Why you really need to upgrade pip
New software releases can bring bug fixes, new features, and faster performance.
For example, NumPy 1.20 added type annotations, and improved performance by using SIMD when possible.
If you’re installing NumPy, you might want to install the newest version.
Unfortunately, if you’re using an old version of pip, installing the latest version of a Python package might fail—or install in a slower, more complex way.
Why?
The combination of glibc versioning, the CentOS end-of-life schedule, and how pip installs packages.
Let’s see what the problem is exactly, how to solve it, and finally—if you’re sufficiently interested—what causes it.
The problem with old pip
Let’s start out with an Ubuntu 18.04 Docker image.
Released in April 2018, this version of Ubuntu has Python version 3.6,