Eirik A

This would require maintaining a map of `ObjectRef` -> `observedGeneration` (int) inside a watcher / reflector and checking if it changed before bubbling up the event (first added event would...

PoC in https://github.com/kube-rs/controller-rs/commit/7e2a15274d0ef99a3b7c841446bc104535ed2118 using the impl trait magic that @teozkr brought for runtime conditions. I think this is viable in terms of inclusion, but there's probably some cleanups. Feel free...

Ah, one benefit on the `client-go` side is less arguments over its main `Patch` type which requires more. The new ones implicitly handle the `.metadata.name` (by forcing it in the...

The KubeCon America talk; [SIG API Machinery Deep Dive](https://www.youtube.com/watch?v=oiC2w1PVjrQ) goes through why these structs (with full optionals) are needed (in the first half of that talk); some required fields get...

It's possible you are on to something. Been a while since I looked at it. But history wise: The 295 is a hard limit request __client__ limit (the value after...

> Now we have some orphaned objects that would no longer validate, but which have effectively been grandfathered in. watcher sees the error and fails immediately, swallowing all objects if...

that sounds great to me! if you have time, then happy to take a pr on this.

This is not really "fixed" but worked around in `master`. We now connect to the apiserver using cluster dns via solution suggested in #587 .

Outside the cluster you can workaround it by tweaking `/etc/hosts` and making a non-ip alias. Not sure that will work for all cases though. I think you're right and that...

A reason that we use `ObjectReference` inside the `Recorder` is because the [`Event` type embeds the ref](https://docs.rs/k8s-openapi/latest/k8s_openapi/api/core/v1/struct.Event.html) inside the spec as the involved object, so we have to convert to...