Ethereum, and uPort, is entering a multi-chain world. As end users increasingly interact with multiple chains, on Ethereum or elsewhere, the risk of users/servers inadvertently transferring value from an address on network X to an address on network Y is growing. This could result in monetary loss. Since uPort is switching to a new test network, we need to solve this issue urgently.
The Bitcoin protocol uses Base58Check encoding to prevent users from sending value off-network, but the ethereum ecosystem has used a raw hex version of the address instead.
My proposal is inspired by the Base58Check encoding as well as EIP77 but also specifies a network identifier, which allows us to programmatically extract the network used by an address as well as provide a visual indicator of the network used.
The following items are encoded:
- 1 byte version number currently
1
- network id or four bytes of genesis block hash (or both)
- actual address data
- Four bytes (32 bits) of SHA3-based error checking code (digest of the version, network and payload)
Then base58 encoding is applied to the end result. The end result is fairly complete but still extendible in the future. We could start by simply using the network id and replace it with the genesis block hash and other meta data in the future.
This works with ethereum blockchains, but can easily be extended to other blockchains or even non-blockchain identifiers in the future. It would also be straightforward to add further details specifying which fork etc.
This can be implemented very easily with few dependencies. It would be trivial to use this to add multichain support to uport-lite for example. Thus even allowing (if desired) the interchange of JWT's verified on different networks.
The following Ethereum hex encoded address 0x00521965e7bd230323c423d96c657db5b79d099f
could be encoded as follows
- main-net:
2nQtiQG6Cgm1GYTBaaKAgr76uY7iSexUkqX
- ropsten:
2oDZvNUgn77w2BKTkd9qKpMeUo8EL94QL5V
- kovan:
34ukSmiK1oA1C5Du8aWpkjFGALoH7nsHeDX
- infuranet:
9Xy8yQpdeCNSPGQ9jwTha9MRSb2QJ8HYzf1u
It would be trivial to add shard ids, fork descriptors (block number and hash) etc to the address. It would also be trivial to encode other kinds of identities that don't correspond directly to an address on a chain.
> var mnid = require('mnid')
> mnid.encode({
network: '0x1', // the hex encoded network id or for private chains the hex encoded first 4 bytes of the genesis hash
address: '0x00521965e7bd230323c423d96c657db5b79d099f'
})
'2nQtiQG6Cgm1GYTBaaKAgr76uY7iSexUkqX'
> mnid.decode('2nQtiQG6Cgm1GYTBaaKAgr76uY7iSexUkqX')
{ network: '0x1',
address: '0x00521965e7bd230323c423d96c657db5b79d099f' }
// Check if string is a valid MNID
> mnid.isMNID('2nQtiQG6Cgm1GYTBaaKAgr76uY7iSexUkqX')
true
> mnid.isMNID('0x00521965e7bd230323c423d96c657db5b79d099f')
false
> mnid.isMNID('1GbVUSW5WJmRCpaCJ4hanUny77oDaWW4to')
false
> mnid.isMNID('QmXuNqXmrkxs4WhTDC2GCnXEep4LUD87bu97LQMn1rkxmQ')
false
Bitcoin's encoding consists of the following 3 items:
- Version prefix - Used more as a type and network field. See list.
- Payload (eg. hash of public key)
- Four bytes (32 bits) of SHA256-based error checking code (digest of the version and payload)
The whole thing is base58 encoded for compactness and URL safety.
The version prefix allows humans to visually recognize the address type from the first few characters in the string. The error checking code ensures that there aren't any obvious errors in the address.
A previous attempt at solving this for ethereum is found in EIP 77 which is similar to Base58Check:
- 1 flag byte - currently undefined. I suppose this could be used to pick a chain. But 1 byte does not seem enough
- Payload (eg. hash of public key)
- Four bytes (32 bits) of SHA3-based error error checking code (digest of the version and payload)