lnd should scan across account/index BIP32 derivations

[alecov]

From lncli wallet accounts list I gather lnd is scanning the following derivations:

m/49'/0'/0' (segwit p2sh)
m/84'/0'/0' (segwit native)
m/86'/0'/0' (taproot)
m/1017'/0'/X' (???)

Why doesn't it scan the standard account/index at each level, eg. m/84'/0'/0'/{0,1}/X?

lnd misses many wallet transactions (account = 0) and wallet changes (at account = 1) since it only looks at the root address (which has usually nothing).

[guggero]

lnd itself does not use any accounts other than account number 0 internally, at least for any standard key scope (1017 is the lnd internal scope for channel/backup/watchtower related keys, there the account number is called the key family, see https://github.com/lightningnetwork/lnd/blob/master/keychain/derivation.go#L67.

The "account" level is the third item in the path. So when you refer to m/84'/0'/0'/{0,1}/X, the {0,1} is the change (which as per BIP044 is always only 0 or 1) and X is the key index. lnd does scan more than just index 0, the value recovery_window in InitWallet dictates how many keys.

With all that said, I'm not sure I fully understand your question. lnd should not "miss" any keys, if they were created with lnd in the first place. It is not recommended to use a seed created by lnd with any other wallet.

[alecov]

With all that said, I'm not sure I fully understand your question.
It is not recommended to use a seed created by lnd with any other wallet.

Why not? Is there a technical reason for this?

I ask specifically why lnd does not scan for m/84'/0'/0'/0/X and m/84'/0'/0'/1/X (for changes) as most or every other wallet software does. This kind of interop is important. Missing that makes it necessary to on-chain transfer between standard BIP32 and lnd-specific wallets.

[guggero]

lnd does scan for addresses on the m/84'/0'/0'/0/X path. But I guess the question is: When does lnd scan for addresses. During normal operation, lnd will only keep track of addresses up to index X for which it has derived addresses by itself (meaning you called lncli newaddress X number of times). So if you derive X+1 in another wallet and send funds to it without also calling lncli newaddress first, lnd will not "see" the funds on that address.
But: If you do a wallet restore from seed, then lnd will scan for recovery_window (so 2500 by default) addresses, so up to m/84'/0'/0'/0/2500.

[alecov]

During normal operation, lnd will only keep track of addresses up to index X for which it has derived addresses by itself (meaning you called lncli newaddress X number of times).

BIP-44 suggests an address gap of 20, so I believe it would make sense for lnd to scan 20 addresses from the latest index X known to lnd during boot (not from initial restore, but after shutdown & restart) and set X accordingly.

But: If you do a wallet restore from seed, then lnd will scan for recovery_window (so 2500 by default) addresses, so up to m/84'/0'/0'/0/2500.

Thanks for clarifying. I was aware of recovery_window but still, I've restored from a root xprv and lnd shows a zero wallet balance. Does the above scan apply for xprv recovery also?

Also, is the seed recovery option BIP-39 compliant or a different scheme? Are 12-word phrases supported?

[guggero]

Unfortunately the btcwallet library we use for the on-chain handling is a bit dated and doesn't have that many options for address derivation and watching them on chain.

lnd uses the aezeed format which uses BIP39 words to encode the seed data as a mnemonic phrase. But the content is not compatible with other seeds.