There is strange missmatch between key.marshal() and .marshalPublicKey

[Gozala]

Have not verified across the key types but at least on Ed25519PrivateKey / Ed25519PublicKey instances .marshal() returns different result from crypto.keys.marshal(key) which is unexpected.

What's worse though that crypto.keys.unmarshalPrivateKey(prk.marshall(), ...) errors.

I would expect that not to be a case.

[mkg20001]

.marshal() returns the data that is later wrapped in a protocol buffer that contains the type and the raw data returned from the .marshal() function by .marshalPrivateKey
That output can later be un-wrapped by .unmarshalPrivateKey

tl;dr Don't use .marshal() directly, use the functions from the crypto.keys APIs

[carsonfarmer]

Is there some good documentation around about interoperability with go-libp2p-crypto in this respect? I'm seeing that one needs to stick with crypto.keys.marshal*Key and crypto.keys. unmarshal*Key, but even this seems to lead to issues in some cases?