Den's fx-pipeline has grown to ~7,300 lines of pipeline infrastructure to coordinate a shared trampoline walk. Ned demonstrates that the same semantics can be expressed in ~277 lines using stream composition and a Cycle.js fixed-point. Rather than migrating Den's internals (carrying architectural debt), this spec proposes rebuilding the resolution engine from scratch on Ned's proven foundation, targeting ~1,000 lines of core implementation with full feature parity.
The user-facing API is an opportunity, not a constraint. Since the fx-pipeline branch hasn't shipped, we can fix known ergonomic issues rather than preserve them.
- Feature parity with Den's fx-pipeline branch — all 8 test suites must be expressible
- ~1,000 lines of resolution engine (vs ~7,300 today)