spec: Come up with a better sender public key header.

[Yawning]

The spec(s) currently specify variations of X-Panoramix-Sender-Identity-Key as the sender public key header.

This needs to be formalized sooner rather than later, and as far as I can tell needs the following adjustments:

• Rename to be in the X-Katzenpost namespace.
• Shortened down such that the full header will be under 78 bytes (excluding the trailing CR/LF), to avoid wrapping (See: RFC 5322 2.1.1).

[meskio]

X-Katzenpost-Identity-Key?

We might want in the future to be able to gossip keys of other recipients in a multiparty conversation. If we don't add addresses to this header might get confusing.

X-Katzenpost-Sender-Key?

In the context of a header I think is obvious that the key is an identity key.

[Yawning]

I think X-Katzenpost-Sender-Key is reasonable.

I would say that X-Katzenpost-Sender-Identity-Key is also ok, but it annoys me that the last 3 characters of the Base64 encoded public key end up on a new line.

[moba]

Why not without the -Key? It is obvious from the context that it is a key, representing a sender (identity), so it is kind of redundant.

[Yawning]

X-Katzenpost-Sender is the most concise and to the point (and is certainly going to be the most common). Anything that wants to add more headers with keys will need to disambiguate themselves somehow anyway, so those can be the things with the longer headers.

[Yawning]

By fiat, this now will be X-Katzenpost-Sender.