Of course you can do limited roll-outs of new features.
You just put it in a feature flag. The app can check w/ Twitter if that user has the flag enabled.
More on Feature Flags: http://code.flickr.com/blog/2009/12/02/flipping-out/
-David
It's harder to do limited roll outs, but not impossible. It's feasible to have all the code in place and only trigger it based on a server response, for instance. That's basically how they handle the web app.
How does that work with the App Store approval process? I could see that also being a strategy to try and sneak stuff past Apple.
The same way that testing apps with Easter eggs works. You have to tell Apple that the feature exists, and let them test it. If you don't, and they find out after approval, you run the risk of getting your app pulled.
Would Apple really pull the official Twitter app? It seems highly unlikely, because Twitter has leverage over Apple's users. People won't switch to some other short message sharing service, they'll switch to a non-Apple phone.
Re:Twitter Kills The #Dickbar
by statictype on Thursday March 31, 2011 @09:59PM
(#2394611)
The iPhone was popular before the existing Twitter client existed. There are also lots of good 3rd party Twitter clients for the iPhone.
People won't switch to some other short message sharing service, they'll switch to a non-Apple phone.
Nah, they'll probably use the web UI or some other native client that supports Twitter.
You can install iOS dev apps on way more handsets than that. The installers simply need your certificate/profile.
As for a limited rollout, they could've rolled a new binary and only turned on the feature for, say, 100 users selected at random.
> As for a limited rollout, they could've rolled a new binary and only turned on the feature for, say, 100 users selected at random.
Be careful with this - developers have previously gotten into trouble for having functionality for users that differs from that for reviewers. You'd have to also add a way for the reviewers to review the "limited rollout" functionality.
If Twitter could have deployed the bar to a hundred or so real users, I think they would have quickly found that people mostly hated it.
I believe you're allowed to distribute test versions of your app to 6-ish handsets for testing, but I'm sure this barely covers the internal folks at Twitter. It'd be nice if they'd had the option to do some more thorough early testing. (Whether they would have done so, or would have listened to the results is outside my point).