task_100926_12 — P3 Wiring Completion: Staking Enforcement, Treasury Forwarding, Fee Floor (Robbert)

[umermjd11]

cc @robertocarlous
Assigned. Full spec: Developer/tasks/task_100926_12.md

The task file has the full specification — exact file/line references, per-item background, and the deliverables checklist. This post is just the summary, assignment notice, and the open questions that need an answer before (or while) you build.

Summary

Comes out of a 2026-09-10 audit of ten open P3 issues (#36-43, #75, #78) — full audit is in both this task file and its sibling, task_100926_13.md. This task is the half where the design decision is already made and only the contract wiring is missing:

• P3 staking: DAO-settable floor, per-model stake requirements, concurrent-registration cap, withdrawal queue #37 (staking): modelMinStakeBounds/maxConcurrentRegistrationsPerStakeUnit have existed as inert storage on DinValidatorStake since Feat/p3 staking #65 — nothing reads them at GI registration. Wire both into registerDINaggregator/registerDINAuditor.
• P3 slashing: conditions taxonomy, penalty tiers, slashed-stake destination, dispute resolution (P3-4.1/4.2/4.3) #38 (partial) + P3 fees & treasury: fee lines, fee-router contract, DIN denomination (P3-5.3) #43 (fees/treasury): three live TODO(task_210726_5): forward to DinTreasury stubs in DINTaskCoordinator.sol/DINTaskAuditor.sol — DinTreasury has existed since Feat/p3 fee router #51 (over a month), nobody's wired the forwarding yet.
• Simulation: model validator gas costs to size ETH network fee for model trainers #78: land the network-fee-floor number gas-simulation-network-fee.md already recommends as an actual DAO-settable contract parameter, after pricing the one gas cost its own "Open items" section flags as missing (commitAuditScore).

Blockers / design choices needing an answer

These are called out explicitly in the task file too, but flagging here since they're real open questions, not just implementation detail:

1. Per-model stake floor, §1(a): the design describes the model owner setting their own per-model stake requirement, bounded by DAO min/max — but the only setter that exists (setModelStakeBounds) is DAO-only (onlyOwner on DinValidatorStake), and there's no "model owner's chosen value" storage slot separate from the DAO bounds. Default proposal in the task: treat modelMinStakeBounds[modelId].min directly as the effective floor for now, defer a model-owner-facing setter as a fast-follow. Robbert — flag in your PR description if you think the setter is small enough to just include.
2. Concurrent-registration cap, §1(b): needs a new cross-contract counter (a validator's concurrent registrations span every model's task contracts, not just one) — no such counter exists today. The task proposes a shape (reuse the existing slasher-authorization trust boundary, increment/decrement around registration/GI-end) but the exact cap formula isn't nailed down anywhere in the design docs beyond "one 10-DIN stake cannot back 20 tasks." Robbert — propose the formula in your PR description.
3. Upheld-dispute bounty top-up, §2(b): this is the one real blocker, not just an open question. DinTreasury.withdrawETH/withdrawERC20 are both onlyOwner — a task contract has no path to pull a bounty out of DinTreasury today. Three options on the table: (a) add an authorized-withdrawer role to DinTreasury, (b) fund the bounty from a different source entirely (GI reward pool, disputed party's own stake), (c) treat the bounty as out of scope for this task and just get the on-chain credit-accounting right. This touches DinTreasury's access control, which is shared platform infrastructure — pick one and flag it here or on the PR before merging, don't decide it silently.
4. Fee-floor consumption point, §3: not obvious from the existing fee-router/model-registration flow where a networkFee/feeFloor constant should actually be read. Flag here if it's not clear once you're in the code.

Reference material

• Developer/tasks/task_100926_12.md — full spec
• Developer/tasks/task_100926_13.md — sibling task, shares §1(b)'s registration-counter infra with its own S6 item
• #37, #38, #43, #78 — tracking issues
• Developer/design/MECHANISM_DESIGN.md §3 (staking), §4 (slashing/disputes), §8 (fees/treasury)

Post progress updates, questions, and PR links here as the task moves — same convention as prior task-tracking discussions (e.g. #47, #106). If anything in the task spec doesn't feel right, is ambiguous, or needs clarification — comment here and cc @umeradl.

[umeradl]

Status update — 2026-09-22

All three PRs for this task are merged into develop:

PR	Closes	Merged
#140	#37	2026-09-16
#141	#43	2026-09-16
#142	#78	2026-09-20

Can this discussion + its tracked issues close? Not yet — two of the three still have open ends.

• P3 staking: DAO-settable floor, per-model stake requirements, concurrent-registration cap, withdrawal queue #37 (staking) — closable now. Both blockers from the original post are resolved as documented defaults: modelMinStakeBounds[modelId].min is the effective per-model floor (model-owner setter deferred as a fast-follow, per plan), and the concurrent-registration cap is wired with an explicit counter + releaseGIRegistrationSlots decrement hook. No known gaps.

• P3 fees & treasury: fee lines, fee-router contract, DIN denomination (P3-5.3) #43 (fees/treasury) — not closable yet. Of the three original TODO(task_210726_5) stubs, only two got wired: resolveDispute's frivolous-forfeiture split and DINTaskAuditor.settleRewards. The third — settleRecomputation's reject-path forfeiture (DINTaskCoordinator.sol ~line 1293, confirmed == false branch) — still only does treasuryAccrued += d.bond with the TODO comment still live, no actual transfer. Separately, the upheld-dispute bounty top-up (blocker Bump web3 from 7.10.0 to 7.15.0 in /dincli in the pip group across 1 directory #3 from the original post) is explicitly flagged out of scope in feat: wire dispute-bond and reward-pool treasury forwarding (closes #43) #141 — DinTreasury.withdrawERC20/withdrawETH are onlyOwner, so a task contract can't pull a bounty. Needs a decision: authorized-withdrawer role on DinTreasury, or fund the bounty from elsewhere (forfeited-stake pool / GI reward pool).

• Simulation: model validator gas costs to size ETH network fee for model trainers #78 (fee floor) — not closable yet. networkFeeFloor now exists as a DAO-settable storage var + setter, and the gas numbers behind it are measured and documented. But it's never read anywhere — no enforcement point checks depositRewards ≥ networkFeeFloor × nValidators at GI start. feat: commitAuditScore gas measurement + networkFeeFloor parameter (closes #78) #142 flagged this explicitly as needing confirmation on where that check belongs before the next release.

• P3 slashing: conditions taxonomy, penalty tiers, slashed-stake destination, dispute resolution (P3-4.1/4.2/4.3) #38 (slashing) was only ever partially in scope here (the treasury-forwarding piece, now folded into the P3 fees & treasury: fee lines, fee-router contract, DIN denomination (P3-5.3) #43 gap above). The rest of it is task_100926_13's territory (#138) — not this discussion's to close.

Recommendation

• Close P3 staking: DAO-settable floor, per-model stake requirements, concurrent-registration cap, withdrawal queue #37 now.
• Keep this discussion and P3 fees & treasury: fee lines, fee-router contract, DIN denomination (P3-5.3) #43/Simulation: model validator gas costs to size ETH network fee for model trainers #78 open pending two decisions from @umeradl: (1) the DinTreasury bounty-top-up access-control question, and (2) the networkFeeFloor enforcement point — plus wiring the one remaining settleRecomputation stub once (1) is decided (same fix covers both).

[umeradl]

Heads-up: dincli dindao is now dincli dinrep — PR No. 175 landed on develop (merge 5864f7d, follow-up 8f9a90b). The DIN-Representative command group is renamed to match the role: dincli/cli/dindao.py → dincli/cli/dinrep.py, Documentation/public/roles/dindao.md → dinrep.md. Breaking, no alias — dincli dindao … now returns "No such command". Subcommands are unchanged (dincli dinrep deploy …, dincli dinrep add-slasher …, dincli dinrep registry …).

For this task: nothing to change — this task file and its branches don't reference the old name. Use dincli dinrep from now on.