[Peras 18] Implement generic weighted Fait-Accompli with Local Sortition by agustinmista · Pull Request #1893 · IntersectMBO/ouroboros-consensus

agustinmista · 2026-02-24T15:39:09Z

This PR implements the initial version of a generic weighted Fait-Accompli with Local Sortition (wFA^LS) committee selection scheme. To allow for both Peras and Leios to reuse the same implementation, the code is abstracted in the following ways:

ElectionId type family: central to committee selection, it allows the client to provide their own different election identifier. Peras will use round numbers, whereas Leios will use something isomorphic to a slot number.
CryptoSupportsVoteSignature: crypto interface needed to sign/verify vote signatures using a public/private key pair.
CryptoSupportsVRF: crypto interface needed to compute/validate VRF outputs using a public/private key pair. This is needed by the Local Sortition fallback scheme.
CryptoSupportsWFALS: Wrapper on top of CryptoSupportsVoteSignature and CryptoSupportsVRF that allows us to derive both their key pairs from a single key pair. This could be used to have potentially different crypto schemes for the vote signing and VRF evaluation routines derived from the same "composite" key pair.
VoteSupportsWFALS: interface needed to extract the bits relevant to committee selection and signature verification from a client vote (i.e., a concrete Peras or Leios vote).

In addition, this PR also provides an initial instantiation of all these type classes for Peras using BLS crypto (the best approximation for the final one we have for now).

Other notes

While this PR only provides wFA^LS, the design should be modular enough to also allow one to reuse some of the interfaces (e.g., ElectionId, and CryptoSupportsVoteSigning) for other committee selection schemes (e.g. wFA^IID).
Conformance tests against the Rust implementation will be added after Peras 16 lands into main.
Because epoch boundaries are a bit special, O.C.Committee.AcrossEpochs defines a small wrapper to keep the committee selections from both the current and previous epochs at hand. This is needed to validate votes cast at the end of the previous epoch arriving at the beginning of the current one.
Because any instance of a weighted-Fait-Accompli-based committee selection for a given epoch would need to aggregate the input stake distribution in the same way, we have extracted this into the ExtCumulativeStakeDistr data type. This would allow both Peras and Leios to share part of the heavy work needed at the beginning of every new epoch.
This PR does not include any other potential optimization beyond the one described above.

This commit implements an initial version of the weigthed Fait-Accompli with Local Sortition committee selection scheme. This implementation is generic w.r.t. both the election and vote types, as well as the crypto scheme used by the client. Co-authored-by: Nicolas BACQUEY <nicolas.bacquey@tweag.io> Co-authored-by: Thomas BAGREL <thomas.bagrel@tweag.io> Co-authored-by: Agustin Mista <agustin.mista@moduscreate.com>

This commit defines an initial instantiation of the weighted Fait-Accompli with Local Sortition scheme for Peras using pre-existing BLS crypto. Co-authored-by: Nicolas BACQUEY <nicolas.bacquey@tweag.io> Co-authored-by: Thomas BAGREL <thomas.bagrel@tweag.io> Co-authored-by: Agustin Mista <agustin.mista@moduscreate.com>

This commit refactors the weighted Fait-Accompli conformance test suite into a composite unit test that accepts a function to preprocess the stake distribution into a value that gets shared accross individual tests. This is useful to instantiate this test suite against the real implementation, which relies on a precomputed stake distribution array that can be shared across all 3000 individual test steps. Co-authored-by: Nicolas BACQUEY <nicolas.bacquey@tweag.io> Co-authored-by: Thomas BAGREL <thomas.bagrel@tweag.io> Co-authored-by: Agustin Mista <agustin.mista@moduscreate.com>

Co-authored-by: Nicolas BACQUEY <nicolas.bacquey@tweag.io> Co-authored-by: Thomas BAGREL <thomas.bagrel@tweag.io> Co-authored-by: Agustin Mista <agustin.mista@moduscreate.com>

perturbing

Functionally, it looks good 👍

Some small details and remarks that we have to crystallize.

perturbing · 2026-03-18T14:31:41Z

+--   * total non-persistent stake
+--
+-- TODO: this can be more efficient if we compute the values directly on the array.
+weightedFaitAccompliSplitSeats ::


I agree with @tbagrel1 that we should at least prevent in some way this request (committee size larger than the number of pools that have non-zero stake).

perturbing · 2026-03-18T14:33:52Z

+    expectedSeats =
+      fromMaybe 0 $
+        taylorExpCmpFirstNonLower
+          3


This value should be scrutinized. That is, we need to understand why this was set at 3

Added a TODO here as well.

perturbing · 2026-03-18T14:36:16Z

+
+-- | Persistent committee size
+newtype PersistentCommitteeSize = PersistentCommitteeSize
+  { unPersistentCommitteeSize :: Word64


Why Word64?

Also for the others below.

I think Word64 is the de-facto type for natural numbers in the codebase. At least I haven't come across any arbitrary-length Natural being used for long-lived values.

I ask because we can also use Word16 with values range from 0 to 65535.

I do not expect the committee size and other so exceed that. Feels like Word16 is overkill, but if it is the defacto choice, I yield ;)

Oh, right. You were actually suggesting to go in the opposite direction haha.

As for Word16, I don't think it matters much unless we're sending these values over the wire and control the memory/packet layout (which we are not).

My rationale here being that GHC will probably not use the remaining 48bits on something else, and instead it will just dedicate the same amount of memory as if it were a Word64 due to memory alignment. See the "Basic Types" section of this old wiki entry.

agustinmista · 2026-03-19T15:10:53Z

FYI: this PR is already large enough, so I started working on its continuation in Peras 20, which includes:

Tiebreaker for pools with the same stake
Conformance tests between model and (Haskell) implementation

perturbing · 2026-03-19T15:38:58Z

@agustinmista and @tbagrel1 , please be aware of this cardano-scaling/leios-wfa-ls-demo#4

dnadales

Commenting for now waiting for a response on:

Underflow checking (mkExtWFAStakeDistr)
Possibly missing isPersistentMember check.

dnadales · 2026-03-19T15:17:52Z

+    ElectionId c ->
+    VoteMessage c ->
+    VoteSignature c ->
+    Either String ()


Why did we decide to use String here instead of a custom error type?

We made it easy to instantiate this class directly with some of the crypto primitives from cardano-crypto-class, which all seem to have Either String () as the return type for verification operations.

That said, I don't have strong opinions, and we can wrap this with a ComitteeSelectionCryptoError newtype if you prefer.

EDIT: these errors are also already wrapped by the InvalidVoteSignature variant of CommitteeSelectionError.

dnadales · 2026-03-19T17:19:05Z

+verifyVote vote selection =
+  getVoteView @c vote $ \case
+    PersistentVote seatIndex electionId message sig
+      | seatIndexWithinBounds seatIndex selection -> do


Don't we also need to check isPersistentMember seatIndex selection?

Yes, good catch!

Likewise, we also need to check in the NonPersistentVote variant that, after retrieving the seat index of someone who claims to be a non-persistent voter, that seat index does not correspond to a persistent member:

NonPersistentVote poolId electionId message vrfOutput sig | Just seatIndex <- Map.lookup poolId (candidateSeats selection) , not (isPersistentMember seatIndex selection) -> ...

dnadales · 2026-03-19T17:27:35Z

+  -- number of seats granted by local sortition and their stake (normalized
+  -- by the total non-persistent stake)
+  NonPersistentCommitteeMember
+    (NonPersistentMemberProof numSeats _sig _vrfOutput)


It seems names are swapped 🤔

Not sure I understand what you mean but, to make it more consistent I renamed these constructors to (Non)PersistentMembershipProof.

Hopefully that sparks joy.

dnadales · 2026-03-19T17:30:54Z

+  stakeDistrArray =
+    listArray
+      ( SeatIndex 0
+      , SeatIndex (fromIntegral (Map.size pools) - 1)


Could there be a potential underflow here if Map.size pools == 0?

Yes, I was planning on discharging this case elsewhere and use a non-empty map as input here, but we may as well do it directly while constructing this structure.

I tweaked this function to return an Either WFAError (ExtWFAStakeDistr a).

agustinmista · 2026-03-20T10:10:58Z

Thanks @dnadales for the review!

I think I addressed all your feedback now (last 5 commits in this PR, to be squashed) 🤞

agustinmista · 2026-04-13T08:40:27Z

Superseded by #1974 and #1975

agustinmista self-assigned this Feb 24, 2026

agustinmista added the peras label Feb 24, 2026

agustinmista force-pushed the peras/committee-selection-rules branch 7 times, most recently from fdbd4da to 96b1d18 Compare March 2, 2026 19:35

agustinmista changed the title ~~WIP implementing committee selection rules~~ WIP implementing generic committee selection Mar 2, 2026

agustinmista force-pushed the peras/committee-selection-rules branch 2 times, most recently from daa1fdd to 5e6723d Compare March 3, 2026 08:50

perturbing reviewed Mar 3, 2026

View reviewed changes

agustinmista force-pushed the peras/committee-selection-rules branch 3 times, most recently from 94579d3 to 21c11dc Compare March 4, 2026 11:24

tbagrel1 reviewed Mar 4, 2026

View reviewed changes

Comment thread ouroboros-consensus/src/ouroboros-consensus/Ouroboros/Consensus/Committee/Crypto.hs Outdated

Comment thread ouroboros-consensus/src/ouroboros-consensus/Ouroboros/Consensus/Committee/Types.hs Outdated

agustinmista force-pushed the peras/committee-selection-rules branch 2 times, most recently from 4545066 to b621259 Compare March 5, 2026 13:05

tbagrel1 force-pushed the peras/committee-selection-rules branch from 92a378e to 2374824 Compare March 6, 2026 11:37

agustinmista force-pushed the peras/committee-selection-rules branch 5 times, most recently from f757764 to 466680c Compare March 9, 2026 13:24

agustinmista changed the title ~~WIP implementing generic committee selection~~ [Peras 18] Generic weighted Fait-Accompli committee selection scheme Mar 9, 2026

agustinmista requested review from nbacquey, perturbing and tbagrel1 March 9, 2026 13:56

agustinmista force-pushed the peras/committee-selection-rules branch from 466680c to 36d9a2f Compare March 11, 2026 08:18

agustinmista requested review from bladyjoker, dnadales and geo2a as code owners March 12, 2026 11:52

agustinmista force-pushed the peras/committee-selection-rules branch 3 times, most recently from 7f18933 to 8fb6ada Compare March 16, 2026 10:26

geo2a force-pushed the main branch from c4a43e1 to 2508010 Compare March 16, 2026 13:20

agustinmista and others added 4 commits March 17, 2026 15:47

Add changelog

22e70f4

Co-authored-by: Nicolas BACQUEY <nicolas.bacquey@tweag.io> Co-authored-by: Thomas BAGREL <thomas.bagrel@tweag.io> Co-authored-by: Agustin Mista <agustin.mista@moduscreate.com>

agustinmista force-pushed the peras/committee-selection-rules branch from 8fb6ada to 22e70f4 Compare March 17, 2026 14:51

perturbing reviewed Mar 18, 2026

View reviewed changes

Address @perturbing's feedback and other minor fixes

412f4df

agustinmista force-pushed the peras/committee-selection-rules branch from 6d6d5d1 to 412f4df Compare March 19, 2026 11:45

dnadales reviewed Mar 19, 2026

View reviewed changes

agustinmista added 5 commits March 20, 2026 09:33

Rename type parameter c to crypto

86a32cf

Rename type parameter m to membership

4dfb674

Add missing check to verifyVote

83a3d94

Add explicit error when trying to construct an empty ExtWFAStakeDistr

3ace31c

Fix typos and minor renamings

d4616a8

This was linked to issues Mar 24, 2026

Mocked Vote and Cert Cryptography tweag/cardano-peras#212

Closed

Voting Committee selection logic tweag/cardano-peras#210

Open

agustinmista mentioned this pull request Mar 25, 2026

[Peras 20] Implement tiebreaker and extra tests for weighted Fait-Accompli #1936

Closed

agustinmista marked this pull request as draft March 25, 2026 14:41

agustinmista closed this Apr 13, 2026

Conversation

agustinmista commented Feb 24, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Other notes

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

perturbing left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

agustinmista Mar 19, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

agustinmista commented Mar 19, 2026

Uh oh!

perturbing commented Mar 19, 2026

Uh oh!

dnadales left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

agustinmista Mar 20, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

agustinmista Mar 20, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

agustinmista commented Mar 20, 2026

Uh oh!

agustinmista commented Apr 13, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

agustinmista commented Feb 24, 2026 •

edited

Loading

agustinmista Mar 19, 2026 •

edited

Loading

agustinmista Mar 20, 2026 •

edited

Loading

agustinmista Mar 20, 2026 •

edited

Loading