Support 'EXTERNAL' for TransactionInput.ScriptType

[RHavar]

The firmware doesn't appear to support 'EXTERNAL' for TransactionInput.ScriptType, which is essential for coinjoins. I'm trying to build a simple bip79 front-end for trezor, but this appears to make it impossible.

Comment from issue: trezor/connect#247

prusnak wrote:

If you want to try implementing it, you should add another case when txi.script_type == InputScriptType.EXTERNAL here:
https://github.com/trezor/trezor-core/blob/b9926a9fffe8a593beb3adb79e5fc429233e41a3/src/apps/wallet/sign_tx/signing.py#L148

[stevenroose]

I'd also very much appreciate EXTERNAL. Currently my workaround is to just put SPENDWITNESS and then remove the signature again afterwards.

[prusnak]

This is not planned at the moment. The issue will be solved via PSBT support which will be added in the future: https://github.com/trezor/trezor-core/issues/410

[matejcik]

Reopening as PSBT support solves a slightly different problem.

To get to the same page: EXTERNAL script type is supposed to mark an input to not be signed by Trezor. Instead, the user explicitly provides a scriptSig which is used when building the transaction.

For some usecases, a possible workaround is using SPENDWITNESS script type, allow Trezor to sign the input, and then replace the Trezor-generated signature in the result.

[matejcik]

For the record, this might end up a prerequisite for signing PSBTs, if Trezor doesn't parse the binary blobs itself. A partially signed transaction might already contain signatures from other sources.

This also means some, uh, challenges, on the UI side. How do we display a transaction where we're sending 1 BTC out of 3 BTC total?

[stevenroose]

Partially signing transaction. Do you want to sign off on spending 4.2 BTC from your wallet in the transaction totalling 42 BTC?
[y]
This transaction sends 22 BTC to address 123...
[enter]
This transaction sends 20 BTC to address 156...
[enter]
Confirm?
[y]

Or otherwise perhaps
Partially signing transaction that sends 42 BTC in total to the following addresses...
[enter]
This transaction sends 22 BTC to address 123...
[enter]
This transaction sends 20 BTC to address 156...
[enter]
Do you want to contribute 4.2 BTC from your wallet for this transaction?
[y]

[tsusanka]

I'm moving this to backlog, but we'll most likely include this into the next milestone.

[prusnak]

As per https://lists.linuxfoundation.org/pipermail/bitcoin-dev/2017-August/014843.html it is necessary to add an ownership proof to every EXTERNAL input.

[matejcik]

It is still possible to sign EXTERNAL inputs without ownership proofs, given that they are treated as risky and confirmed explicitly by the user.

This (esp. the segwit part) also means that we would require prev_tx data for every EXTERNAL input, so that we can validate the ownership proof signature and validate the amount for segwit inputs. Not great - maybe the ownership proof could be extended to cover the amount too?

In any case this probably stops this feature until we have some sort of standard for the proof of ownership.

[prusnak]

Let's refactor the signing code first: #617

[tsusanka]

Implementing SLIP-19 is needed for this. Moved to #1052.