Get inputs public keys via `getrawtransaction` RPC

[louisinger]

For optimizations purposes, some of the previous output are not fetch while extracting public keys. Thus, silentiumd is using the witness data to identify input types. This is not safe to rely on witness, the scriptpubkey should always be verified before extracting the public key.

[knocte]

This is not safe

What do you mean exactly? What's the worst case scenario?

[louisinger]

BIP352 defines "eligible transaction for silent payment v0" with the following rule:

The transaction does not spend an output with SegWit version > 1

However, extracting the public key of an input with the witness data only could break this rule because we don't know how the next segwit versions will look like.
In other words, let's imagine a segwit v2 script type is released, allowing a witness data to be similar to a P2PKH. The current silentium code will detect the input as P2PKH and thus mark the transaction eligible for a silent payment v0.

So I would say that it's not urgent to fix this because segwit v2 is not for tomorrow (same for silent payment v1). Also, without this kind of optimization, the time to index a block increases a lot because of getrawtransaction RPC performance.