[bitcoin-dev] BIP Proposal: Wallet Interface

[monokh <mnokhb@gmail.com>]

[-- Attachment #1: Type: text/plain, Size: 6144 bytes --]

Hi

This is a first draft of a BIP we intend to submit. The main intention is
to define a simple interface that wallets and applications can agree on
that would cover the vast majority of use cases. This can enable writing
bitcoin applications (e.g. time lock, multi sig) on the web that can be
seamlessly used with any compatible wallets. We have implementations of
such examples but I don't want to turn this thread into a promotion and
rather focus on the spec.

Appreciate input from the list. Please share if there are existing efforts,
relevant specs or use cases.

------------------------------

A wallet interface specification for bitcoin applications

## Abstract

This BIP describes an API for Bitcoin wallets and applications as a
standard.

## Summary

Bitcoin wallets should expose their address derivation and signing
functions to external applications. The interface would be expressed as
follows in javascript:

```
{
// Wallet Metadata
wallet: {
name: 'Bitcoin Core'
},

// Request access to the wallet for the current host
async enable: (),

// Request addresses and signatures from wallet
async request ({ method, params })
}
```

In the web context the interface could be exposed at the top level of a
webpage, for example under `window.bitcoin`. However this spec does not
intend to define any standards for how and where the interfaces should be
exposed.

## Motivation

Due to the seldom available APIs exposed by wallets, applications (web or
otherwise) are limited in how they are able to interact. Generally only
simple sends have been available. A more robust API that introduces other
requests will promote richer Bitcoin applications.

Additionally, wallet APIs have frequently included inconsistencies in their
interfaces and behaviour. This has required applications to build and
maintain a separate client for each wallet, increasing the risk of bugs and
unintended behaviour as well as being a limiting factor for the adoption of
usable bitcoin applications.

With a standardised wallet API:

- Wallets have a clear API to implement
- Applications have a clear expectation of wallet interface and behaviour
- Applications become agnostic to the wallet specifics, increasing choice
for users

If more wallets implement the specification, applications will be developed
more confidently by benefiting from the wallet interoperability. This
creates a positive feedback loop.

## Specification

For simplicity, the interface is defined in the context of web applications
running in the browser (JS) however, they are simple enough to be easily
implemented in other contexts.

### General Rules

- For sensitive functions (e.g. signing), wallet software should always
prompt the user for confirmation

### Types

**UserDeniedError**
An error type indicating that the application's request has been denied by
the user
Type: Error

**Hex**
Type: String
Example:
`"0000000000000000000a24677957d1e50d70e67c513d220dbe8868c4c3aefc08"`

**Address**
Address details
Type: Object
Example:

```
{
"address": "bc1qn0fqlzamcfuahq6xuujrq08ex7e26agt20gexs",
"publicKey":
"02ad58c0dced71a236f4073c3b6f0ee27dde6fe96978e9a9c9500172e3f1886e5a",
"derivationPath": "84'/1'/0'/0/0"
}
```

### API

The wallet must implement the following methods.

**enable**

The enable call prompts the user for access to the wallet.

If successful, it resolves to an address (`**Address**` type) of the
wallet. Typically the first external address to be used as an identity.

**`UserDeniedError`** will be thrown if the request is rejected.

**request**

The request method must take one parameter in the following format:

```
{
"method": "wallet_methodName",
"params": ["foo", "bar", "baz"]
}
```

For a list of mandatory methods see Table

The wallet should reject request calls unless `enable` has been resolved.

Sensitive requests that involve signing should always prompt the user for
confirmation

On success the request should resolve to the response as defined in the
method table.

**`UserDeniedError`** will be thrown if the request is rejected.

**Mandatory methods**

method: `wallet_getAddresses` params: [`index = 0, numAddresses = 1, change
= false`]
return: `[ Address ]`
error: UserDeniedError

method: `wallet_signMessage` params: `[ message, address ]`
return: Signature `Hex`
error: UserDeniedError

method: `wallet_signPSBT` params: `[ [psbtBase64, inputIndex, address] ]`
return: `psbtBase64`
error: UserDeniedError

method: `wallet_getConnectedNetwork` params: `[]`
return: Network object `mainnet` | `testnet` | `regetst`
error: UserDeniedError

## Rationale

The purpose of the API is to expose a set of commonly used wallet
operations. In addition, it should be flexible enough to serve for other
requests such as node RPC calls.

**Why is there a singular request call instead of named methods?**
The transport layer for the requests cannot be assumed, therefore it is
much more flexible to instead define an abstract format.

**Why are the mandatory methods so primitive? Where is getBalance,
getUtxos, ... ?**
A wallet need not worry about providing every possible scenario for usage.
The primitives of keys and signing can expose enough to applications to do
the rest. Applications should have flexibility in how they implement these
functions. It is the role of a library rather than the wallet.

## Security Implications

Great care should be taken when exposing wallet functionality externally as
the security and privacy of the user is at risk.

### Signing

Operations that trigger signing using private keys should be guarded behind
confirmation screens where the user is fully aware of the nature of the
transaction. In the example of a PSBT signature request, the outputs, the
inputs and which key is being used should be clearly marked.

### Privacy

Some api methods expose metadata about the user, such as public keys.
Depending on how privacy focused the wallet intends to be, the wallet could
protect these behind a confirmation. Commonly the wallet just needs to give
the origin access to all of its public keys, however it could also allow
the option to expose only selected derivation paths.

-monokh

[-- Attachment #2: Type: text/html, Size: 6884 bytes --]