Interview- Guido van Rossum- The Early Years of Python

I worked at this place, CWI, and for my then current project, which was the Amoeba operating system, which I was working under Sape Mullender. One of our projects was to write application-level utilities for this new Amoeba operating system. And so, there was an operating system kernel and it was very good at talking to the network. And it was very good at managing processes. But there wasn’t much user-level software. There was barely a shell, I think there was a ported, very old Unix shell something like the original Unix version six shell, but because the file system model on Amoeba was so very different we couldn’t take an existing suite of Unix utilities. And so we wanted it to be self-hosting. And in order to that we realized we need a large amount of user-level sort of applications, utilities, tools, whatever you call it, from an editor to a mail program to a login utility and a backup tool. And one of the things I realized, as we had a small team of people sort of working on those things, that it was very slow going writing all that stuff in C, and it wasn’t particularly interesting or sort of novel or difficult and often we also didn’t really care how fast the code ran. All the reasons why you would write a piece of software in C didn’t really apply. The only reason left was that C was the only language that for which we had a compiler. There was this ABC language that I had in the back of my mind that would be a much better language to write a whole bunch of utility application tools for Amoeba, except that ABC was so high and abstract that it really. It wasn’t a good language to talk to servers and file systems and processes and it’s sort of, the whole operating system thing was abstracted away from ABC. It was almost something like it, I don’t know. In an alternate universe ABC could have become the language of spreadsheets. It was very good for talking about a user’s data. And doing all sorts of clever stuff with that data. And it did that using very, sort of general all-purpose data structures like lists and dictionaries and of course it had very nice code structuring devices like a small number of simple statements that could be combined in any way you wanted to create other constructs. The usual function and procedure abstractions. It was not at all an object oriented language although the implementation had some objects sort of shining through. Anyway ABC as it existed still wasn’t usable but I somehow, since I had worked on the ABC project, I knew exactly how it was built. And I had this idea in my head that, given how much time we had available for our project, I could actually build a whole new language, design and implement it from scratch, given what I learned about both the design and the implementation of ABC, and start using it to implement our suite of tools and still be ahead of the game compared to a situation where we would just have sort of clunked on writing the things we wanted to write in C. For three months, I did my day job and at night and whenever I got a chance I kept working on Python. I believe that within three months I was to the point where I could tell people, look here. This is what I built. And it had an interactive interpreter loop. So the first demos were all let’s assign an expression to a variable and print it back. And let’s define a small function and call it. And let’s put some things in an array and iterate over the array. And all those things worked. And somehow, and I don’t exactly know sort of how fast this happened, but certainly my two office mates were almost instantly taken with it and started helping out. And they and a few others within the institute were very excited about this thing that I’d built. And started sort of, of course we didn’t instantly use it on Amoeba because it wasn’t mature enough to actually write the Amoeba utilities that we wanted. But it was already instantly useful enough to run on our Unix system. People within CWI, even outside my own department, started using it. And recognizing that it was fun and productive to use. And they used it for small scripts and people started contributing things like bug fixes. Somehow things then went very quickly during that first year. Because I think by the end of 1990, so a year after I started, we developed a plan to do an open source release of this language. And this was before the word open source had even been coined. So we didn’t call it that. But we did have some models. We had like X11, the window system at the time, had a distribution that was one of the sort of open source examples. With two colleagues I worked sort of building a distribution and we, actually it turned out to be very simple to get management to sign off on this release. That was an an incredibly lucky stroke. I just sort of, I asked my manager’s manager what should we do about this and he said oh, you gotta talk to this and this person in the administrative branch of the institute. And I talked to this woman and she was like in charge of all legal affairs I believe and I said well I have written this source code and I would like to release that and I have sort of made up a license that’s like identical to the MIT license, I could say MIT releases software under exactly this license and we just put like the formal name of our institute in there instead of the regents of the institute. She asked me some questions like did someone pay for this to be developed? And my answer was no, I started it all on my own time and it was for this research project and that sort of. My manager’s fine with it. And so she said sure, go ahead. And we did it and that’s so in February ‘91, we did the first Python release. It felt like, at the time like an incredible milestone. We needed to post it to Usenet. There was this Usenet newsgroup that would receive source code for random free projects. I got immediately started getting useful positive feedback from, well initially just from random people who picked up free software from Usenet. And we quickly sort of got into a routine of doing new Python releases. And every time, I mean there were the obvious improvements to the language and the library and bug fixes, a very important category of things that were often contributed were ports or ported fixes where people had different architectures, different compilers. The Unix world was much less homogenous at the time. There were lots of small Unix vendors that had their own compilers, or their own hardware. All sorts of things. So the big things that happened during say the first half of the ’90s was a community of Python users and developers self-organized. Then came the invitation to come to the United States for a couple of months from NIST, the National Institute for Standards and Technology. So I spent two months there. We organized the first Python workshop, which was hosted by NIST. Through that Python workshop I met people who offered me a job and from ‘95, I mean I went back to the Netherlands for a few months and then from ‘95 to 2000 I worked in US in northern Virginia at CNRI. And so there we worked through a lot of growth of the community and the infrastructure. We created the Python website, I got in touch with a bunch of people who are still active in the Python community like Barry Warsaw. I think when I started there Python 1.3 was about to be released. And then while I was there we released several subsequent versions leading up to 1.5.2 which for some reason 1.5.0 was nothing but 1.5.2 remained the sort of the standard, the gold standard of Python for a very long time afterwards.