server.address or peer.addres may refer to service by the same across namespaces

[esara]

In the current implementation, the server address or the client address - when resolved from k8s - is only the short name (not fully qualified)

If the client and/or the server are in different namespaces - while the resource attributes spells out the namespace, it is not obvious if the span.client or span.server refers to a client/server in the same namespace or another namespace

[grcevski]

Thanks for reporting this Endre, I am curious if this is desired for traces and metrics or just traces?

[esara]

I am primarily looking at traces with beyla. In this case specifically, the nature of a span to show the resource and span attributes, often referring to the client as the resource and the server as the server address in the span attributes.

And while making assumption that communication is often within a namespace is bad, if the span is created in a crossnamespace span, it is not clear what server the server address is referring to.

For metrics, if it is just server metrics metrics, there would be no confusion, because it is only one side.