---
title: WAKU-RENDEZVOUS
name: Waku rendez vous discovery
editor: Simon-Pierre Vivier <simvivier@status.im>
contributors:
---

## Abstract
This document describe the goal,
strategy and usage of the libp2p rendezvous protocol by Waku.

Rendezvous is one of the discovery methods that can be used by Waku,
it supplements Discovery v5 and Waku peer exchange but fill a specific role.

## Background and Rationale
How do new nodes join the network is the question that discovery answers.
Discovery must be fast, pertinent and resilient to attacks.
Rendezvous is both fast and allow the discovery of relevant peers,
although by it's design can be easily abused.
The properties of rendezvous complements well the slower but safer methods like Discv5.
To contribute well, a Waku nodes must know a sufficient number of peers with
a wide variety of capabilities.
By using rendezvous in combination with
[Discv5](https://github.com/ethereum/devp2p/blob/master/discv5/discv5.md#node-discovery-protocol-v5) and
[Waku peer exchange](https://github.com/waku-org/specs/blob/master/standards/core/peer-exchange.md#abstract),
Waku nodes will reach a good number of meaningful peers
faster than through a single discovery method.

## Semantics
Waku rendezvous fully inherit the [libp2p rendezvous semantics](https://github.com/libp2p/specs/blob/master/rendezvous/README.md#rendezvous-protocol).

## Specifications
The namespaces used to register and request will be in the format `rs/cluster-id/shard`.

Every [Waku Relay](https://github.com/vacp2p/rfc-index/blob/main/waku/standards/core/11/relay.md) node will also be initialized as a rendezvous point.

Each relay node will register with random rendezvous points every 10 seconds,
once for each shard it supports and register only the namespace corresponding to that shard.
Each registrations will expire after 1 minute.

At startup, every Waku nodes will discover peers by sending requests to random rendezvous points,
once for each shard it supports.
A maximum of 12 peers will be requested each time.

This number is enough for good GossipSub connectivity and
minimize the load on rendezvous points.
It is assumed that the bootstrap node used is a rendezvous point and
that other discovery methods are used in conjunction and
will continue discovering peers for the lifetime of the node.

## Future Work

Namespaces will not contain capabilities yet but may in the future. If the need arise nodes could use rendezvous to discover peers with specific capabilities.