F7 Plan (draft)

[Dave Jones]

On Tue, Dec 19, 2006 at 12:59:05AM -0500, Bill Nottingham wrote:
 > Hey, F7 needs a plan. So here's one

Nice work. Hannibal Smith would be proud.

 > 2. Merge Core and Extras in source control
 > Accountable: Jesse Keating, EVERYONE
 > 
 > Important dates: Leaf packages may start to be moved at <near future>.
 > All packages WILL be moved by Test 2, gating on feature #1. Reviews will
 > take place, there will be an automated system to determine what needs
 > review.

slight tangential thing:
Something that will be needed at some point is an external variant of
qabot (for outsiders, we have an irc bot internally that allows things like..

/m qabot whoowns kernel
<qabot> kernel is owned by Dave Jones <davej at redhat.com>

This would make a nice 'easy' project for a community member wanting to
contribute to Fedora.  It's been a gripe of mine for a while that no
equivalent existed for extras. Now, I guess the issue is being forced.

 > 4. A Fedora Desktop spin of F7
 > Accountable: Jesse Keating (rel-eng), Jonathan Blandford, the Fedora
 >   Desktop SIG (what, you say there isn't one?)
 > 
 > One of our release targets. Needs defined by Test 2. Test 1 can
 > be an old-style tree.
 > 
 > 5. A Fedora Server spin of F7
 > Accountable: Jesse Keating (rel-eng), YOUR NAME HERE!
 > 
 > Another one of our release targets. Needs defined by Test 2.

I'm prepared to step up and lend a hand defining this if Jesse feels
overwhelmed (I note he has a number of the big ticket items).

 > 8. Rock Solid Wireless
 > Accountable: Dan Williams, Dave Jones
 > 
 > For every wireless card we support, we should have working, tested
 > support - it comes up on boot, it associates, it works with WEP and
 > WPA, all out of the box. A requisite for hitting the laptop market.
 > Drive fixes through NM, wpa_supplicant, kernel.

Yes.  One thing needed for this will be at least one of every piece
of hardware we care about, are we even close to this today ?

 > 9. Wireless Firmware
 > Accountable: Fedora-legal, me!
 > For every wireless card we support and can get firmware to, we should
 > work to make it included so things work out of the box.

Oh yes. bring. it.

 > 10. Boot and shutdown speedup
 > Accountable: me!, David Zeuthen, YOUR NAME HERE
 >
 > We do enough Stupid Stuff that we can make easy improvements to startup
 > and shutdown time without large systemic changes. Includes: tagging of
 > scripts that don't need shut down, profiling of boot, potential changes
 > to how we organize disk blocks, and more.

I'll have at least a partial involvement with this when I get around
to integrating danpb's systemtap profiling goodness into the kernel rpm.
Plus I love watching a good trainwreck :)

 > 16. KVM virtualization support
 > Accountable: Dan Berrange, Daniel Veillard
 > 
 > Get KVM support, including QEMU, into virt-manager and libvirt. Make
 > it so the same tools can use KVM and Xen. Integrate networking support
 > so the same networking can be used for Xen, KVM, and maybe baremetal.
 > Special bonus: lhype?

lhype may take quite a while.  I'm optimistically thinking we can squeeze
2.6.21 into FC7. I expect .20 to land say, mid to end of January.
given where lhype is today though, that's still not going to be a huge
improvement, so at best we can hope for a 'tech preview' type thing that
may be useful for people actually hacking on it.

I fear we may actually have to support Xen for at least one more Fedora release.

 > 18. Move to using libata for PATA support
 > Accountable: Dave Jones, Peter Jones, potentially other Joneses

Jeff Garzik and Alan Cox can be honorary Joneses if they lend a hand where needbe :)

 > Make upgrades seamless, because not booting is bad. Must be done
 > by TEST 1.

This is the #1 things that makes me worry about meeting test1 on time.
Worse case scenario, we can reenable the old PATA drivers rather than slip though,
and then try again for test2.  Still gives me the creeps though, but
then again, it's only been a week so far, so all the scary looking bug reports
may all disappear by test1.   The xmas break doesn't make it any easier :-/

Worse yet, pretty much everything blocks on this.

 > 20. Improved firewire support
 > Accountable: Kristian Høgsberg
 > 
 > Make it sane/supportable/etc.

Can't really be any worse than the current situation, so what the hell.
And it's going upstream eventually.

 > 21. Real-time kernel
 > Accountable: Ingo Molnar, Dave Jones
 > Because fake-time kernels are so last year

Scary. Invasive. Bits going upstream periodically.
Not convinced this is a good idea for us to carry with all the
other huge stuff we have that we're trying to shed.
Right now for eg, all the big stuff is busted to hell.
Tux, Xen, signed modules etc.

 > 22. Tickless kernel support
 > Accountable: Ingo Molnar, Dave Jones
 > 
 > Saving power is good, mmkay?

Indeedy. Hopefully Ingo can whack the remaining oddnesses and this
will get upstream real soon.

 > 23. Fix wakeups across the distribution
 > Accountable: me, EVERYONE
 > 
 > Because, without this, having a tickless kernel is pointless.

Yes++
Arjan had a useful tracking bug on this problem. I recommend using it
(or making a new one) so that interested parties can subscribe to it.

 > 25. LiveCD
 > Accountable: David Zeuthen, Jeremy Katz, Jesse Keating
 > 
 > Get LiveCDs built and shipped as part of the release process. Make
 > them installable.
 > 
 > Section the Third: Additions, Constructive Criticism, and Bellyaching
 > 
 > Got more? Let me know!

26. Nouveau.
Accountable: Me, Ajax.

Because binary drivers are the suck, and pimping rev-eng efforts like
this should be what distros should be advocating instead of whining
about how they 'need' to ship binary drivers for their cracktastic desktop effects.
Lets lead by example by showcasing this work, even if its rough around
the edges. We can always improve on it more later.

		Dave

-- 
http://www.codemonkey.org.uk