I think you have, but I don’t think you’ve realised that I still have no idea what you mean. And from what I can see, there are a lot of other people struggling to understand you, too. I’m genuinely trying to get what you’re saying, but all I keep hearing is circular definitions. You clearly don’t think that they are circular, but whatever insight you have as to why they aren’t is somehow getting lost.
Let me try again.
There are two parts to the spec:
- What does it mean for a wheel to be manylinux_X" compliant?
- What does it mean for a system to be “manylinux_X” compliant?
I think you’re focusing on the idea that a wheel is manylinux_X compliant if it works on all manylinux_X compliant systems. Let’s assume I’m OK with that (I have some reservations, but on the whole I’m willing to go with it). But that still leaves open the question of how I know if a system in manylinux_X compliant. And I don’t know how you are specifying that - there’s no document I can refer to which lets me check if my system is manylinux_X compliant, and no set of checks I can run to do so.
And suppose I take it the other way around and assume that a system is manylinux_X compatible if it runs manylinux_X compatible wheels. I’m much less happy with that as a definition, but once again, let’s go with it. Now, how do I check if my wheel is manylinux_X compatible? If the answer is “run auditwheel on it”, do you then see that this means we have a definition that’s implementation defined (specifically, defined by the implementation of auditwheel)? And that’s not sufficient, for all of the reasons why we have standards in the first place, to avoid implementation-defined behaviour. And if the answer is something other than “run auditwheel on it”, I don’t know what that answer is, so a (possibly repeated) explanation would be helpful.
Whichever way you choose to frame the argument, one of the parts needs to be clearly defined, and you don’t seem to be doing that. Which is why people feel your arguments are circular.
If you’re still struggling to understand my confusion, here’s another suggestion - can someone else who’s arguing for the perennial manylinux proposal try to clarify? Maybe a different perspective will help break the impasse here. Maybe @takluyver can help to clarify? Once we can get past this issue of not understanding each other, we’ll hopefully be able to get the discussion moving forward again.