Setup instrumentation library

[dmathieu]

Following @mwear's comment in #261, this is a proposal to implement the instrumentation library pattern, as I understand it from open-telemetry/oteps/pull/84.

[mwear]

Thanks for looking taking this on @dmathieu.

The instrumentation library should be the name and version of the instrumentation, not the underlying library. It really should be the name of the thing generating a span. Sorry for not explaining this better before. The reason for that, is the longer term goal is for OpenTelemetry to be built in to libraries, such as excon. The library itself would use excon and its version for spans it creates, while the instrumentation would identify itself as the excon adapter, and use the adapter version.

What you're doing in the PR is definitely on the right track. I can think we can do the following:

• initialize the instrumentation library in Tracer using the name and version arguments, no need to set this explicitly from instrumentation
  • the name will be the adapter name
  • the version will be the adapter version
• pipe the instrumentation library through to spans, and then span data as you are

Another note, is that the resource we're storing on the Tracer (with name and version) is not what we want longer term. I would leave it as is so @robertlaurin can use that to pipe through a process-level resource from #263.

[dmathieu]

Thank you for the feedback @mwear. I've updated my PR to reflect your comment.

[mwear]

Thanks @dmathieu. This looks great!

I just wanted to double check with @robertlaurin. Will you be repurposing the library_resource to be just a resource in #263? If so, we should just leave it as is. If not, we should create an issue to remove it.

[mwear]

There could be changes on the horizon if this is accepted: open-telemetry/opentelemetry-proto#150. I'm merging this PR as it represents the current state of the specification.