Arvados

Because arv-copy expects to work with multiple clusters, it needs to pull credentials from multiple sources. To get credentials for the source cluster, it looks at the source object UUID, and searches for a configuration file named after the cluster prefix in that UUID, like xyzzy.conf.

It would be cool if all of our CLI tools worked this way: given that it has an object UUID it's editing, or a project UUID to put a new object in, it should get the cluster ID from that UUID and look for credentials in xyzzy.conf before it falls back to settings.conf. This would make it easier to work across multiple clusters in parallel without the need for separate tooling.