Skip to content

Congestion control v2 (design discussion first) #3

Description

@lgoyal6

v1 uses a fixed send window (config.h window_pkts, DESIGN.md §5.8 defers congestion control on purpose: the target is small urgent messages on closed meshes). For lossy links that are congestion-related rather than radio-related, a fixed window is either wasteful or unfair.

Looking for design input before code: something minimal and predictable (AIMD on the window, or a BBR-lite pacing rate from the existing RTT estimator) that preserves determinism under SimNet. Proposals welcome as comments here; a SimNet-reproducible experiment beats an opinion.

Metadata

Metadata

Assignees

No one assigned

    Labels

    help-wantedBigger surface, design input welcome

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions