This document should be viewed as additive to the [document from Phase 0](../phase0/p2p-interface.md) and will be referred to as the "Phase 0 document" hereafter.
Readers should understand the Phase 0 document and use it as a basis to understand the changes outlined in this document.
Altair adds new messages, topics and data to the Req-Resp, Gossip and Discovery domain. Some Phase 0 features will be deprecated, but not removed immediately.
The `MetaData` stored locally by clients is updated with an additional field to communicate the sync committee subnet subscriptions:
```
(
seq_number: uint64
attnets: Bitvector[ATTESTATION_SUBNET_COUNT]
syncnets: Bitvector[SYNC_COMMITTEE_SUBNET_COUNT]
)
```
Where
-`seq_number` and `attnets` have the same meaning defined in the Phase 0 document.
-`syncnets` is a `Bitvector` representing the node's sync committee subnet subscriptions. This field should mirror the data in the node's ENR as outlined in the [validator guide](./validator.md#sync-committee-subnet-stability).
The derivation of the `message-id` has changed starting with Altair to incorporate the message `topic` along with the message `data`. These are fields of the `Message` Protobuf, and interpreted as empty byte strings if missing.
Altair changes the type of the global beacon block topic and adds one global topic to propagate partially aggregated sync committee messages to all potential proposers of beacon blocks.
The following validations MUST pass before forwarding the `signed_contribution_and_proof` on the network; define `contribution_and_proof = signed_contribution_and_proof.message`, `contribution = contribution_and_proof.contribution`, and the following function `get_sync_subcommittee_pubkeys` for convenience:
- _[IGNORE]_ The contribution's slot is for the current slot (with a `MAXIMUM_GOSSIP_CLOCK_DISPARITY` allowance), i.e. `contribution.slot == current_slot`.
- _[REJECT]_ `contribution_and_proof.selection_proof` selects the validator as an aggregator for the slot -- i.e. `is_sync_committee_aggregator(contribution_and_proof.selection_proof)` returns `True`.
- _[REJECT]_ The aggregator's validator index is in the declared subcommittee of the current sync committee --
i.e. `state.validators[contribution_and_proof.aggregator_index].pubkey in get_sync_subcommittee_pubkeys(state, contribution.subcommittee_index)`.
- _[IGNORE]_ The sync committee contribution is the first valid contribution received for the aggregator with index `contribution_and_proof.aggregator_index`
for the slot `contribution.slot` and subcommittee index `contribution.subcommittee_index`
(this requires maintaining a cache of size `SYNC_COMMITTEE_SIZE` for this topic that can be flushed after each slot).
- _[REJECT]_ The `contribution_and_proof.selection_proof` is a valid signature of the `SyncAggregatorSelectionData` derived from the `contribution` by the validator with index `contribution_and_proof.aggregator_index`.
- _[REJECT]_ The aggregate signature is valid for the message `beacon_block_root` and aggregate pubkey derived from the participation info in `aggregation_bits` for the subcommittee specified by the `contribution.subcommittee_index`.
The `sync_committee_{subnet_id}` topics are used to propagate unaggregated sync committee messages to the subnet `subnet_id` to be aggregated before being gossiped to the global `sync_committee_contribution_and_proof` topic.
- _[IGNORE]_ The message's slot is for the current slot (with a `MAXIMUM_GOSSIP_CLOCK_DISPARITY` allowance), i.e. `sync_committee_message.slot == current_slot`.
- _[REJECT]_ The `subnet_id` is valid for the given validator, i.e. `subnet_id in compute_subnets_for_sync_committee(state, sync_committee_message.validator_index)`.
- _[IGNORE]_ There has been no other valid sync committee message for the declared `slot` for the validator referenced by `sync_committee_message.validator_index`
(this requires maintaining a cache of size `SYNC_COMMITTEE_SIZE // SYNC_COMMITTEE_SUBNET_COUNT` for each subnet that can be flushed after each slot).
Note this validation is _per topic_ so that for a given `slot`, multiple messages could be forwarded with the same `validator_index` as long as the `subnet_id`s are distinct.
Individual validators can be duplicated in the broader sync committee such that they are included multiple times in a given subcommittee or across multiple subcommittees.
Aggregated sync committee messages are packaged into (signed) `SyncCommitteeContribution` along with proofs and gossiped to the `sync_committee_contribution_and_proof` topic.
With any fork, the fork version, and thus the `ForkDigestValue`, change.
Message types are unique per topic, and so for a smooth transition a node must temporarily subscribe to both the old and new topics.
The topics that are not removed in a fork are updated with a new `ForkDigestValue`. In advance of the fork, a node SHOULD subscribe to the post-fork variants of the topics.
Subscriptions are expected to be well-received, all updated nodes should subscribe as well.
Topic-meshes can be grafted quickly as the nodes are already connected and exchanging gossip control messages.
Messages SHOULD NOT be re-broadcast from one fork to the other.
A node's behavior before the fork and after the fork are as follows:
Pre-fork:
- Peers who propagate messages on the post-fork topics MAY be scored negatively proportionally to time till fork,
to account for clock discrepancy.
- Messages can be IGNORED on the post-fork topics, with a `MAXIMUM_GOSSIP_CLOCK_DISPARITY` margin.
Post-fork:
- Peers who propagate messages on the pre-fork topics MUST NOT be scored negatively. Lagging IWANT may force them to.
- Messages on pre and post-fork variants of topics share application-level caches.
E.g. an attestation on the both the old and new topic is ignored like any duplicate.
- Two epochs after the fork, pre-fork topics SHOULD be unsubscribed from. This is well after the configured `seen_ttl`.
The length of this bitfield is `SYNC_COMMITTEE_SUBNET_COUNT` where each bit corresponds to a distinct `subnet_id` for a specific sync committee subnet.
