/c crate, which also serves as the C SDK distribution (containing CMake configs, headers, and tests). This creates unnecessary friction when other language bindings (like Go) simply need to link against the core C ABI.

Goal:

Decouple the C ABI implementation into a standalone crate to facilitate easier integration for future language bindings and reduce code duplication.

Proposed Plan:

1. Create a Dedicated FFI Crate (sdks/ffi)

   • This crate will be a pure Rust library (cdylib / staticlib).
   • It will house all extern "C" function exports currently in sdks/c/src/ffi.rs.
   • It will be responsible for generating the raw dynamic/static library artifacts (libboxlite) and the C header (boxlite.h).

2. Refactor sdks/c as a Consumer

   • Transform sdks/c into a consumer-facing SDK package.
   • It will primarily contain curated C headers, CMake build scripts, pkg-config files, and C-language examples/tests.
   • It will link against the artifacts produced by sdks/boxlite-c-abi.

Reasoning:

By isolating the C ABI implementation into its own module (boxlite-c-abi), we create a clean, reusable foundation for any language that can interface with C (FFI). This eliminates the need for each SDK (like Go) to maintain its own redundant Rust bridge layer or depend on the full C SDK build system, streamlining the addition of future language support and preventing duplicated effort.