[madnak]

[ QUOTE ]
mad/gump,

how does your answer change when the problem is changed to "fit the set of programs that people wished to / did program over the past 5 years (still only computation programs) into those 3 categories"

[/ QUOTE ]

Number 1 definitely wins. People just don't want to run a program on their computer that takes over a year to spit out a given piece of information. I'm not sure how distributed programming fits into your model, since it really doesn't meet the constraints (it has to "phone home" after all), but even if we include those programs there aren't that many of them.

The development of a 1 year+ program isn't viable, so even though those will represent the majority of the programs in theory, they'll represent a minority in practice.

In terms of group 2, I'm not sure. I can't think of much other than 3d graphics software that takes so long just to give a certain output. 1 definitely wins.

There might be some long simulations brewing, but probably most are written for supercomputers in research environments. And 1 year is still a long time. I mean, even most servers won't even see 1 full year of uninterrupted uptime.