FOCIL: get_inclusion_list_committee and signature domain may diverge from current Heze spec

[rahulbarmann]

Background

While adding tests for the FOCIL helpers (#17090) I compared the focil branch against the current Heze spec and found a couple of places that seem to have drifted. Since the code predates recent spec revisions, checking whether these are lags or intentional.

Description

1. Committee selection (beacon-chain/core/helpers/inclusion_list.go:47)

Current impl unshuffles all active validators and slices:

start := uint64(slot%SlotsPerEpoch) * InclusionListCommitteeSize
shuffledList, _ := UnshuffleList(shuffledIndices, seed)
return shuffledList[start:end], nil

Current spec (specs/heze/beacon-chain.md) concatenates the slot's beacon committees and cycles:

for i in range(committees_per_slot):
    indices.extend(get_beacon_committee(state, slot, CommitteeIndex(i)))
return Vector[...]([indices[i % len(indices)] for i in range(INCLUSION_LIST_COMMITTEE_SIZE)])

Different validator sets, so inclusion_list_committee_root won't match other clients. This is the one with interop impact.

2. Signature domain epoch (beacon-chain/core/helpers/inclusion_list.go:92)

Impl derives the domain from the head-state slot, spec uses the IL's own slot:

currentEpoch := slots.ToEpoch(st.Slot())        // impl

compute_epoch_at_slot(message.slot)             # spec (is_valid_inclusion_list_signature)

Agrees within an epoch, diverges at epoch boundaries.

For completeness: the gossip rules in beacon-chain/sync/inclusion_list.go also follow an older previous-or-current-slot variant with REJECT severities vs the spec's current-slot [IGNORE], and IL timing is seconds-based rather than INCLUSION_LIST_DUE_BPS.

Happy to open a fix PR for (1) and (2) against focil if they're confirmed as lags. Thanks!

cc @terencechain