Is Google the new IBM, is Ice Cream Sandwich the new OS/2?
Well if you follow Android news you'll know that Google is pushing the requirement of having ALL phones released after Ice Cream Sandwich comes out they HAVE to be all on the same OS.
I fear Google will find themselves in a very difficult position with this one. By open-sourcing Android, effectively they no longer control 2.x. They can not stop it from being available and they can not stop it from being used. Furthermore, as it is open source, anybody can develop it in their own direction (i.e. fork it).
Consider the following examples from PC history:
In the early '80s there were dozens of incompatible PCs out there. None was made by IBM, which was a major player in the mainframe market, and wanted a slice of the PC market. They brought out a PC at a low price point after a short development time and achieved a dominant market share. They achieved this by using overwhelmingly off the shelf hardware and 3rd party software (MS DOS in particular). This dominant position and weakly controlled architecture allowed any number of clones into the market. Companies like Compaq reverse engineered the small number of non-standard components, bought everything else off the shelf, including the OS, and marketed a 100% IBM compatible computer, at a lower price. Realising they were losing market share, IBM decided to switch to a new OS, namely OS/2, effectively shutting out the cheap clones and regaining market share. Instead what happened was the large installed user base of old IBM machines and clones remained dominant, and OS/2 failed to gain any market share. IBM is now no longer even in the PC business (sold to Lenovo).
In the second half of the '90s, apart from a few Mac hold outs, pretty much every PC sold ran Windows. We all knew that Windows was a load of rubbish ('95, '98 and particularly ME), so when newer versions came out, everyone upgraded to the new version. When a well established, stable, usable version was finally achieved with XP, development slowed, and it became dominant. When MS tried to get everyone to move on to the next version, Vista, and now Windows 7, a very large number of people had no desire to move on, and because of the large installed user base of XP, it remains the dominant OS. If a software developer wants to sell their software, they have to make it work with XP.
So what does this have to do with Android? At present, Google is in a position like IBM was at the beginning of the 1990s. They do not control the hardware (the standard Android phone is not a concept controlled by Google), and they do not control the OS (because it's open source, anybody can do whatever they want with 2.x). OEMs like 2.x because they can control and customise the OS in ways that suit them. Even if a large number of new phones come out with 4.x on them, the user base will still be dominated by 2.x users (the Windows XP problem, exacerbated by the fact that a significant number of current handsets can not be upgraded without the users rooting them). Developers will therefore need to continue to support 2.x in their apps if they want to reach the bulk of the user base.
The problem is this: Google can not force an upgrade from 2.3, in the same way that IBM could not force an upgrade to OS/2. Because it is open source, OEMs can continue to install 2.3 (and modify it as much as they like, effectively forking Android), and there is nothing Google can do to prevent this. More importantly, the talk of adding restrictions to how OEMs and networks are allowed to implement 4.x provides a positive disincentive to the OEMs adopting it. Old users won't (in many cases can't) upgrade, so developers continue to support the old standard. OEMs lose commercial freedom by upgrading, and the need to support the existing user base severely limits the uptake of new functionality by developers (diluting the "pull" factors for users to upgrade).
The net result here is either further fragmentation (several forks of Android), stagnation (everyone sticks with 2.3 or 100% compatibles) or Google backs down on its attempts to enforce user experience (by open sourcing new versions). None of these is particularly good for the end user, IMO.