+/// Signing services could be implemented on a hardware wallet. In this case,
+/// the current ChannelKeys would be a front-end on top of a communication
+/// channel connected to your secure device and lightning key material wouldn't
+/// reside on a hot server. Nevertheless, a this deployment would still need
+/// to trust the ChannelManager to avoid loss of funds as this latest component
+/// could ask to sign commitment transaction with HTLCs paying to attacker pubkeys.
+///
+/// A more secure iteration would be to use hashlock (or payment points) to pair
+/// invoice/incoming HTLCs with outgoing HTLCs to implement a no-trust-ChannelManager
+/// at the price of more state and computation on the hardware wallet side. In the future,
+/// we are looking forward to design such interface.
+///
+/// In any case, ChannelMonitor or fallback watchtowers are always going to be trusted
+/// to act, as liveness and breach reply correctness are always going to be hard requirements
+/// of LN security model, orthogonal of key management issues.
+///