Dropbox Started as a Python Script on a Bus Ride - https://www.youtube.com/watch?v=69HHFdoz34M

Well, I started working on Dropbox because I kept forgetting my thumb drive. And eventually I just got sick of having this problem and I wanted to solve it permanently, first for myself, but then it was clear that many other people would be were dealing with a lot of the same issues. So it was on this one bus ride where I forgot my thumb drive at home. This was 2006. I had my laptop with me, but I realized that I didn't have my thumb drive and so I was super frustrated. And this was before the iPhone. You didn't have WI FI on the bus. So I had something like four hours of time to kill. And so I opened up the editor and started writing some Python code. Not really envisioning starting a company or anything, but really just to mechanically solve this problem of never carrying around my thumb drive again. And the first few lines of code were basically around, all right, I need to watch a folder or a directory for changes, and anytime a file changes, I need to identify the change and sync the difference up to a server. If you think about Dropbox, it's a cross platform piece of software by design and by necessity. Dropbox helps you sync your files across Windows and Mac and Linux. So if you think about it, the conventional way to do that would have been to build a Windows version in C and use the Win32 API and use a different stack for Mac and a different stack for Linux. I didn't have time for that. So I was hoping and kind of praying that I could be able to write one version of the app that would run anywhere. This was mainly out of necessity because I was just one person. And then I partnered up with my co founder shortly thereafter. But then there were two people. So my co founder is basically writing the server of Dropbox and I was writing the client software. And the conventional way to do it would have been to have to be like in very low level systems code on all these different platforms. But with Python, my hope, and with the little bit of experience I had, I'd used Python on every major operating system and it broadly worked the same way. And in general I could write a script on Windows and have it run on Linux and vice versa. But Dropbox was also very tightly integrated in the operating system, so that posed a bit of a challenge because we still needed the capability and the performance and these really tight integrations with the operating system. But. But I just started writing in Python and hoping that none of the roadblocks would completely blow me up. So the BASIC architecture was like I had the main version of the app, which was OS OS agnostic. It had the basic syncing logic. And then there were operating system specific backends for each of the major operating systems. But I was always concerned, like, hey, are we going to have performance problems? Because it's an interpreted language, There was no precedent. So when you've used, if you download Firefox or something back then it would be a broadly compiled or the core of the app would be written in the operating system native language or something that's C or C. And then there are other challenges, like Dropbox would have to be multi threaded because there's a lot going on simultaneously. One thread would be talking to the network, another might be reading the hard disk, another might be doing a lot of sync engine logic. And fortunately, Python had solutions for each of those problems. I didn't know it at the time, I had to kind of walk into the jungle and hope that I could get out the other side. But there were a lot of benefits from, in addition to just the expressivity of the language itself, lending itself to clean design, then there was also a really mature standard library. And so a lot of other programmers had figured out all the contours of the different operating systems and had kind of batteries included on every platform. And then from some of the trickier things around multi threading, even things like memory usage or optimizing memory usage, and then these deep OS integrations or things that were pretty CPU intensive, like these hashing algorithms or compression or encryption, either there was a library available for, that was usable, that already had a binding for Python, or I could take something that was in C and wrap it with a Python binding. And so fortunately we just started marching and then there was a lot of navigating around little different potholes, but it ended up being probably the best technical decision we ever made to use Python.