Python 3.7

Posted Aug 15, 2019 8:40 UTC (Thu) by amacater (subscriber, #790)
In reply to: Python 3.7 by yodermk
Parent article: EPEL 8.0 released

Python 3 is in RHEL 7.7, as just released. When CentOS get to package that once 8.0 is finished, Python 3 will be available just as it is upstream.

This _finally_ because 2.* dies next year and this is the last feature release for RHEL 7.* as it enters maintenance phase. [Maintenance phase 1 until 2020, 2 until 2024]
This point release was slightly earlier than expected

Just as EPEL 8 is significantly different from that for 7, so the release of SCL 8 for RHEL / EPEL 7 is the last in that form, I think.

Python 3.7

Posted Aug 15, 2019 9:25 UTC (Thu) by yodermk (subscriber, #3803) [Link] (1 responses)

RHEL 7.7 has Python 3.6, not 3.7. 3.6 has been available via IUS and maybe other repos for a long time. I'm specifically wondering how it can be that no reputable repo has pushed 3.7 RPMs. Is it that much harder to package than 3.6?

Python 3.7

Posted Aug 15, 2019 10:53 UTC (Thu) by dsommers (subscriber, #55274) [Link]

It's not that 3.7 necessarily is hard to package. But the whole "instrumentation" around it when Red Hat provides packages needs to be considered.

When Red Hat decides to package something and distribute it, they put a lot of resources into it. They will have allocated dedicated package maintainers for it, security teams are tasked to keep an eye on what happens with these packages, QA teams needs to allocate resources for testing and ensure regression tests are present and running. Then the release teams which needs to ensure packaging is done accordingly to the RHEL standards, that packages are installable, can be upgraded, downgraded and uninstalled without issues, etc. And on top of that, the support teams needs to get knowledge about this package. And all this is an effort which they dedicate for multiple years.

So with all this, it is a noticable cost for Red Hat to commit themselves to ship yet another Python release. I'm not saying it won't happen, but I would more expect it to happen at a later point where Python 2.x truly is dead and there are even more interesting Python releases which has stabilized.

Of course you can argue that the gap between 3.6 and 3.7 is so small lots of the current 3.6 efforts can be reused on 3.7. But then it comes into a prioritization challenge. If there are urgent issues needed to be solved in both 3.6 and 3.7, which should be resolved first? And will Red hat customers be happy if 3.7 gets priority when their own software stack depends on 3.6?

We can't solve all this in a discussion, and at least not here in the LWN comment section. But it is important to understand that when Red Hat ships a package, they're really highly committed to ensure it is maintainable over a longer time and strive to ensure it doesn't explode in the face of their users.