It would break many of the programs which rely on 2.6.x naming scheme. Linus always was on the side of not breaking things if they do work. It seems that he changed his views for this particular case.
I can't actually think of the use case for needing to parse the complete version string outside the kernel, but it doesn't sound stupid to me that you'd assume a canonical format that's been there for a decade and a half. Apparently wrong, and decidedly unimportant & bikesheddy, but not necessarily stupid.
I agree with you that assuming the kernel would always be 2.6.x is not OK, I'd just assumed noone would be that shortsighted and stupid. :)
Hopefully by the time the final 3.0 is out, we'll have that extra zero all figured out."
Source: http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6...
When Torvalds bumps the major version number with little warning, he signals that anything which depends on it always being '2.x.x' or, worse, '2.6.x', is not acceptable and will not be tolerated.
I've lived in both worlds. I prefer Linux. I'm not alone.