Scope
Scope of the B2 milestone "Hybrid Hauling"
Research stage
B2 is in the research/design stage. This scope is a working draft, not a committed plan; it will firm up as the link contract and experiments produce numbers.
This milestone inherits goals, constraints, and no-goals from C1. Manual Mesh and assumes the control plane direction sketched in B1's Coordinator. For how B2's pieces fit together, continue with the Design Sketch.
Goals
Use Case
Edge Mesh: a rete of mobile/embedded nodes (robots, drones, vehicles) connected over heterogeneous links — where no single medium covers all pairs, and links appear and disappear with mobility.
The B2 prototype must demonstrate:
- A working mesh over at least two dissimilar link types simultaneously (reference target: ad hoc WiFi + Ethernet), with multi-hop forwarding across mixed hops — exercising C1's parallel-link-vertices design under real radio conditions.
- Link-quality-aware behavior: the mesh survives a degrading/disappearing radio link by failing over to an allowed alternative path.
- A minimal, operational Radio Resource Allocation (RRA) — the first Coordinator RRA extension, targeting the contention medium itself: the smallest radio-resource control that is still useful (working candidate: centrally computed per-node TX-power caps within operator-signed regulatory bounds, improving spatial reuse and mitigating hidden terminals). This completes the control-plane roadmap: C0 delivers mgmt only → C1 adds the minimal ctrl stream → B2 adds the first RRA extension.
- The link contract: a published, vendor-neutral statement of what any L1/L2 must provide to carry a Florete link — validated against WiFi as the reference CSMA medium.
Radio-vendor readiness
B2 makes Florete integrable by radio/hardware vendors:
- The link contract is public and testable (each requirement has a number or a validation procedure, not just prose).
- A custom-radio link integrates as a 3rd-party link adapter per C1's connection-manager design — no changes to mesh-flor required.
- The RRA-placement design states where resource allocation lives for each MAC family (node fast-path vs. Coordinator extension), including the single-rank/decentralizability constraint on scheduled MACs.
Development Constraints
Same as C0/C1 (Rust, Tokio, cross-platform), plus: the reference prototype must run on single-board computers (ARMv7+, Linux) with commodity WiFi adapters.
No-Goals
- No custom PHY in the reference scope. Vendor-specific radio integration is a variant layered on the same contract, not part of the open reference.
- Single host, omnidirectional radios. A B2 node is one computing host with directly attached omni, non-beam-steering radios. Multi-host/multi-radio carriers (supernodes) and directional/beam-steering (phased-array) antennas are the evolution axes the contract and RRA are shaped to extend into — see Radio MACs > Evolution.
- No scheduler. B2's minimal RRA targets contention MACs; scheduled-MAC machinery (allocator interface, semi-persistent grants, slot maps) stays design-only (Radio MACs) and is pulled in by a concrete vendor integration.
- No asymmetric/unidirectional links. Supported by the architecture (directed vertex graph); B2 neither implements them nor makes decisions that would foreclose them.
- No coordinator-less/gossip mode. Churn beyond Coordinator convergence is post-B2.