You want to put a web application on a Linux-ish web server. Let’s say your application is written in Python. It should be easy! Most Linux distributions come with a Python package.

These days, most experienced Python developers will tell you to never deploy your application into System Python -- that’s the version of Python that’s packaged by your Linux distribution. This is sound advice: System Python often lags the state of the art, and lacks features that you may want to use. Its upgrade cycle may vary from your maintenance cycle.

But it doesn’t stop the fact that System Python exists.

If you choose not to use System Python, you need to figure out how to configure an environment with a version of Python, with support libraries, and packages, and how to keep that version of Python isolated from System Python.

So how do you isolate your environment from the System version?

Regardless of your choice of language, the history of deployment environments is the story of a constant trade-off between isolation and available resources. Resources tend to be limited. As resources -- be it money, processing power, physical space, or storage -- have become less limiting, new approaches to isolation have become possible.

In this talk, we’re going to (very quickly) trace the history of environments, from shared hosting through to massively orchestrated container clusters, and show what problems each type of environment introduced, and what problems each type of environment solved.

After this talk, you’ll understand what factors various environments solve, and why some environments that used to be popular have fallen out of favour.