What type of Android application is the most difficult to support (fragmentation)

Question

The other day I saw an interesting and curious scary blog post. It was a mobile collection of Android device developers on which they tested. It was about 400. → http://feedproxy.google.com/~r/Techcrunch/~3/0LybOzd0l0U/

I’m a solo guy, I couldn’t support the application if even a part of such testing and support were required. I know that you can emulate many devices, but there will still be a lot of time on testing on more than a few AVDs. Sounds like it could be a nightmare.

For those who have been chewing Android for a while, any data or tips on which applications process different devices is easiest? The blog developer made a lot of games, are these the most difficult?

I'm sure Hello World works very well on all Android devices, but there aren’t many players, you know?

It would be great to know before starting with an ambitious application, for example, GPS is easy, consistent, but the native code can be a nightmare, or the photos are fine, the video is disgusting to support. SMS, database, SDK access? OpenGL, gestures, etc. Such things...

If someone has general advice or especially the simplest and most complicated list, which can be VERY useful for us, beginners.

thanks

PS and please do not say "develop on iOS ...", this is not a question, and worse, it is too predictable .; -)

Answer 1

They are developing at the level of 70 million downloads , and they are quite significant for the Asian market, which means that the number of new devices from lower-end Asian manufacturers

Netflix tests on just a dozen or so devices , those that represent the majority of their users, some user ROMs, various computing powers, and various playback architectures.

Since you do not need to test the device on every Android device that has ever been made, just make sure your application looks good and solid .

Answer 2

The general advice already given, creating something solid and good looking, is a good starting point. In general, the more autonomous your application is, the easier it is to be sure that it will work on a large number of devices after testing on a basic subset of the best sellers.

The main problem you are facing is that you interact with other applications, such as Gallery, Address Book, etc. Any place that can change or replace the OEM application you want to integrate can cause problems. This is not a problem for most applications. In particular, I worked on a messaging application that required testing on all phones of a particular operator; which is not the kind of project I would expect from someone to solo anyway.

So the simple answer is to avoid integration with other applications, looking more like an iOS sandbox. Unfortunately, this removes some of the Android values, but this is the easiest way to avoid fragmentation issues.

Actually, although you need to test on some real devices, if it is important for you to have commercial success with your application. Personally, I would say that the minimum for testing will be on the phone from each major OEM-manufacturer, both Gingerbread and ICS, and if it is a network application on each medium. This level of financial commitment is unrealistic for someone just starting out, and at that moment you are simply using your users as testers.

Answer 3

Well, since you like scary articles, here are 2 more scary (and informative) articles:
http://opensignalmaps.com/reports/fragmentation.php
http://burnsmod.com/development/2012/05/01/Android-Fragmentation-Hurts-Everyone/

As a rule, everything that uses access to low-level equipment is a nightmare, so when possible, stick with the Java API to access the functionality.

Camera APIs, whether high or low, can be problematic. For example, I had problems with specific devices with high-level camera APIs when called through startActivityForResult using android.media.action.IMAGE_CAPTURE .

Additional information about device errors here:
Is there a compilation of device errors for Android devices?
Android device features

It’s best to build “regular” devices, and then based on the frequency of user complaints about researching specific devices.