psbt: satisfy any ToPublicKey key

[darosior]

The satisfier was only implemented for bitcoin::PublicKey and could not be reimplemented for a more generalistic key downstream.

Also, it makes sense after #131 (i have a rebased version for my own needs here https://github.com/darosior/rust-miniscript/commits/satisfy_xpubs). All signatures are looked for by raw public key, and thus should be added to the mapping with to_public_key() (which derives the one pointed by the DescriptorPublicKey in #131).

Could have rebased it on #131 to add a test xpub-descriptor, but seemed overkill as the change is pretty straightforward.

[sanket1729]

As discussed on IRC, we probably need to implement Satisfier on a structure that also has information for (nSequence, nLocktime) etc. The current implementation on psbt::Input does not have sufficient information to implement the satisfier for after and older.

We require the complete psbt to get the nSequence, nLockTime fields.

[sanket1729]

I guess this PR is independent of the above issue, we can merge this

[apoelstra]

I'm actually gonna concept NACK this, because it plays badly with Rust's trait coherency rules. It prevents users from implementing Satisfier themselves on their types if those types also implement MiniscriptKey.

[apoelstra]

@darosior why do you say it "could not be reimplemented for a more generalistic key downstream"?

[darosior]

@darosior why do you say it "could not be reimplemented for a more generalistic key downstream"?

IIRC: As it's implemented for bitcoin::PublickKey which i use and is MiniscriptKey + ToPublicKey, i had an error trying to implement it for a <Pk: Miniscript + ToPublicKey> as it was considered to reimplement a foreign trait.

[apoelstra]

Right, if you try to implement it for all Pk: Miniscript + ToPublicKey the compiler will prevent it because it may cause breakage with your other dependencies which might have conflicting implementations ...it's legal for us to do it in rust-miniscript but I think it's still likely to be a problem for users.

[sanket1729]

@apoelstra Why do you think that that downstream user would need another implementation of Satisfier for psbt? As far as I can see, there can only be one way of doing it?

[apoelstra]

Oh, oops, I misread this PR. I thought we were implementing on Pk itself, not on psbt::Input.

[apoelstra]

concept ACK then :)

[darosior]

concept ACK then :)

:)

More generally, what about we either:

• implement the default satisfier for every key, if downstream users want an exotic implementation they should use a newtype
• don't implement the satisfier at all, leaving the implementation design up to the user

As the middle ground seems confusing to me (you likely don't need a specific implem for just one key type but rather for all MiniscriptKey + ToPublicKey keys).

[apoelstra]

Is the first option what's being proposed in this PR?

[darosior]

Conceptually, i think so. Was not the original motivation though.