On Fri, May 31, 2013 at 2:43 PM, Marcel Kinard <[email protected]> wrote: > Starting off, specifically, I'm asking if we can keep Android 2.2 in Cordova > head. For how long? Until the OS usage in these markets drops into the > "doesn't matter" threshold. I suspect that will not be just a few months. And > does the definition of "keep" mean "actively support" or "just avoid breaking > it"? I'm open to suggestions. If I'm the only person asking for this, I > understand I need to have some skin in the game. >
No. In fact, I'll say hell no. We base our deprecation of Android platforms on the good old Android Pie Chart found at https://developer.android.com/about/dashboards/index.html. The pie chart shows which people actually download applications on the Play Store. I don't care about Android 2.2 devices that don't connect to the store because they don't connect to the store and Cordova isn't distributed to these people. These people don't matter because they don't use apps, whether it be Cordova or a native Android application. Supporting users who will never use apps is insane! Now, the Chinese market was problematic until recently, because Play was blocked until a month or so ago. That being said, I think the Android Pie Chart is a very solid way to tell whether the version matters or not because these are the people who download apps. In fact, if I was an application developer, I'd want to know about the people who actually buy apps and in-app items, and what they run, and I wouldn't support any of the freeloaders. That's where the group of android developers who tweet about minApiLevel=14 come from. If we don't use the Android Pie Chart to determine what to support, what do we use? Stories from the guy who hasn't upgraded their phone in years? The fact is that the store is the only real way that we can have any metrics on people who actually use apps, including people who use Cordova apps. Finally, one of the big problems with supporting old versions for so long is maintaining old devices. Devices eventually break. When you install and uninstall something on a phone enough times, things get weird, and even when you factory reset the device, things tend to not work the same after three years of testing. We have one Android 2.1 device and one Android 2.2 device. They tend to not work on the device wall for some weird reason, and it's time consuming to run mobile-spec on them such that it's not a worthwhile use of time to actually make sure that we don't break Android 2.1 and 2.2 in the real world. When is the last time anyone else who works on Android tested on Froyo? Does anyone remember the last time they tested Eclair when we claimed to support that? The emulator doesn't count! So, no, I see zero value in extending our deprecation window larger than it currently is. We should support users who actually use apps, not people who don't.
