[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: kernel development approach for fedora



Mail Lists <lists <at> sapience.com> writes:
>    Linus switched kernel development away from large releases (odd/even
> major numbers) with infrequent release cycles and instead switched to
> something more continuous - essentially small  rapid changes and
> frequent snapshots to stable.

This is essentially how Fedora already works.

>    Would the kernel release style be suited to fedora - for much the
> same reasons possibly. They seem to manage getting big changes in there
> too. And it would be in spirit with the bleeding edge of we desire in
> fedora.

It is, in fact Fedora is already using it. ;-)

>    This mode would be basically always updating and never/seldom
> installing ..

Nothing prevents you to update from one Fedora release to the next. It can even 
be done with yum (or apt or smart if that's what you fancy, and zypper coming 
soon as yet another option).

> perhaps by some measure the rawhide to stable is similar
> .. but there are definite differences. As rawhide is not merged into
> stable ..

Stable releases are cut from Rawhide at regular points in time (and frequent - 
6 months is about the shortest cycle which makes sense for a distribution!) 
just like the kernel does it. And we do also backport bugfixes from Rawhide to 
the stable releases, that's what updates are for. In fact, we even sometimes 
backport new features, which is something the kernel doesn't do.

What you seem to implicitly suggest is a "rolling release" model, but that's 
not how the kernel works, in fact we're closer to it than the kernel (because 
we often push out new features to the stable releases). If you really want 
a "rolling release", then just use Rawhide, because that's what it is. But then 
you get to keep the pieces. ;-) There's no way to do development without 
occasionally breaking something, you have to give us time to stabilize things 
for a release.

> so our current method seems to resemble the older kernel
> development approach.

I don't see how that follows at all. The older kernel approach was closer to 
how RHEL works, not Fedora. (You can compare Fedora with 2.5 and RHEL with 2.4, 
a new RHEL release is like a new kernel release was at the time.)

        Kevin Kofler


[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]