I'm probably missing the point (not sarcasm). I'm about to start writing my first mobile app - a UI client to a ReSTful image rating API. I've been debating between native and HTML5 for the past few days, and reading this article makes me lean ever more toward native. I see that the author's goal is possible. However, he's also illustrated that getting an HTML5 app to feel native on iOS requires a deep understanding…
The author here. You bring up a very good point. The caveats will be different for all different devices. And this does depend on a very deep knowledge of HTML5 hacks and tricks as well as a lot of experimentation. Have you considered titanium appcellerator for your project?
Not to mention lock-in. I'm already at the mercy of the respective distribution channels - I don't need to add to that pain.
Edit:
Don't get me wrong, I'm happy that these things exist. I like anything that lowers the barrier of entry for people who otherwise wouldn't write software. It's just that you're trading a lot of maintenance/optimization time for a shallower learning curve - and it's important to make that known.
Also one last point - the reason why this kind of stuff typically needs to be learned through painful iterative experience is because it's typically undocumented. In most cases it's undocumented specifically because it's not a supported development path, and as a result it's very likely the techniques you discover today will break tomorrow.
It's a lot like leveraging unspecified/implementation-specific language behaviors in C.