Erik,

I couldn't find anything but a reference to the Do-while Jones article. I suggest you find it and read it - then you'll understand the point I'm making.


http://www.dtweed.com/circuitce...do.htm#617


Jack also makes reference to the article and adds his comments.

http://www.ganssle.com/articles/acodeisr.htm

To fully test all the program paths in a system without interrupts is quite easy - add one interrupt and the task becomes monumental. Whilst I agree with the points that Do-while Jones makes, I tend to follow Jack's more pragmatic approach.

What we're discussing here is a computer scientists view vs an engineering view. Theory vs pragmatism. Theory is fine, but if you can't make it work in the real world then it remains a curiosity. In most instances, it would be extremely difficult (if not impossible) to build the average app without interrupts.
Therac 25 shows what happens with the misuse of interrupts.

Back to your point#5 - I think it is too 'hard' and too broad a statement. Way too prescriptive. I doubt you could prove its validity in many cases.