I wonder if it's correct to use Vec<u8> instead of ScriptBuf. If I use ScriptBuf I need to use a generic, which would percolate up until NetworkMessage, which would be a big breaking change.

I wonder if it's correct to use Vec<u8> instead of ScriptBuf. If I use ScriptBuf I need to use a generic, which would percolate up until NetworkMessage, which would be a big breaking change.

You mean for leaf data? That is always a script pubkey, can't you use the tag for it and avoid being generic?

On ReconstructableScript. If the tag is 0x00 we need to send the actual script (see https://github.com/bitcoin/bips/blob/bd1e2425872450b4b9d80cdcb47874d9659a3bda/bip-0183.md#reconstructable-script):

pub struct ReconstructableScript {
    /// The kind of locking script the UTXO is locked to.
    pub tag: ReconstructableScriptTag,
    /// The actual script, if [`ReconstructableScriptTag`] is of kind `Other`.
    pub script: Option<Vec<u8>>,
}

I've implemented this. However, you can use ScriptBuff<ScriptPubkey> safely.

Can you just use a Rust enum? This looks like a manual implementation of a tagged union, which is exactly what a Rust enum is.

Can you just use a Rust enum? This looks like a manual implementation of a tagged union, which is exactly what a Rust enum is.

Makes sense, this would also prevent the illegal state where the tag is != 0x00 and a script exists.

For scripts, is the script always a ScriptPubKey (locking script)? Are there any constraints (size, canonicality) enforced?

Where the hell are these test vectors?

@Davidson-Souza Where the hell are these test vectors?