idea: a lambda / OpenWhisk style programming model?

[jstrachan]

so folks like the simplicity of the Amazon Lambda and OpenWhisk kinda programming models. e.g. here's some examples https://github.com/awslabs/lambda-refarch-iotbackend

What might be nice is to let folks pick any language (go lang, nodejs, swift, rust, whatever) to write some code ('a function') and then bind it to events via a subscription.

e.g. folks could write a simple REST service in nodejs or swift like these [examples](https://github.com/fabric8-quickstarts/swift-example].

Then folks could define their subscription rules in a declarative file somewhere:

---
# triggers for local HTTP URIs based on subscriptions to external URLs (which use Camel components)
/foo: 
  uri: "activemq:Some.Queue?concurrentConsumers=20"
/bar:
  uri: "timer:foo?period=5000"

then they write a normal nodejs / swift / whatever app that listens to events foo and bar. Then the build process deploys a sidecar container which is implemented using Apache Camel that subscribes to the URIs and invokes the local endpoints.

e.g. the subscription rules are converted into simple routes something like...

from("activemq:Some.Queue?concurrentConsumers=20").to("http://localhost:8080/foo")

Maybe using JSON marshalling in there or something...

We could hide the HTTP plumbing behind a little language/framework specific bit of glue; so the go/node/swift programmer just writes native language functions called foo and bar and the build system automatically binds those functions to local HTTP endpoints.

After the build system, folks get a docker image for their functions and a ReplicationController which in the pod template has a sidecar container that implements the subscription manager. Also we should try get the sidecar container to figure out auto-scaling; as the subscription should hopefully be able to figure out a good algorithm to autoscale the pods based on the upstream event size. e.g. for HTTP use the metrics of HTTP volume; for messages, use the queue depth and so forth

[jstrachan]

gonna mark this one as fixed since we have funktion now:
https://github.com/fabric8io/funktion/