Should the glibc version be locked in for the lifetime of a Python's release?

[brettcannon]

If it's locked in then wheels will have a floor they could reasonably target for a specific Python version. But I'm assuming there could be some benefit to changing it as time goes on.

[emmatyping]

I think it should be locked because if we raise it we are effectively dropping support for certain Linux distros that have lower libc versions. That would put users in an awkward position and make it harder to rely on these releases.

[geofft]

Yes, in the sense that you want to be able to install a security patch successfully. If this project does not guarantee that security patches are meaningfully installable then I think that massively reduces its practical value.

(Also, I can't see much of a reason why it would benefit to increase it. In practice, increasing the floor is useful to pick up new features like new syscalls, but those would also need bindings in os or whatever, or because it's the easiest way to pick up a new compiler, but you aren't going to need that for point releases.)