>I'm feeling in a good mood tonight, so I will just ask this: how often
>do you boot a 1.4 or 1.3 kernel on a 1.5 system ? 

I cant speak for David, but I boot 1.5.3-pre kernels on a -current
system often enough that I reworked top and netstat to fallback to the
old, 32-bit APIs if the new APIs for UVM stats failed.  (I'll test
pcmcia/cardbus support , but I am *NOT* going to retrograde my laptop
to do it!)  I keep wondering whether to commit those changes.
It only helps 1.5 branch testers, tho.



>A 1.5 kernel on a 1.4 system maybe, or more
>likely mixing -current and 1.5.  I do see the problem though, but I'm
>sure there have been more (or at least equal) drastic things in the past ?

Its never been officially supported, but those of us who try it are
pleasantly surprised by the _very_ low number of "flag days" which
broke stuff so badly you couldn't at least boot and login single-user
with an "old" kernel on a "new" filesystem.

Introduction of pread()/pwrite(), is the one I remember off-hand.
And that only broke due to changing startup code to use it, more-or-less
"just to show its really there"