Is it a good time to remove all nostd library implementations and start to use standard c++ libraries instead?

[alex-q-chen]

It's quite error prone if API users forget to define macros like OPENTELEMETRY_STL_VERSION=2023 in their client applications, so do we think it's a good time to remove all nostd library implementations and start to use standard c++ libraries instead?

[malkia]

Ideally when users are using it, no additional -DOPENTELEMETRY_ macros should be needed. Somehow the Otel they use should have some generated "config.h" file, which gets included by all other headers, and set things right.

It has to go further and ensure that anything pre-defined before config.h should raise alarms, as it may lead to ODR violations (https://en.wikipedia.org/wiki/One_Definition_Rule)

Engineers using it, would not be right away aware of these specifics, so it must be guarded.

[lalitb]

so do we think it's a good time to remove all nostd library implementations and start to use standard c++ libraries instead?

nostd constructs are not just for the features missing in C++11/14, but also to provide ABI compliant interfaces - https://github.com/open-telemetry/opentelemetry-cpp/blob/v1.17.0/docs/abi-policy.md