Release Cycles

frontend swift

I work on the internet, where I mostly build and maintain websites. The web has this really interesting property that people often don't even think about: on the web, when I make a change, people can see it instantly.

Sculpture of a tortise and a hare
The tortise and the hare. (c) flickr.com/cle0patra

But recently, I worked on an iPhone app where this isn't the case. And it got me thinking about release cycles and their impact.

A change on the web

On the web, changes are immediate. I make a change, and it shows up.

If there's a bug, I fix it, and there's no more bug. If there's an awesome feature that needs to be added, I add it, and people can get it right away.

A change on the iPhone

For iPhone apps, changes are not so immediate. I make a change; then I have to submit the app to the app store for re-review. That takes a while. And after that, ultimately, the choice to upgrade an app is in the hands of the user. iOS7 introduced auto-update functionality, but it isn't required. A user could have a version of an application that is 3 or 4 versions old. Or even the first version ever released.

So if there's a bug, I fix it, and it takes between 7 days and never to get to the user. Same thing for features.

But on the plus side, a user doesn't have to redownload the application for every use. Once a version is downloaded, it's downloaded, it's there. Things load immediately. No waiting for the network. No dreading a 2G connection. The application is already downloaded and ready to go.

The aggregate

So the web is instantaneous, and an iPhone app has a longer release cycle. Now this is where it gets interesting. As I've said before, an app isn't just what's on the phone, it's so much bigger than that. Most apps have both an iPhone component and a backend API component. Guess where that API component is? That's right, it's on the web.

So we have an app where we can update one component as frequently as we want but the other component must move through a long release cycle.

Structuring the system

How then do we structure the system? Where do we put the logic? If we're aiming for maximum flexibility, we can try to push as much of the logic to the API, the component with the fastest release cycle. That lets us adapt quickly.

But logic that's on the iPhone happens almost instanteously. Logic that's on the API side requires a slower network request. So for a fast, smooth experience that people expect from an app, we need to keep some of the logic on the phone. Logic on the iPhone also shields us from network connection issues.

Ultimately, I settled on the middle ground. I didn't go full on data-driven and use the API to provide everything, because I was scared of connectivity issues. But for the logic that made sense to be in the API, I left it in the API, which gives me flexibility.

And ultimately, if there's a bug that needs to be fixed, the Apple approval process is only about a week. At least I don't have to mail out CDs like AOL did.