Best practices with ppc64le/aarch64 migration

For a while now we have been using Azure with QEMU to virtualize architectures we didn’t have easy access to. In a few cases we have used Travis CI to do ppc64le builds on native infrastructure. Recently Drone came online and that allows us to do native aarch64.

It was my understanding that we should try to use Azure wherever possible for these architectures as they have generously provided us resources and only use native builds when there is an issue with emulation.

Based on a recent discussion with @isuruf, it seems that my knowledge here may be out-of-date. Given the number of feedstocks involved and the number of people that will be involved, I think it is very important that we establish best practices for how we want to handle builds for these architectures and communicate that clearly to people (IOW document). It’s totally possible that the best practices have been established by those close to the migration, but the communication thereof does not appear to be happening. Perhaps even these are documented and we are just lacking visibility.

Anyways am opening this issue to make sure we have firmed up what those best practices are, document them, and make sure that info is visible.

Apologies if I’m just missing lots of things here. Figured that if I was missing things, there are many maintainers who are also missing them, which seems worth correcting if nothing else.

cc @conda-forge/core