default IdP metadata for Shib IdP V3

[Eric Goodman]

>I'm not a cryptography expert, but there has never been any argument I've seen suggesting 
>that's a misguided idea. Every scheme and best practice out there urges separation of keys.

Not to get too off topic here (probably too late) but I haven't generally heard recommendations for separation of Signature and Encryption keys presuming we're talking about the inverse operations on similar messages (so technically my messages signature and *de*cryption keys being the same). Just as a simple example, most PGP use cases I'm aware of use just the one key for both operations.

Separation of Signature/Decryption keys and TLS keys I'm more familiar with. Also separating the TLS "server" keys vs. TLS "client" certificates (though I think the latter is more pragmatically because a "typical" web server is not considered to initiate outbound connections, but rather separate processes do that, and they have their own "client" identities). 

I don't see any particular problem with having or having the ability to define separate signature and decryption keys (other than key management complexity), just haven't heard it as a "best practice".

--- Eric