Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

[RFC] peer discovery with mDNS #80

Merged
merged 14 commits into from
May 6, 2019
93 changes: 93 additions & 0 deletions discovery/mdns.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,93 @@
# Multicast DNS
Author: Richard Schneider (makaretu@gmail.com)


## Overview

The goal is to allow peers to discover each other when on the same local network with zero configuration.
MDNS uses a multicast system of DNS records; this allows all peers on the local network to see all query responses.

Conceptually, it is very simple. When a peer starts (or detects a network change), it sends a query for all peers.
As responses come in, the peer adds the other peers information into is local database of peers.

## Definitions

`service-name` is the DNS-SD service name for all peers. It is defined as `_p2p._udp.local`.

`host-name` is the fully qualified name of the peer. It is derived from the peer's name and `p2p.local`.

`peer-name` is the case-insenstive ID of the peer and less than 64 characters. It is normally is the base-32 encoding of peer's ID.
Copy link
Member

@lidel lidel Sep 17, 2018

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I guess until ipfs/kubo#5287 is solved this would require apps to do a manual step of converting back to Base58, yes?

peer-name is just a unique name, we should not assume it's the peer id. The dnsaddr ends with the peer ID.

What if PeerID is longer than 64 characters? Should we split after 63 characters like noted in ipfs/in-web-browsers#89 ?

I think Workaround A: Split at 63rd character makes sense.

richardschneider marked this conversation as resolved.
Show resolved Hide resolved

Note that all names are case insensitive.

## Peer Discovery

### Request

To find all peers, a DNS message is sent with the question `_p2p._udp.local PTR`.
Peers will then start responding with their details.

Note that a peer must respond to it's own query. This allows other peers to passively discover it.

### Response

On receipt of a `find all peers` query, a peer sends a DNS response message (QR = 1) that contains
the **answer**

<service-name> PTR <peer-name>.<service-name>

The **additional records** of the response contain the peer's discovery details

<peer-name>.<service-name> TXT "dnsaddr=..."

The TXT record contains the multiaddresses that the peer is listening on. Each multiaddress
is a TXT attribute with the form `dnsaddr=.../p2p/QmId`. Multiple `dnsaddr` attributes
and/or TXT records are allowed.

## DNS Service Discovery

DNS-SD support is not needed for peers to discover each other. However, it is
extremely usefull for network administrators to discover what is running on the
network.

### Meta Query

This allows discovery of all services. The question is `_services._dns-sd._udp.local PTR`.

A peer responds with the answer

_services._dns-sd._udp.local PTR <service-name>
richardschneider marked this conversation as resolved.
Show resolved Hide resolved

### Find All Response

On receipt of a `find all peers` query, the following **additional records** should be included

<peer-name>.<service-name> SRV ... <host-name>
<host-name> A <ipv4 address>
<host-name> AAAA <ipv6 address>

### Gotchas

Many existing tools ignore the Additional Records and always send individual queries for the
peer's discovery details. To accomodate this, a peer should respond to the following queries:

- `<peer-name>.<service-name> SRV`
- `<peer-name>.<service-name> TXT`
- `<host-name> A`
- `<host-name> AAAA`

## Issues

- MDNS requires link local addresses. Loopback and "nat busting" addresses should not sent and must
be ignored on receipt?

## References

- [RFC 1035 - Domain Names (DNS)](https://tools.ietf.org/html/rfc1035)
- [RFC 6762 - Multicast DNS](https://tools.ietf.org/html/rfc6762)
- [RFC 6763 - DNS-Based Service Discovery](https://tools.ietf.org/html/rfc6763)
- [Multiaddr](https://github.com/multiformats/multiaddr)

## Worked Example

**TODO**