Resource conventions: Compute Instance vs Compute Unit unclear

[Oberon00]

See https://github.com/open-telemetry/opentelemetry-specification/tree/c04041e16c9571b7ee5cfd5f102fc15e94ffacd7/specification/resource/semantic_conventions#compute-unit and https://github.com/open-telemetry/opentelemetry-specification/tree/c04041e16c9571b7ee5cfd5f102fc15e94ffacd7/specification/resource/semantic_conventions#compute-instance

Compute Unit currently only has Host as an entry.

On the other hand a Serverless Function (FaaS) is currently a Compute Unit but has attributes like faas.instance, suggesting that it might be an instance. Also containers are the runtime incarnations/instances of images. Probably we should just merge these sections.

To bring in even more confusion, under "Deployment Services" (https://github.com/open-telemetry/opentelemetry-specification/tree/c04041e16c9571b7ee5cfd5f102fc15e94ffacd7/specification/resource/semantic_conventions#deployment-service) we find k8s.pod.name and according to https://kubernetes.io/docs/concepts/workloads/pods/pod/ "A Pod (as in a pod of whales or pea pod) is a group of one or more containers" so if a container is an instance, so is a POD.

[andrewhsu]

from the spec sig mtg today, discussed this issue and bumped priority to p3 for the required-for-ga release

[Oberon00]

Relatedly the TOC hierarchy for "Deployment" is a bit odd: https://github.com/open-telemetry/opentelemetry-specification/pull/851#discussion_r475660328

[Oberon00]

This is just an editorial change (changing section headers or merging/splitting section), I think this is after GA.