John Fremlin's blog: HTML5 is a terrible trap

Waiting for updates: connected

What's hot for aspiring dotcom billionaires? Building apps for phones. And justifiably — there's a huge opportunity to quickly ship something to improve people's lives. But there's a diversity of mobile platforms. There are a few companies trying to make cross-platform development frameworks but generally the decision is between focusing on Apple iOS or Android or making a browser based HTML5 app — probably packaged as an app with a webview browser wedged inside it.

The HTML idea is initially attractive and seems to make sense from a business standpoint, because it promises to reduce the effort needed to cover a large market. But this is an illusion. Why?

True, cobbling together something in HTML5 may be possible. But the user experience will inevitably be bad:

— offline access is a really important part of the mobile user experience. Devices go offline all the time: in buildings, in transit, in reception blackspots: in real life an always on connexion is not really available yet. The reason people have an app not a website is that they want to access it on the move and that means that ideally it should provide a decent user experience offline, as much as realistically possible.

— integration with other apps on the platform, user interface conventions, and access to device specific APIs like camera, accelerometer, notifications etc. is much more painful and even impossible.

— performance and responsiveness will be worse. From network access to user interface widgets, native implementations have access to alternatives that are simply unavailable in HTML5.

Now these things are well-understood and very obvious a priori. But they're not the real killer: the real killer is that a webview HTML5 solution is incredibly expensive to polish and make smooth. How do you control the loading experience when to control the loading experience you inevitably need to load a bunch of JavaScript? And then the bugs, the bugs: from SSL not working well on old Android 2 webviews, to crashing on some API calls, to handling loss of connectivity to server-side error responses, to handling running out of memory, it's vastly difficult to go from a demo to a polished production ready system. Essentially, it's a technical trap: doing easy things is easy but getting the details right is impossible sometimes and very tricky and tiresome generally.

In the end, HTML5 doesn't save development time: it increases it and the extra time is spent fiddling with bugs in broken browser implementations. In the end, it's quicker to ship one app per platform, a better user experience, better for morale, and better for finding market fit.