While numpy provides pre-compiled wheels for Alpine on both x86-64 and aarch64, scipy only provides alpine wheels for x86-64. This means people with newer ARM-based Macs (myself included) would have a hard time working on the project, either having to build and run the Docker image using emulation, or we'd have to modify the build process to compile scipy (which is both difficult and very slow).
Switching to Debian means we can use the scipy "manylinux" flavour of wheel, which is available for aarch64 too.
For https://github.com/ietf-tools/wagtail_website/issues/173, we're integrating the wagtail-ab-testing package, which unfortunately depends on numpy and scipy.
While numpy provides pre-compiled wheels for Alpine on both x86-64 and aarch64, scipy only provides alpine wheels for x86-64. This means people with newer ARM-based Macs (myself included) would have a hard time working on the project, either having to build and run the Docker image using emulation, or we'd have to modify the build process to compile scipy (which is both difficult and very slow).
Switching to Debian means we can use the scipy "manylinux" flavour of wheel, which is available for aarch64 too.