Localisation
A discussion I was reading elsewhere linked me to this old gem on “falsehoods programmers believe about names.” I laughed, but it was one of sympathy rather than surprise. I’ve tackled localisation on a bunch of projects in my time, and the thing that takes the time is not content wrangling, or getting the right unicode fonts in. It’s dealing with the assumptions that the code team have already made and implemented in the early days of development.
It’s a print-to-string line that formats currency for display as $1.23, regardless of the user’s locale. Or a user signup form that has one box for First Name and one box for Surname, and expects exactly one word in each. Simple things, taken from the programmer’s own experience as ‘obviously’ the way it is, thrown in because they have to get a working implementation done quickly, and no-one has asked them to take localisation into account. That can all get sorted later, right? No. Not when you build in assumptions at the very base level that simply aren’t true.
For example, Steam. I was probably one of the early adopters of it. I didn’t want to be, annoying system tray icon that it was, but I wasn’t going to wait for Half-Life 2. To sign up, I had to use my email address as a username. Sure. Whatever. It’s now over 12 years later. That email address, tied to an ISP I moved away from, is long since gone. Can I change my username? No. Because they insisted on a particular form of unique ID at the time, and they insisted that usernames can never change. New users don’t have the same restriction, they can pick whatever username they want, but mine is frozen in time. Even though there is an actual email field on my account which is wholly separate from the username, I still have to enter a 30 character email address that has no relation to reality every time I want to log in. While I can just treat it as not an email address but just an oddly formatted unique identifier string, it jars with me, every single time.
Arguably this is just the meat of development – changing requirements over time invalidating initial assumptions. But for me it’s a plea to other developers – to slow down and take some time thinking about your initial implementation. If it’s not on a specification handed to you and you’re winging it based on how you think it should be; maybe think to yourself what the ramifications of how you choose to implement it will be. What won’t you be able to do if you implement it this way? What are the awkward cases, the potential ranges of input. Will it be possible to fix it later once the system is live and populated with data, or are you building in something that’s a fundamental to the system?