It would be nice to see some justification for this matrix. I did some googling, and though I was surprised at seeing the 3G recommended for 'No Support', the data would seem to back that up: http://www.14oranges.com/2012/04/ios-version-statistics-apri...
It's easy understand that developers don't really want support older platforms as that might be quite of a burden. But please don't drop support if you can. Not everyone likes updating to newest version of OS on their phone all the time especially if there are no significant changes coming. (IMO There haven't been any killer features in 5.x and 6.x revisions of iOS.)
The thing that annoys me most is Mac applications that don't support 10.6 any more even though that's shoudn't be hard at all as desktop APIs are quite stable and don't change much. Or applications that support full screen only on 10.7. (Seriously, Evernote?)
iOS 5 has a far less buggy memory management system (strong/weak vs strong/unsafe_unretained). You may not realize this is a killer feature, but it is, it makes your apps function in cleaner ways, and allows more functionality to be correctly done done with less work.
Additionally, the UI is tons easier to correctly customize, making apps more colorful with less custom graphics, making apps smaller. This means more apps can be downloaded OTA, and apps can have more stuff in them before hitting the OTA limit.
Supporting iOS4 causes both of the above to be missing. I don't even want to talk about iOS 3, which makes it so you can't use ARC and Blocks, two of the things that make iOS dev much better today than it was 24 months ago.
While there may not have been a lot of user-facing changes in iOS5, there were a LOT of development changes that has reduced the amount of hackery needed. For example, iOS5 included the ability to style many aspects (navbar, buttons,tabbar) that previously had to be hacked around and were unstable. Developers either gave up on the styling and used the new features in iOS5, or would have to maintain the hack + new appearance classes.
This graphic needs more context/metrics. Are these judgments made on past trends? Usage numbers? Something like how MacRumors buying guide averages out past release schedules to determine the next likely update:
We're also not supporting the original iPad on some new apps since the hardware is too underpowered to deliver great performance, and we believe it'll quickly become a liability in the future (better to have not supported it from the beginning than having to drop support a few versions down the line).
