Theres no need to get snippy.
I think my basic point has remained constant - although I concede I may have different assumptions, leading to my glossing over different areas.
I also take your point about the reduction in prevalance of ROS users. I don't think that affects my underlying argument though.
Let me try to put it another way though, and see if you agree.
* 'Many' OSes are capable of being compiled for different architectures.
* However, in practice, for commodity OSes  at any one point in time, there will be a single preferred architecture. This is demonstrated by Macs migration from one architecture to another. This is a support issue at several levels.
* There don't seem to be the resources to make ROS easily portable.
* As architectures vary in relative power and the current leader may in a sense be maxing out, it may be wise to try to target where general processing power increases are to be had.
I you read my previous posts carefully I think you'll see this position in them. I'm not saying that it wouldn't be nice to have completely portable sources for ROS. Just that it isn't necessarily something to panic about.
 I may also be wrong of course, but humour me..
 Yes yes I know, although most of those are UNIX. Also note I'm writing about practice, not theory.
 Windows, Macs, RISC OS, (perhaps) Amiga OS etc..
 This needn't be the case, even for proprierary software, if a secure universal system of compilation on installation existed. Obviously this system would itself need to be portable.
 not just (third party?) developers as you imply. Hardware is affected by the CPU. To give a far-reaching example, RAM requirements vary dependent on code density.