Discussion: priority on environment variable vs compiling time value

[TommyCpp]

Most of our creates generally allow two ways to configure values:

• Use environment variable fetched during runtime
• User provided value during compiling time(with_xxx)

However, this is unclear when both of the configurations present which one should take priority. Regardless we should align on this and use it as a tenet across all our creates.

cc @open-telemetry/rust-approvers

[TommyCpp]

We discussed it in weekly meetings. Some notes

• Compiling time value
  • @TommyCpp @cijothomas agree with this
  • @shaun-cox is not objecting if we advise users to not configure anything in code and leave it to ENV.
  • NET, Go, Python, follow this.
• Environment variables
  • @shaun-cox agrees this
  • C++ follows this

[djc]

I find it very confusing to make the compiled-in value take priority over the run-time value (from an environment variable). Code is usually written (and compiled) to function across many environments, whereas the runtime environment is specific to a particular invocation. Clearly the runtime environment, with its better specificity, should be trusted more than the code.

Maybe we should just not provide any way to configure these values in code? Or clarify with function names etc that the values provided in code are just fallbacks if the environment doesn't provide the required information?

See also this recent discussion in Cargo on a related problem:

rust-lang/cargo#12515

In this case, there is no value provided in code, but there's a clear hierarchy of specificity.

[TommyCpp]

I think most of people favors the environment variables. I think we can favor env vars here and make sure it's consistent.

Please leave comments if you disgree @open-telemetry/rust-approvers

[hdost]

I think most of people favors the environment variables. I think we can favor env vars here and make sure it's consistent.

Please leave comments if you disgree @open-telemetry/rust-approvers

So while this is not a deviation (more of a clarification), it might also be useful to document. #1297