Damien

It's possible that we've been asked for shim tokens but never managed to fully start the container for some reason. When the runtime tries to clean up in the error...

megacheck found a bit of dead code, we should fix this and re-enable megacheck.

The proxy API has notifications. Right now only the shim written in C needs those notifications (ProcessExited). If we ever need notifications from a client written in go, we'll need...

gometalinter takes a long time to get and compile. We should be able to cache the source and binary and always do a `gometalinter --install --update`

Thinking about it, runtime and shim should be able to always give us the container they are logging for.

James commented in https://github.com/clearcontainers/proxy/pull/61 that it'd be useful to log the pid of the actors involved.

It'd be useful to have the hypervisor pid in the aggregated logs. One good option to know about qemu's pid is for the runtime/virtcontainers to tell us in RegisterVM.

Currently the proxy really only knows about containers, because of its root in docker. AttachVM takes an argument called containerID, but that's really a podID. We may or may not...

We have a few data structures to garbage collect if runtime & shim don't behave: - If a token is never claimed by a shim, it's never garbage collected at...

It may be nice to control the amount of logging remotely (maybe only if the proxy has been started with a debug option to avoid clients abusing of that facility...