My general thoughts on Resque Next is that it should be more of a protocol and
an interface than an end-all-be-all solution for delayed work in Ruby, that it
should provide at least one implementation for each of the interfaces, but that
it should leave further implementation up to third-parties and separate
include-if-you-want-it gems (that may be published to either the resque org or
elsewhere).
This includes the dependency on Redis. When a consumer out-grows Redis and
needs to use a queueing service that has actual, non-emulated queueing semantics,
they shouldn't have to rewrite all of their worker code and start over from