feat(mesh): volunteer-mesh superset design + Needs/Wants placement firewall - #26
Merged
Conversation
…rewall
Synthesizes the volunteer-computing corpus (13 docs: Dual-Orchestration, Agent-First Node,
Urbit/CoreOS, BOINC OOBE, CluBORun + CMS@home DataBridge, SETI@home motivation, Pando,
Radicle/Wikipedia governance, Science United UX, Needs/Wants firewall, 2019 State-of-VC) into
one design and ships the crispest primitive from it.
docs/VOLUNTEER_MESH.md — the superset design. North star: Folding@home hit top-5 supercomputer
on 400k volunteers; the constraint is participation + trust, not silicon. Seven moves that make
the mesh more elegant than a scheduler-centric anchor:
1. participation is the system (intent enrollment, close the contribution->outcome loop,
cooperative teams default) — the scheduler is secondary.
2. pull/lease streaming dispatch (Pando StreamLender/Limiter), quorum as a stakes×reputation
overlay, not always-on byzantine.
3. split control/data planes; "done" = data durably landed (CMS@home signed-URL DataBridge +
verify-download barrier); data gravity/RTT first-class.
4. trust by earned reputation + soft security + cooperative games, not stake/crypto; credits
additive-only, never slashable.
5. every signal an instrument with a refusal mode; every policy a logged gate.
6. a Needs/Wants placement firewall.
7. gate the incentive loop vs Goodhart; domain-scoped reputation.
Needs/Wants firewall (move 6) — SHIPPED: compute_plane gains BACKEND_CAPS (attested capabilities
per backend) + place(workload.needs=...): a hard Need (tee/fips/residency/no_egress) may only be
satisfied by a backend that PROVABLY provides it — a soft Want (prefer/nearby/cheap) can never
masquerade as a Need. Fail-closed: no backend provably provides the need -> blocked.
Tests: +4 needs-firewall = 122 tools tests green. Remaining moves (pull/lease scheduler, DataBridge
data plane, instrumented signals, domain reputation, participation on-ramp) named in the doc.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Synthesizes the 13-doc volunteer-computing corpus into one design (
docs/VOLUNTEER_MESH.md) and ships its crispest primitive.North star: Folding@home hit top-5 supercomputer on 400k volunteers — the constraint is participation + trust, not silicon. The corpus converges on seven moves that make the mesh more elegant than a scheduler-centric anchor: (1) participation is the system; (2) pull/lease streaming dispatch, quorum as a stakes×reputation overlay; (3) split control/data planes, "done"=data durably landed (CMS@home DataBridge); (4) earned-reputation + soft-security trust, additive-only credits (no stake/slash); (5) instrumented signals + logged gates; (6) Needs/Wants placement firewall; (7) Goodhart-gated incentives + domain-scoped reputation.
Shipped this PR (move 6):
compute_planegainsBACKEND_CAPS(attested capabilities per backend) +place(workload.needs=…). A hard Need (tee/fips/residency/no_egress) may only be satisfied by a backend that provably provides it — a soft Want (prefer/nearby/cheap) can never masquerade as a Need. Fail-closed. (Verified: a TEE-needing workload lands on HPC even when k8s is preferred.)122 tools tests green. The other moves (pull/lease scheduler, DataBridge data plane, instrumented signals, domain reputation, participation on-ramp) are named in the doc as the next builds — with
work_unit.py's verification plane (#25) already covering result-trust.