In just a few short months, Python 2 will officially reach the end of its supported life. đź’€ This means that anyone building applications in Python will need to have moved to Python 3 if they want to keep getting updates including, importantly, fixes for any security vulnerabilities in the core of Python or in the standard library. How did we get here?
Python 3 was initially released on December 3, 2008 and included a variety of major compatibility-breaking changes. Overall, these changes are welcomed by Pythonistas and remove a lot of hacks and workarounds that had evolved over time. One of my favorites is that things like dict.items() no longer return a list, so you don’t have to use dict.iteritems() to get a lower memory and more performant way to iterate over dictionary items.
Others, while still welcome, are more challenging from a compatibility perspective as they bring along syntax changes to the core language. This meant that many of the Python libraries that we use for building applications weren’t ready for Python 3. Django, Flask, urllib3, etc... none were ready for the initial release of Python 3. But they now are and have been for quite a while. The efforts to support multiple Python versions have been great but can’t continue forever.
This isn’t the first time this kind of event has happened in the Python world, though. Way back in October 2000, Python 2 came out. This major release of Python had a number of incompatible changes that impacted developers, especially surrounding how one worked with strings and Unicode.
At that time I was working for Red Hat and maintaining Anaconda, the installer for Red Hat Linux. We had decided that migrating all of the Python usage within Red Hat to Python 2 was a priority. There were many less Python modules back then and a small group of us (employed by Red Hat!) were able to do the work to update the modules we shipped to support Python 2. We sent patches upstream, in some cases taking over upstream maintenance of the module, and were able to help move the world forward to Python 2.
But today is different. There are now over 200,000 Python libraries. It’s not practical for one company to help drive all of the changes in the ecosystem to support this new and incompatible release. And the vast majority of the Python packages out there are maintained by volunteers—people who are doing this in their spare time and as a labor of love.
This challenge of how to migrate successfully from Python 2 to Python 3 is exactly the sort of situation where having an incentive for maintainers to support a new version and work with the incompatibilities would be so much better. It’s a perfect example of why we need to pay the maintainers of the open source libraries that all of our applications depend upon. With strong financial incentives in place, the speed and comprehensiveness of our preparation for Python 3 could have been accelerated.
For users, major incompatible changes like those involved in the migration to Python 3 are an important part of keeping software vibrant, alive, and performant. But without being a psychic, it is simply impossible to understand how the world will change and evolve and require modifications to our software.
One of the things I’ve been excited to work on here at Tidelift is our new version guidance functionality. It can help application development teams choose well-supported versions of software that will work together and with Python 3—so when future scenarios like this emerge, we can be more prepared well in advance.