Closed c0c0n3 closed 5 years ago
From my point-of-view, the two top priorities for backends should cover:
Don't need to go too much farther than that initially.
Most definitely yes! In fact, I need to have the single VM install option to support the deployment model here at MRI. Re: queue distribution, this is a low hanging fruit, so we should have that too. (The default being the all-in-one option like you said.)
The messaging API is built on top of a few primitive interfaces---classes in core.msg package. A separate queue provider package provides an implementation of those primitive interfaces using Artemis---classes in providers.q package. We should make the
q
package pluggable so that we can use different implementations along the current one based on Artemis.So for example, going forward we should be able to easily add a
q
implementation backed by RabbitMQ or even one that uses (abuses?) Vert.x's event bus to provide an implementation of the abstract communication channels in themsg
package.