FC5T2 and ndiswrapper

[Arjan van de Ven]

On Sat, 2006-01-21 at 15:02 +0100, David Nielsen wrote:
> lør, 21 01 2006 kl. 08:50 -0500, skrev Neal Becker:
> 
> > It's not quite correct, but OK for fedora devel (problem is kernel claims to
> > be 2.6.15 when it's really 2.6.16gitsomething)
> 
> I wouldn't call that a problem really, more a sane and natural design
> decision - afterall the final 2.6.16 isn't out yet why should we give
> people the impression they are running it.
> It's my impression that rpm doesn't take kindly to the kernel versioning
> scheme, that is to say we can't name kernels by the git release as part
> of the version number - the current solution must therefore be the most
> natural way to name kernels

it's more complex than that.. even if it were called 2.6.16... it's
quite likely that the api will change again before final 2.6.16, so if
the kernel started pretending to be 2.6.16 early, then those later api
changes would be in the same boat but worse, since then the "2.6.16"
ifdef couldn't be used reliably for things that REALLY are in 2.6.16.
Think of it in terms of api additions..

"2.6.16" means "all the functionality as of the 2.6.16 release".

the muddy middle area.. well there's no way to get it really right no
matter, and then there's of course the patches added in the spec as well
that may change api.

Maybe that's just many words for saying "the kernel has no stable
internal API, live with it".