Jul 6, 2020
Haarlem, 1956. No, this isn’t an episode about New York, we’re talking Haarlem, Netherlands. Guido Van Rossum is born then, and goes on to college in Amsterdam where he gets a degree in math and computer science. He went on to work at the Centrum Wiskunde & Informatica, or CWI. Here, he worked on BSD Unix and the ABC Programming language, which had been written by Lambert Meertens, Leo Geurts, and Steven Pemberton from CWI.
He’d worked on ABC for a few years through the 1980s and started to realize some issues. It had initially been a monolithic implementation, which made it hard to implement certain new features, like being able to access file systems and functions within operating systems. But Meertens was an editor of the ALGOL 68 Report and so ABC did have a lot of the ALGOL 68 influences that are prevalent in a number of more modern languages and could compile for a number of operating systems. It was a great way to spend your 20s if you’re Guido.
But after some time building interpreters and operating systems, many programmers think they have some ideas for what they might do if they just… started over. Especially when they hit their 30s. And so as we turned the corner towards the increasingly big hair of the 1990s, Guido started a new hobby project over the holiday break for Christmas 1989.
He had been thinking of a new scripting language, loosely based on ABC. One that Unix and C programmers would be interested in, but maybe not as cumbersome as C had become. So he got to work on an interpreter. One that those open source type hackers might be interested in. ALGOL had been great for math, but we needed so much more flexibility in the 90s, unlike bangs. Bangs just needed Aquanet.
He named his new creation Python because he loved Monty Python’s Flying Circus. They had a great TV show from 1969 to 1974, and a string of movies in the 70s and early 80s. They’ve been popular amongst people in IT since I got into IT.
Python is a funny language. It’s incredibly dynamic. Like bash or a shell, we can fire it up, define a variable and echo that out on the fly. But it can also be procedural, object-oriented, or functional. And it has a standard library but is extensible so you can add libraries to do tons of new things that wouldn’t make sense to be built in (and so bloat and slow down) other apps. For example, need to get started with big array processing for machine learning projects? Install TensorFlow or Numpy. Or according to your machine learning needs you have PyTorch, SciPi, Pandas, and the list goes on.
In 1994, 20 developers met at the US National Standards Bureau in Maryland, at the first workshop and the first Python evangelists were minted. It was obvious pretty quickly that the modular nature and ease of scripting, but with an ability to do incredibly complicated tasks, was something special. What was drawing this community in. Well, let’s start with the philosophy, the Zen of Python as Tim Peters wrote it in 1999:
Those are important enough to be semi-official and can be found by entering “import this” into a python shell. Another reason python became important is that it’s multi-paradigm. When I said it could be kinda’ functional. Sure. Use one big old function for everything if you’re moving from COBOL and just don’t wanna’ rethink the world. Or be overly object-oriented when you move from Java and build 800 functions to echo hello world in 800 ways. Wanna map reduce your lisp code. Bring it. Or add an extension and program in paradigms I’ve never heard of. The number of libraries and other ways to extend python out there is pretty much infinite.
And that extensibility was the opposite of ABC and why Python is special. This isn’t to take anything away from the syntax. It’s meant to be and is an easily readable language. It’s very Dutch, with not a lot of frills like that. It uses white space much as the Dutch use silence. I wish it could stare at me like I was an idiot the way the Dutch often do. But alas, it doesn’t have eyeballs. Wait, I think there’s a library for that.
So what I meant by white space instead of punctuation is that it uses an indent instead of a curly bracket or keyword to delimit blocks of code. Increase the tabbing and you move to a new block. Many programmers do this in other languages just for readability. Python does it for code.
Basic statements included, which match or are similar to most languages, include if, for, while, try, raise, except, class, def, with, break, continue, pass, assert, yield, import and print until python 3 when that became a function. It’s amazing what you can build with just a dozen and a half statements in programming. You can have more, but interpreters get slower and compilers get bigger and all that…
Python also has all the expressions you’d expect in a modern language, especial lambdas. And methods. And duck typing, or suitability for a method is determined by the properties of an object rather than the type. This can be great. Or a total pain. Which is why they’ll eventually be moving to gradual typing.
The types of objects are bool, byte array, bytes, complex, dict, ellipsis (which I overuse), float, frozen set, int, list, NoneType (which I try to never use), NotImplementedType, range, set, str, and tuple so you can pop mixed tapes into a given object. Not to be confused with a thruple, but not to not be confused I guess…
Another draw of python was the cross-compiler concept. An early decision was to make python cable to talk to c. This won over the Unix and growing Linux crowds. And today we have cross-compilers for C and C++, Go, .Net, Java, R, machine code, and of course, Java.
Python 2 came in 2000. We got a garbage collection system and a few other features and 7 point releases over the next 10 years. Python 3 came in 2008 and represented a big change. It was partially backward-compatible but was the first Python release that wasn’t fully backward-compatible. We have had 7 point releases in the past 10 years as well. 3 brought changes to function print, simpler syntax, moved to storing strings in unicode by default, added a range function, changed how global variables react inside for-loops, implemented a simpler set of rules for order comparisons, and much more.
At this point developers were experimenting with deploying microservices. Microservices is an a software development architecture where we build small services, perhaps just a script or a few scripts daisy chained together, that do small tasks. These are then more highly maintainable, more easily testable, often more scalable, can be edited and deployed independently, can be structured around capabilities, and each of the services can be owned by the team that created it with a contract to ensure we don’t screw over other teams as we edit them.
Amazon introduced AWS Lambda in 2014 and it became clear quickly that the new micro services paradigm was accelerating the move of many SaaS-based tools to a micro services architecture. Now, teams could build in node or python or java or ruby or c# or heaven forbid Go. They could quickly stand up a small service and get teams able to consume the back end service in a way that is scalable and doesn’t require standing up a server or even a virtual server, which is how we did things in EC2. The containerization concept is nothing new. We had chroot in 1979 with Unix v7 and Solaris brought us containerization in 2004. But those were more about security. Docker had shown up in 2013 and the idea of spinning up a container to run a script and give it its own library and lib container, that was special. And Amazon made it more so.
Again, libraries and modularization. And the modular nature is key for me. Let’s say you need to do image processing. Pillow makes it easier to work with images of almost any image type you can think of. For example, it can display an image, convert it into different types, automatically generate thumbnails, run sooth, blur, contour, and even increase the detail. Libraries like that take a lot of the friction out of learning to display and manage images.
But Python can also create its own imagery. For example, Matplotlib generates two dimensional graphs and plots points on them. These can look as good as you want them to look and actually allows us to integrate with a ton of other systems.
Van Rossum’s career wasn’t all python though. He would go on to work at NIST then CNRI and Zope before ending up at Google in 2005, where he created Mondrian, a code review system. He would go to Dropbox in 2013 and retire from professional life in 2019. He stepped down as the “Benevolent dictator for life” of the Python project in 2018 and sat on the Python Steering Council for a term but is no longer involved. It’s been one of the most intriguing “Transfers of power” I’ve seen but Python is in great hands to thrive in the future. This is the point when Python 2 was officially discontinued, and Python 3.5.x was thriving.
By thriving, as of mid-202, there are over 200,000 packages in the Python Package Index. Things from web frameworks and web scraping to automation, to graphical user interfaces, documentation, databases, analytics, networking, systems administrations, science, mobile, image management and processing. If you can think of it, there’s probably a package to help you do it. And it’s one of the easier languages.
Here’s the thing. Python grew because of how flexible and easy it is to use. It didn’t have the same amount of baggage as other languages. And that flexibility and modular nature made it great for workloads in a changing and more micro-service oriented world. Or, did it help make the world more micro-service oriented. It was a Christmas hobby project that has now ballooned into one of the most popular languages to write software in the word. You know what I did over my last holiday break? Sleep. I clearly should have watched more Monty Python so the short skits could embolden me to write a language perfect for making the programmers equivalent, smaller, more modular scripts and functions. So as we turn the corner into all the holidays in front of us, consider this while stuck at home, what hobby project can we propel forward and hopefully end up with the same type of impact Guido had. A true revolutionary in his own right.
So thank you to everyone involved in python and everyone that’s contributed to those 200k+ projects. And thank you, listeners, for continuing to tun in to the history of computing podcast. We are so lucky to have you.