Starting from 4.0, many APIs were cleaned up, many bugs were fixed, and it seems that many carriers/manufacturers have stopped shipping bizarrely broken custom modifications.
Unless you have a large development team and absolutely must reach every possible customer, targeting 4.0 and upwards seems to be the most efficient choice.
Less importantly, they only take into account devices that have the Play Store (which you might consider your audience, but there are tons of people in the world happy to buy things from alternative markets available on their devices).
However, and this is somewhat insidious: they measure devices used in the last seven days. How often do normal users "actively visit the Play Store"? I find it highly unlikely that every normal user does so at least once every week.
In fact, I would be much less surprised by a universe where people only "visit the Play Store" once a month at best, and only when there's something specific for them to be doing (maybe they hear about a new product, and go to download).
I would then further submit that this frequency decreases the older the device is: during the first month of owning your device, you probably have tons of things you suddenly realize you want, because the device comes with no apps installed.
With an expectation curve on the usage of the device thereby working against older devices, you will see a chart that only looks at the most recent 7 days biased towards recent versions as the probability of older versions dips.
Let's look at what happens then if we posit that the average 2.3 user is using the Play Store once every four weeks, while the average 4.x user is using the Play Store once every two weeks: this mild skew makes 30/70 become almost 50/50.
Are these people who are using 2.3 really outside of your target market because they only download an app every month instead of every couple weeks? If you market to them, and make it clear your app supports them, will they not buy it?
While that might be true, it isn't something you can support with just these numbers from Google. It's even worse, in fact, when you remember that when Google was tracking "active devices" instead of "devices actively visiting Play" the 2.3/4.x split was so far in the other direction as to be unbelievable, and when Google decided "fine, we are limiting it to active users" the numbers jumped.
And now, we see the numbers creeping better and better... but did you notice that they changed the time range from 2 weeks to 7 days in the few months since Apple lambasted their uptake at WWDC? If not, you aren't paying enough attention ;P, as that's a really critical detail: Google keeps moving the goal post, so you can't take any solace in "we are making progress".
On a side note, I'd love to update the software on it, but Samsung/Telus have not offered an upgrade path. So, 2.3 it is.
How many devices are out there that could be running 4.0+ but aren't due to manufacturer or carrier interference? Not everyone is capable of rooting and installing custom versions of the OS.
Android fragmentation is purely a manufacturer/carrier issue. To the article's point, maybe developers can help that process by enforcing 4.0+.
Read the blog post then decide.
Wrote perspective from different Android devs in the blog post.
I'm resizing to a bigger plan.
Cool story bro.
Not to mention there are still plenty on new GB phones sold in developing markets.
He's advocating dropping potentially 30% of his potential consumers because it's technically easier. Some of us can probably afford that, some can't. But I just think broad statements like "Just drop support for the older version, it'll push people to upgrade and make the entire system better" is a stupid argument.
The reality is people paid a lot of money for their sweet new Android 2.3 phones 2 years ago or less. Why should you punish them now and not spend the extra time to make your app work well there?
The biggest problem with GB that I've seen isn't necessarily the lack of certain APIs , but the constraints in terms of resources on phones with GB, which tend to have worse hardware (because they are older devices or lower-end devices). If you think by developing on ICS/JB you magically can stop worrying about performance or resource constraints, you'd be mistaken. Some of the new cheaper Android phones in developing markets are very resource constrained.
Cheap devices aren't going away. Dropping GB support does not mean you can stop writing performance-optimized code or profiling your app.
People with older phones that they paid only a few hundred bucks for two years ago should not have the expectation for it to keep working. It is not punishment, they can continue to use their devices as is perfectly fine."
Site is back up now.
They're still out there, absolutely.
Also, writing code to work well on lower end phones might be a pain, but it means it's going to make your app even smoother on the higher end devices. It requires more time, buy you get more of a performance buffer.
Well, ...not _nothing_
http://forum.xda-developers.com/showthread.php?p=26920609
...but I understand that only techies dare to go download random images from the interwebz and attempt flashing their device. Personally, my Kindle Fire is running this SimpleCM9 Android 4.0.4 quite well.
Thanks for that link!
I've been making this argument for a while, whenever someone cries about fragmentation, mainly from my personal experience. I'm a "techy"; I want the latest and greatest. But in my family there are 4 Android users, 3 of which don't know what version they're running, or care in the least. They use the phone to make calls, check email, text, browse the internet and maybe download an app or two. Every version of Android has those capabilities.
Hipster Mountain View/Palo Alto is not the world. Most people who manage to afford to buy a smartphone, use the thing until it dies. I personally just upgraded from an Android 2.3 device, and my latest Apple device is on iOS 2. And I read Hacker News for christ's sake.
Maybe for your app, your community the norms are different. But be careful not to bias your data - you better be sure that you are not unknowingly turning away people who would be customers if only you supported their device.