Should I use Command or Message to send commands to AMPS?

You should (almost) always use a Command and execute or executeAsync rather than a Message.

The Command interface is designed to be equivalent to the named commands, and includes error checking and support from the Client and HAClient infrastructure. The Message/send interface intentionally bypasses most of this infrastructure.

For example, when you use an HAClient to subscribe to a topic using the Command interface or the named convenience method, the HAClient registers the subscription in the MemorySubscriptionManager so the subscription can automatically be resumed on failover. In contrast, if you simply send a Message to AMPS, the HAClient does not track the subscription, and the subscription will not be resumed on failover.

The 60East engineers recommend using the Command interface. If your application has the need to bypass most of the Client infrastructure for some reason, the Message/send interface provides the flexibility to do so (at the expense of managing any part of that you need in your application). For most uses, however, the Command infrastructure provides the same capabilities to customize the command to AMPS, with full support from the client infrastructure.