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

add HTTP spec #508

Open
wants to merge 35 commits into
base: master
Choose a base branch
from
Open
Changes from 30 commits
Commits
Show all changes
35 commits
Select commit Hold shift + click to select a range
bc1aa59
add HTTP spec
marten-seemann Jan 22, 2023
1f075f6
2nd attempt for server auth
marten-seemann Jan 29, 2023
12f86b8
require client to authenticate the server when doing client auth
marten-seemann Jan 29, 2023
146c09a
better motivation for libp2p+HTTP (#515)
marten-seemann Feb 14, 2023
5398f5d
fix a few typos
marten-seemann Feb 14, 2023
b6c1bc2
http: use .well-known/libp2p.json for configuration
marten-seemann Mar 2, 2023
8a57943
http: nest libp2p.json config to allow for future configuration
marten-seemann Mar 2, 2023
d506145
Merge pull request #529 from libp2p/http-well-known-configuration
MarcoPolo Jun 1, 2023
946f516
Reformat the spec from the Point of View of an implementer
MarcoPolo Jul 7, 2023
3681472
Add link
MarcoPolo Jul 7, 2023
dd5d07c
Merge comments
MarcoPolo Jul 10, 2023
46d1857
Merge pull request #556 from libp2p/marco/http-update
MarcoPolo Jul 10, 2023
ebe612c
Add note about how this is just one possible auth mechanism
MarcoPolo Jul 10, 2023
7e5a077
Add lidel to interest group
MarcoPolo Jul 14, 2023
db2b3b5
Update http/README.md
MarcoPolo Jul 17, 2023
6319458
Formatting
MarcoPolo Jul 17, 2023
c7c9c43
Add thomas
MarcoPolo Jul 17, 2023
454e25c
Use metadata map and call it protocols
MarcoPolo Jul 17, 2023
a25267b
Add mermaid diagrom for HTTP semantics vs transport
MarcoPolo Jul 17, 2023
3014b22
Grammar fixes
MarcoPolo Jul 17, 2023
f96359b
Lidel suggestions
MarcoPolo Jul 17, 2023
1e87960
Define where the libp2p-token will be
MarcoPolo Jul 17, 2023
d0f0d93
Grammar fix
MarcoPolo Jul 17, 2023
8fbd64a
Specify IX vs NX in auth scheme
MarcoPolo Jul 19, 2023
71415b0
Add SNI and HTTP_libp2p_token to Noise extensions
MarcoPolo Jul 19, 2023
4a03bb0
Reword Namespace section a bit
MarcoPolo Aug 2, 2023
877899d
Remove SNI and token from extensions
MarcoPolo Aug 2, 2023
d8850aa
update protocol name for IPFS gateway
marten-seemann Oct 4, 2023
78e8ca1
Be clear about no pipelining
MarcoPolo Mar 14, 2024
d30efda
Use SHOULD instead of MUST
MarcoPolo Mar 18, 2024
8628b5a
Update RFC for connection: close
MarcoPolo Apr 3, 2024
3c0ac40
Rename well-known
MarcoPolo Apr 3, 2024
75bc635
Add sentence on why POST and other mappings
MarcoPolo Apr 3, 2024
f95e4db
Sukun's review comments
MarcoPolo Apr 15, 2024
e3eb9dc
Small typo fixes
MarcoPolo Apr 15, 2024
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Jump to
Jump to file
Failed to load files.
Diff view
Diff view
133 changes: 133 additions & 0 deletions http/README.md
@@ -0,0 +1,133 @@
# HTTP

| Lifecycle Stage | Maturity | Status | Latest Revision |
| --------------- | ------------- | ------ | --------------- |
| 1A | Working Draft | Active | r0, 2023-01-23 |

Authors: [@marten-seemann], [@MarcoPolo]

Interest Group: [@lidel], [@thomaseizinger]

[@marten-seemann]: https://github.com/marten-seemann
[@MarcoPolo]: https://github.com/MarcoPolo
[@lidel]: https://github.com/lidel
[@thomaseizinger]: https://github.com/thomaseizinger

MarcoPolo marked this conversation as resolved.
Show resolved Hide resolved
## Introduction

This document defines how libp2p nodes can offer and use an HTTP transport alongside their other transports to support application protocols with HTTP semantics. This allows a wider variety of nodes to participate in the libp2p network, for example:
MarcoPolo marked this conversation as resolved.
Show resolved Hide resolved

- Browsers communicating with other libp2p nodes without needing a WebSocket, WebTransport, or WebRTC connection.
- HTTP only edge workers can run application protocols and respond to peers on the network.
- `curl` from the command line can make requests to other libp2p nodes.

The HTTP transport will also allow application protocols to make use of HTTP intermediaries such as HTTP caching, and layer 7 proxying and load balancing. This is all in addition to the existing features that libp2p provides such as:
MarcoPolo marked this conversation as resolved.
Show resolved Hide resolved

- Connectivity – Work on top of WebRTC, WebTransport, QUIC, TCP, or an HTTP transport.
- Hole punching – Work with peers behind NATs.
- Peer ID Authentication – Authenticate your peer by their libp2p peer id.
- Peer discovery – Learn about a peer given their peer id.

## HTTP Semantics vs Encodings vs Transport

HTTP is a bit of an overloaded term. This section aims to clarify what we’re talking about when we say “HTTP”.


```mermaid
graph TB
subgraph "HTTP Semantics"
HTTP
end
subgraph "Encoding"
HTTP1.1[HTTP/1.1]
HTTP2[HTTP/2]
HTTP3[HTTP/3]
end
subgraph "Transports"
Libp2p[libp2p streams]
HTTPTransport[HTTP transport]
end
HTTP --- HTTP1.1
HTTP --- HTTP1.1
HTTP1.1 --- Libp2p
HTTP --- HTTP2
HTTP --- HTTP3
HTTP1.1 --- HTTPTransport
HTTP2 --- HTTPTransport
HTTP3 --- HTTPTransport
Comment on lines +37 to +57
Copy link

Choose a reason for hiding this comment

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

h2c to show how it can be multiplexed and negociated in many ways (header compression, binary based protocol, ...) ?
Feel free to ignore this comment if you think it's making the graph too complex.

```

- *HTTP semantics* ([RFC 9110](https://www.rfc-editor.org/rfc/rfc9110.html)) is
the stateless application-level protocol that you work with when writing HTTP
apis (for example).

- *HTTP encoding* is the thing that takes your high level request/response
defined in terms of HTTP semantics and encodes it into a form that can be sent
over the wire.

- *HTTP transport* is the thing that takes your encoded reqeust/response and
sends it over the wire. For HTTP/1.1 and HTTP/2, this is a TCP+TLS connection.
For HTTP/3, this is a QUIC connection.

When this document says *HTTP* it is generally referring to *HTTP semantics*.

## Interoperability with existing HTTP systems

A goal of this spec is to allow libp2p to be able to interoperate with existing HTTP servers and clients. Care is taken in this document to not introduce anything that would break interoperability with existing systems.
Comment on lines +72 to +76
Copy link
Contributor

Choose a reason for hiding this comment

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

So this is a bit confusing to me. Above you are saying the you generally refer to HTTP semantics and the next sentence says that a goal is to interoperate with existing HTTP servers and clients which refers to the transport, correct?

Copy link
Contributor

Choose a reason for hiding this comment

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

Refers to both actually


## HTTP Transport

Nodes MUST use HTTPS (i.e., they MUST NOT use plaintext HTTP). It is RECOMMENDED to use HTTP/2 and HTTP/3.

Nodes signal support for their HTTP transport using the `/http` component in
their multiaddr. E.g., `/dns4/example.com/tls/http`. See the [HTTP multiaddr
MarcoPolo marked this conversation as resolved.
Show resolved Hide resolved
component spec](https://github.com/libp2p/specs/pull/550) for more details.

## Namespace

libp2p does not squat the global namespace. libp2p application protocols can be
discovered by the [well-known resource](https://www.rfc-editor.org/rfc/rfc8615)
`.well-known/libp2p`. This allows server operators to dynamically change the
Copy link
Member

Choose a reason for hiding this comment

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

We use well-known resources elsewhere, e.g. .well-known/libp2p-webtransport. Perhaps we should namespace this one too?

Suggested change
`.well-known/libp2p`. This allows server operators to dynamically change the
`.well-known/libp2p-http`. This allows server operators to dynamically change the

Copy link
Contributor

Choose a reason for hiding this comment

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

Aesthetically it's nicer if we aren't referencing HTTP multiple times. .well-known/libp2p is a HTTP resource. It's kind of like saying "ATM machine".

Technically this doesn't matter, but my vote is for .well-known/libp2p

Copy link
Member

Choose a reason for hiding this comment

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

I'm not sure I agree. Given we have other well-known resources, .well-known/libp2p is ambiguous, .well-known/libp2p-http is not.

Though yes, from a technical perspective it just has to be a predictable string with a low chance of collision.

Copy link
Contributor

@MarcoPolo MarcoPolo Mar 28, 2024

Choose a reason for hiding this comment

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

@lidel can you be our tie breaker? .well-known/libp2p vs .well-known/libp2p-http or something else

Copy link
Member

@lidel lidel Apr 3, 2024

Choose a reason for hiding this comment

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

Squatting .well-known/libp2p for a file may cause us problems in the future if we need to add something else.

I think if we want to avoid that, there is still time to can change it (this spec is still a draft), we should go with either .well-known/libp2p-http ..or make a directory and put protocol mapping configuration in .well-known/libp2p/http.

The latter (libp2p/foo) feels a bit better to me, as it creates libp2p-specific directory/namespace, which is easier to map via reverse proxies, but other than that it is mostly aesthetics.

If we don't want 'http' twice, could be .well-known/libp2p/protocols.

Copy link
Contributor

Choose a reason for hiding this comment

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

Just a note that this is a map, so we could always add more things in a backwards compatible way.

I like namespacing with .well-know/libp2p. This lets us get a single IANA URI suffix if we want while still giving us flexibility in what goes under it.

I'll make the change to .well-known/libp2p/protocols, thanks. Mark this one as another success for IPFS' Chief Renaming Officer.

Copy link
Member

Choose a reason for hiding this comment

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

Just to point out that we use .well-known/libp2p-webtransport for WebTransport things already, .well-known/libp2p-http would be consistent with that, .well-known/libp2p/protocols looks like something different.

Copy link
Contributor

Choose a reason for hiding this comment

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

This is different though. This is metadata about a peer's supported protocols. The well-know webtransport URI is about where to send the HTTP CONNECT request to.

Copy link
Member

Choose a reason for hiding this comment

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

I'm not sure it is, I think how you interact with the well-known resource is a detail of the protocol that's unrelated to the path it uses.

Copy link
Contributor

@MarcoPolo MarcoPolo Apr 3, 2024

Choose a reason for hiding this comment

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

The .well-known/libp2p/ suffix has the benefit of being a single thing we would need to register. Future versions of libp2p-webtransport may be placed under that suffix. Remember WebTransport isn't even out of draft status yet. We'll probably need to make a new multiaddr for webtransport-v1 just like we did with QUIC.

Any other well-known resource we would want could also fit under that suffix.

URLs of the application protocols offered, and not hard-code any assumptions how
a certain resource is meant to be interpreted.

```json

{
"protocols": {
"/kad/1.0.0": {"path": "/kademlia/"},
"/ipfs/gateway": {"path": "/"},
}
}
```

The resource contains a mapping of application protocols to a URL namespace. For
example, this configuration file would tell a client

1. That the Kademlia application protocol is available with prefix `/kademlia`
MarcoPolo marked this conversation as resolved.
Show resolved Hide resolved
and,
2. The [IPFS Trustless Gateway API](https://specs.ipfs.tech/http-gateways/trustless-gateway/) is mounted at `/`.
Copy link
Member

Choose a reason for hiding this comment

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

If I understand correctly, this only specifies the path but not the method (GET / POST) to use when accessing this protocol over HTTP and that's up to the specific protocol to define how to run it over HTTP?
If so, should we add this explainer in the spec?

Copy link
Member

Choose a reason for hiding this comment

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

Hm.. the methods will be specific to each protocol at each mount point, so not part of this spec?

Copy link
Member

Choose a reason for hiding this comment

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

That makes sense. As I see it there are two sections in this document:

  1. Running libp2p protocols over standard http like h2 or h3
  2. Running http protocols over libp2p streams.

So should we mention it in the specs that a libp2p protocol supporting http transport should specify the http method and headers to be used for the protocol. For the path they can expose it via the wellknown endpoint /.well-known/libp2p/protocols

Copy link
Contributor

Choose a reason for hiding this comment

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

I'm not sure I understand. An application protocol would be built using HTTP semantics, and that protocol would then be able to run on libp2p streams or "standard" http transports like h2, h3.

What do you mean by:

Running libp2p protocols over standard http like h2 or h3

This spec does not define how you would take an existing libp2p protocol and map it to HTTP semantics. That is best done by the specific protocol itself. But maybe I'm misunderstanding your point?


It is valid to expose a service at `/`. It is RECOMMENDED that implementations facilitate the coexistence of different service endpoints by ensuring that more specific URLs are resolved before less specific ones. For example, when registering handlers, more specific paths like `/kademlia/foo` should take precedence over less specific handler, such as `/`.

## Peer ID Authentication

When using the HTTP Transport, peer id authentication is optional. You only pay
MarcoPolo marked this conversation as resolved.
Show resolved Hide resolved
for it if you need it. This benefits use cases that don’t need peer
authentication (e.g., fetching content addressed data) or authenticate some
other way (not tied to libp2p peer ids).

Specific authentications schemes for authenticating Peer IDs will be defined in
MarcoPolo marked this conversation as resolved.
Show resolved Hide resolved
a future spec.

## Using HTTP semantics over stream transports
MarcoPolo marked this conversation as resolved.
Show resolved Hide resolved

Application protocols using HTTP semantics can run over any libp2p stream transport. Clients open a new stream using `/http/1.1` as the protocol identifer. Clients encode their HTTP request as an HTTP/1.1 message and send it over the stream. Clients parse the response as an HTTP/1.1 message and then close the stream. Clients SHOULD NOT pipeline requests over a single stream. Servers SHOULD set the [`Connection: close` header](https://datatracker.ietf.org/doc/html/rfc2616#section-14.10) to signal to clients that this is not a persistent connection.
MarcoPolo marked this conversation as resolved.
Show resolved Hide resolved

HTTP/1.1 is chosen as the minimum bar for interoperability, but other encodings of HTTP semantics are possible as well and may be specified in a future update.
Jorropo marked this conversation as resolved.
Show resolved Hide resolved

## Using other request-response semantics (not HTTP)

This document has focused on using HTTP semantics, but HTTP may not be the common divisor amongst all transports (current and future). It may be desirable to use some other request-response semantics for your application-level protocol, perhaps something like rust-libp2p’s [request-response](https://docs.rs/libp2p/0.52.1/libp2p/request_response/index.html) abstraction. Nothing specified in this document prohibits mapping other semantics onto HTTP semantics to keep the benefits of using an HTTP transport.

To support the simple request-response semantics, for example, the request MUST be encoded within a `POST` request to the proper URL (as defined in the Namespace section). The response is read from the body of the HTTP response. The client MUST authenticate the server and itself **before** making the request.
MarcoPolo marked this conversation as resolved.
Show resolved Hide resolved
MarcoPolo marked this conversation as resolved.
Show resolved Hide resolved