Skip to main content

64/WAKU2-NETWORK

  • Status: draft
  • Category: Best Current Practice
  • Editor: Hanno Cornelius <hanno@status.im>

Abstract

This specification describes an opinionated deployment of 10/WAKU2 protocols to form a coherent and shared decentralized messaging network that is open-access, useful for generalized messaging, privacy-preserving, scalable and accessible even to resource-restricted devices. We'll refer to this opinionated deployment simply as the public Waku Network, the Waku Network or, if the context is clear, the network in the rest of this document. All The Waku Network configuration parameters are listed here.

Theory / Semantics

Routing protocol

The Waku Network is built on the 17/WAKU2-RLN-RELAY routing protocol, which in turn is an extension of 11/WAKU2-RELAY with spam protection measures.

Network shards

Traffic in the Waku Network is sharded into eight 17/WAKU2-RLN-RELAY pubsub topics. Each pubsub topic is named according to the static shard naming format defined in WAKU2-RELAY-SHARDING with:

  • <cluster_id> set to 1
  • <shard_number> occupying the range 0 to 7. In other words, the Waku Network is a 17/WAKU2-RLN-RELAY network routed on the combination of the eight pubsub topics:
/waku/2/rs/1/0
/waku/2/rs/1/1
...
/waku/2/rs/1/7

A node MUST use WAKU-METADATA protocol to identify the <cluster_id> that every inbound/outbound peer that attempts to connect supports. In any of the following cases, the node MUST trigger a disconnection:

Roles

There are two distinct roles evident in the network, those of: 1) nodes, and 2) applications.

Nodes

Nodes are the individual software units using 10/WAKU2 protocols to form a p2p messaging network. Nodes, in turn, can participate in a shard as full relayers, i.e. relay nodes, or by running a combination of protocols suitable for resource-restricted environments, i.e. non-relay nodes. Nodes can also provide various services to the network, such as storing historical messages or protecting the network against spam. See the section on default services for more.

Relay nodes

Relay nodes MUST follow 17/WAKU2-RLN-RELAY to route messages to other nodes in the network for any of the pubsub topics defined as the Waku Network shards. Relay nodes MAY choose to subscribe to any of these shards, but MUST be subscribed to at least one defined shard. Each relay node SHOULD be subscribed to as many shards as it has resources to support. If a relay node supports an encapsulating application, it SHOULD be subscribed to all the shards servicing that application. If resource restrictions prevent a relay node from servicing all shards used by the encapsulating application, it MAY choose to support some shards as a non-relay node.

Bootstrapping and discovery

Nodes MAY use any method to bootstrap connection to the network, but it is RECOMMENDED that each node retrieves a list of bootstrap peers to connect to using EIP-1459 DNS-based discovery. Relay nodes SHOULD use 33/WAKU2-DISCV5 to continually discover other peers in the network. Each relay node MUST encode its supported shards into its discoverable ENR, as described in WAKU2-RELAY-SHARDING. The ENR MUST be updated if the set of supported shards change. A node MAY choose to ignore discovered peers that do not support any of the shards in its own subscribed set.

Transports

Relay nodes MUST follow 10/WAKU2 specifications with regards to supporting different transports. If TCP transport is available, each relay node MUST support it as transport for both dialing and listening. In addition, a relay node SHOULD support secure websockets for bidirectional communication streams, for example to allow connections from and to web browser-based clients. A relay node MAY support unsecure websockets if required by the application or running environment.

Default services

For each supported shard, each relay node SHOULD enable and support the following protocols as a service node:

  1. 12/WAKU2-FILTER to allow resource-restricted peers to subscribe to messages matching a specific content filter.
  2. 13/WAKU2-STORE to allow other peers to request historical messages from this node.
  3. 19/WAKU2-LIGHTPUSH to allow resource-restricted peers to request publishing a message to the network on their behalf.
  4. WAKU2-PEER-EXCHANGE to allow resource-restricted peers to discover more peers in a resource efficient way.

Store service nodes

Each relay node SHOULD support 13/WAKU2-STORE as a store service node, for each supported shard. The store SHOULD be configured to retain at least 12 hours of messages per supported shard. Store service nodes SHOULD only store messages with a valid rate_limit_proof attribute.

Non-relay nodes

Nodes MAY opt out of relay functionality on any network shard and instead request services from relay nodes as clients using any of the defined service protocols:

  1. 12/WAKU2-FILTER to subscribe to messages matching a specific content filter.
  2. 13/WAKU2-STORE to request historical messages matching a specific content filter.
  3. 19/WAKU2-LIGHTPUSH to request publishing a message to the network.
  4. WAKU2-PEER-EXCHANGE to discover more peers in a resource efficient way.

Store client nodes

Nodes MAY request historical messages from 13/WAKU2-STORE service nodes as store clients. A store client SHOULD discard any messages retrieved from a store service node that do not contain a valid rate_limit_proof attribute. The client MAY consider service nodes returning messages without a valid rate_limit_proof attribute as untrustworthy. The mechanism by which this may happen is currently underdefined.

Applications

Applications are the higher-layer projects or platforms that make use of the generalized messaging capability of the network. In other words, an application defines a payload used in the various 10/WAKU2 protocols. Any participant in an application SHOULD make use of an underlying node in order to communicate on the network. Applications SHOULD make use of an autosharding API to allow the underlying node to automatically select the target shard on the Waku Network. See the section on autosharding for more.

RLN rate-limiting

The 17/WAKU2-RLN-RELAY protocol uses RLN-V2 proofs to ensure that a pre-agreed rate limit of x messages every y seconds is not exceeded by any publisher. While the network is under capacity, individual relayers MAY choose to freely route messages without RLN proofs up to a discretionary bandwidth limit, after which messages without proofs MUST be discarded by relay nodes. This bandwidth limit SHOULD be enforced using a bandwidth validation mechanism separate from a RLN rate-limiting. This implies that quality of service and reliability is significantly lower for messages without proofs and at times of high network utilization these messages may not be relayed at all.

RLN Parameters

The Waku Network uses the following RLN parameters:

  • rlnRelayUserMessageLimit=100: Amount of messages that a membership is allowed to publish per epoch. Configurable between 0 and MAX_MESSAGE_LIMIT.
  • rlnEpochSizeSec=600: Size of the epoch in seconds.
  • rlnRelayChainId=11155111: Network in which the RLN contract is deployed, aka Sepolia.
  • rlnRelayEthContractAddress=0xCB33Aa5B38d79E3D9Fa8B10afF38AA201399a7e3: Network address where RLN memberships are stored.
  • staked_fund=0: In other words, the Waku Network does not use RLN staking. Registering a membership just requires to pay gas.
  • MAX_MESSAGE_LIMIT=100: Maximum amount of messages allowed per epoch for any membership. Enforced in the contract.
  • max_epoch_gap=20: Maximum allowed gap in seconds into the past or future compared to the validator's clock.

Nodes MUST reject messages not respecting any of these parameters. Nodes SHOULD use Network Time Protocol (NTP) to synchronize their own clocks, thereby ensuring valid timestamps for proof generation and validation. Publishers to the Waku Network SHOULD register an RLN membership.

RLN Proofs

Each RLN member MUST generate and attach an RLN proof to every published message as described in 17/WAKU2-RLN-RELAY and RLN-V2. Slashing is not implemented for the Waku Network. Instead, validators will penalise peers forwarding messages exceeding the rate limit as specified for the rate-limiting validation mechanism. This incentivizes all relay nodes to validate RLN proofs and reject messages violating rate limits in order to continue participating in the network.

Network traffic

All payload on the Waku Network MUST be encapsulated in a 14/WAKU2-MESSAGE with rate limit proof extensions defined for 17/WAKU2-RLN-RELAY. Each message on the Waku Network SHOULD be validated by each relayer, according to the rules discussed under message validation.

Message Attributes

  • The mandatory payload attribute MUST contain the message data payload as crafted by the application.
  • The mandatory content_topic attribute MUST specify a string identifier that can be used for content-based filtering. This is also crafted by the application. See Autosharding for more on the content topic format.
  • The optional meta attribute MAY be omitted. If present, will form part of the message uniqueness vector described in 14/WAKU2-MESSAGE.
  • The optional version attribute SHOULD be set to 0. It MUST be interpreted as 0 if not present.
  • The mandatory timestamp attribute MUST contain the Unix epoch time at which the message was generated by the application. The value MUST be in nanoseconds. It MAY contain a fudge factor of up to 1 seconds in either direction to improve resistance to timing attacks.
  • The optional ephemeral attribute MUST be set to true if the message should not be persisted by the Waku Network.
  • The optional rate_limit_proof attribute SHOULD be populated with the RLN proof as set out in RLN Proofs. Messages with this field unpopulated MAY be discarded from the network by relayers. This field MUST be populated if the message should be persisted by the Waku Network.

Message Size

Any 14/WAKU2-MESSAGE published to the network MUST NOT exceed an absolute maximum size of 150 kilobytes. This limit applies to the entire message after protobuf serialization, including attributes. It is RECOMMENDED not to exceed an average size of 4 kilobytes for 14/WAKU2-MESSAGE published to the network.

Message Validation

Relay nodes MUST apply gossipsub v1.1 validation to each relayed message and SHOULD apply all of the rules set out in the section below to determine the validity of a message. Validation has one of three outcomes, repeated here from the gossipsub specification for ease of reference:

  1. Accept - the message is considered valid and it MUST be delivered and forwarded to the network.
  2. Reject - the message is considered invalid, MUST be rejected and SHOULD trigger a gossipsub scoring penalty against the transmitting peer.
  3. Ignore - the message SHOULD NOT be delivered and forwarded to the network, but this MUST NOT trigger a gossipsub scoring penalty against the transmitting peer.

The following validation rules are defined:

Decoding failure

If a message fails to decode as a valid 14/WAKU2-MESSAGE, the relay node MUST reject the message. This SHOULD trigger a penalty against the transmitting peer.

Invalid timestamp

If a message has a timestamp deviating by more than 20 seconds either into the past or the future when compared to the relay node's internal clock, the relay node MUST reject the message. This allows for some deviation between internal clocks, network routing latency and an optional fudge factor when timestamping new messages.

Free bandwidth exceeded

If a message contains no RLN proof and the current bandwidth utilization on the shard the message was published to equals or exceeds 1 Mbps, the relay node SHOULD ignore the message.

Invalid RLN epoch

If a message contains an RLN proof and the epoch attached to the proof deviates by more than max_epoch_gap seconds from the relay node's own epoch, the relay node MUST reject the message. max_epoch_gap is set to 20 seconds for the Waku Network.

Invalid RLN proof

If a message contains an RLN proof and the zero-knowledge proof is invalid according to the verification process described in RLN-V2, the relay node MUST ignore the message.

Rate limit exceeded

If a message contains an RLN proof and the relay node detects double signaling according to the verification process described in RLN-V2, the relay node MUST reject the message for violating the agreed rate limit of rlnRelayUserMessageLimit messages every rlnEpochSizeSec second. This SHOULD trigger a penalty against the transmitting peer.

Autosharding

Nodes in the Waku Network SHOULD allow encapsulating applications to use autosharding, as defined in WAKU2-RELAY-SHARDING by automatically determining the appropriate pubsub topic from the list of defined Waku Network shards. This allows the application to omit the target pubsub topic when invoking any Waku protocol function. Applications using autosharding MUST use content topics in the format defined in WAKU2-RELAY-SHARDING and SHOULD use the short length format:

/{application-name}/{version-of-the-application}/{content-topic-name}/{encoding}

When an encapsulating application makes use of autosharding the underlying node MUST determine the target pubsub topic(s) from the content topics provided by the application using the hashing mechanism defined in WAKU2-RELAY-SHARDING.

Copyright and related rights waived via CC0.

References