The problem with this argument is that we don't have any more extreme strides in mobile performance in the next 5 years. We already have phones that can effectively run any intensity of "eye candy" with spiff - by the time such a policy came to market in 2 years, it would be 2014 - unless we have a revolution in tech that lets us crank up the ghz without dramatically increasing the voltage to match, any arm based device for Android will be able to run any degree of eye candy they can add in in the "short term".
Back in 2008, we could easily see that the 500mhz single core ARM chip was going to be short lived. We had tremendous strides in performance available to us in just upping core clocks, and then we had the mobile multicore revolution.
We are done with both the main vein and the secondary digs of the mobile CPU goldmine (in reference to Herb here) and we are already pretty deep into heterogeneous computing. Remote processing is next, when the telecom monopoly is torn down so we can get the data speeds we are actually capable of today without paying monopoly rates.
The hardware WILL get significantly better between 2014 and 2019 - a 2019 smartphone will almost assuredly be as powerful as the 2014 consoles that will be coming out for gaming - but we can actually predict that. A 2014 console will have approximately a GCN 7850 graphics card and quad core 3.5 ghz cpu of some sort. So expect the mobile chip of 2019 to operate at ~4ghz with some number of cores and a lot of compute nodes approximately equivalent to modern mainstream desktop graphics at similar wattages to what we have now.
If the version of Android in 2019 can't run on an Android device of 2014, then that isn't the hardware not keeping up with software - that is just terrible design. Just like how desktop hardware peaked in 2006, mobile hardware will peak soon as well, after which performance is "good enough" for average joe consumer, and the UX won't change.
Another good example is how Google can keep pumping out new versions of Chrome without hesitation that the last version's user can run the new version. The browser is not getting exponentially more power or resource hungry, and in many ways they optimize its performance as much as they sacrifice it each release for better experience (of course things like webgl are pressuring browsers upwards with new usage in an environment without resource contraints) but the basic device, or in this case application, is not growing more and more computationally intensive by large jumps each release, or they would not be able to push out silent updates like they do.
Same thing with Android. Once it plateaus, there is no excuse not to use the silent continuous always backwards compatble update model, because the device is now "good enough". Just like how a dual core 3ghz pc from 2005 with 4gb of ram is still good enough today for 95% of use cases.
Desktop didn't really peak out at 2006. The clock speeds are not really raising anymore, true, but both architectural and fab improvements make a difference. If you take a look at benchmarks on anandtech for example, you'll find that current CPUs are about five or six times faster in well parallelizable tasks like video encoding or about three times faster in tasks that don't parallelize well.
While this might not seems like much, waiting for something one minute instead of five makes a difference, and 2006 computer won't play FullHD video, and you'll feel the difference on js heavy web pages.
Won't play FullHD video? Maybe if you're talking about crazy everything-x264-can-do encodings, but a decent 2006 computer will have a geforce 7xxx or similar and play "normal" FullHD video like you buy from itunes just fine.
But without the GPU (and I'm not sure how geforce 7xxx driver support looks now) even 720p videos might get problematic, some becoming pretty much a slideshow, some being just slightly uncomfortable when going around 20 fps. I know they do on my 2007 macbook.
The problem is that the general use case doesn't have a media center pc from 2007, "A significant portion" of consumer hardware is EXCUSIVELY for web browsing and word processing. Sources: every single relative I have except for my father who uses trading software to watch stocks, only use a combination of word (if I haven't touched their boxes), libre office (if I have), ie (if I haven't) or firefox (if I have). Those two cases cover 90% of pc use, maybe some photo browsing, but none of them watch video on a general purpose pc. (yet). Netflix probably changes that.
Back in 2008, we could easily see that the 500mhz single core ARM chip was going to be short lived. We had tremendous strides in performance available to us in just upping core clocks, and then we had the mobile multicore revolution.
We are done with both the main vein and the secondary digs of the mobile CPU goldmine (in reference to Herb here) and we are already pretty deep into heterogeneous computing. Remote processing is next, when the telecom monopoly is torn down so we can get the data speeds we are actually capable of today without paying monopoly rates.
The hardware WILL get significantly better between 2014 and 2019 - a 2019 smartphone will almost assuredly be as powerful as the 2014 consoles that will be coming out for gaming - but we can actually predict that. A 2014 console will have approximately a GCN 7850 graphics card and quad core 3.5 ghz cpu of some sort. So expect the mobile chip of 2019 to operate at ~4ghz with some number of cores and a lot of compute nodes approximately equivalent to modern mainstream desktop graphics at similar wattages to what we have now.
If the version of Android in 2019 can't run on an Android device of 2014, then that isn't the hardware not keeping up with software - that is just terrible design. Just like how desktop hardware peaked in 2006, mobile hardware will peak soon as well, after which performance is "good enough" for average joe consumer, and the UX won't change.
Another good example is how Google can keep pumping out new versions of Chrome without hesitation that the last version's user can run the new version. The browser is not getting exponentially more power or resource hungry, and in many ways they optimize its performance as much as they sacrifice it each release for better experience (of course things like webgl are pressuring browsers upwards with new usage in an environment without resource contraints) but the basic device, or in this case application, is not growing more and more computationally intensive by large jumps each release, or they would not be able to push out silent updates like they do.
Same thing with Android. Once it plateaus, there is no excuse not to use the silent continuous always backwards compatble update model, because the device is now "good enough". Just like how a dual core 3ghz pc from 2005 with 4gb of ram is still good enough today for 95% of use cases.