Capture net.host.name and net.host.port for SERVER spans?

[trask]

I think we should probably at least capture net.host.name and net.host.port when there's no http.host (i.e. HTTP 1.0) in order to satisfy:

Additional attribute requirements: At least one of the following sets of attributes is required:

• http.scheme, http.host, http.target
• http.scheme, http.server_name, net.host.port, http.target
• http.scheme, net.host.name, net.host.port, http.target
• http.url

I'm guessing we would add these to NetServerAttributesGetter(?)

And then the question is, should we:

• somehow make capturing net.host.name and net.host.port conditional on absence of http.host, or
• always capture net.host.name and net.host.port

[mateuszrzeszutek]

• somehow make capturing net.host.name and net.host.port conditional on absence of http.host

Hmm, that'd probably make net attributes extractors depend on the http getter (if we want to preserve the order of importance in the attribute sets list). It would probably be cleaner (from the code perspective) to capture both net.host.* and http.host -- also, we do a similar thing in the HTTP client instrumentations, we capture both http.url and net.peer.*.

[mateuszrzeszutek]

I think it's been implemented already in one of my HTTP semconv PRs.