Skip to contentSkip to conversation

PROPOSAL: New NEW development model

DISCUSSION John Richard Moser 4 participants 6 messages Oct 26, 2004

lore
Share & tools
lore
#0 lore raw

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

In lieu of the recent unpleasantness[1] about the new development model, and of some unexpected nastiness[2] as well, I propose that a very slight modification be made to the current development model. This would merge the previous and new development models (as far as my understanding is on them) and create a solution for all.

[1] http://lkml.org/lkml/2004/10/22/497
[2] http://lkml.org/lkml/2004/10/26/155

Previously, there were "Stable" and "Development" branches. The "Stable" branches, such as 2.2 and 2.4, would stagnate, and rarely have backported features. The "Development" branches such as 2.3 and 2.5 would be hammered and mixed with patches, then cleaned up until they were fairly stable. Then they'd be hammered with more patches, then cleaned up, then frozen, cleaned up until "stable" and then released.

Under the new development model, the "Stable" branch appears to be more "volatile." New patches and heavy VM and scheduling changes are freely added, as long as they're stable. It's not quite "Unstable," but it's closer to "Development" than "Stable."

I propose that a new development model involving "Stable" and "Volatile" branches be made. This would allow development to progress, but still provide a stable kernel for patch maintainers to work with.

The "Stable" kernel will be the even releases, 2.6 2.8 3.0 and so on. These should only have security fixes and bug fixes. No backporting should be done. Drivers are a grey area; although this development model aims to evade the backporting of drivers.

The "Volatile" kernel will be the odd releases, 2.7 2.9 3.1 and so on. These should follow the current development model of 2.6; new schedulers, new VM, new drivers, new features all around should be thrown at the "Volatile" kernel *IF* *THEY* *ARE* *MEANINGFULLY* *STABLE*. In this way, the "Volatile" kernel will be suitable for general use, and will remain fairly "stable."

A stable release cycle should be set up. Approximately once every six months, we'll say for example January 31 and July 31 to avoid the whole Christmas/New Years holiday glob interfering around Jan., the "Volatile" branch should be frozen into "Stable." At this point, the new "Stable" will immediately be equivalent to the latest state of the previous "Volatile." Also, because "Volatile" does not break to wait for "Stable" to "Sableize," it would immediately fork to the next "Volatile" series.

In effect, there will be no freezes ever. Once every 6 months, Volatile will fork into Stable and the next Volatile simultaneously, without a pause. Because Volatile will be continually kept up as a functional, usable kernel as 2.6 is now, there will be no need for a grace period for bug fixes.

Comments?

- --
All content of all messages exchanged herein are left in the Public Domain, unless otherwise explicitly stated.

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.6 (GNU/Linux)
Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org

iD8DBQFBfoQihDd4aOud5P8RAvMWAJwKqPxmwD6lMFVchDlhf2RrujhsIwCgkTrZ a6uQLECNDF68nfC4mMCTX8g=
=HLYr
-----END PGP SIGNATURE-----

#1 lore raw
Quoted 5 lines · John Richard Moser

In lieu of the recent unpleasantness[1] about the new development model, and of some unexpected nastiness[2] as well, I propose that a very slight modification be made to the current development model. This would merge the previous and new development models (as far as my understanding is on them) and create a solution for all.

Quoted 8 lines

[1] http://lkml.org/lkml/2004/10/22/497
[2] http://lkml.org/lkml/2004/10/26/155
Previously, there were "Stable" and "Development" branches. The "Stable" branches, such as 2.2 and 2.4, would stagnate, and rarely have backported features. The "Development" branches such as 2.3 and 2.5 would be hammered and mixed with patches, then cleaned up until they were fairly stable. Then they'd be hammered with more patches, then cleaned up, then frozen, cleaned up until "stable" and then released.

This is not useful to distinguish your "suggestion" from the status quo.

-- wli

#2 lore raw

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

William Lee Irwin III wrote:
| On Tue, Oct 26, 2004 at 01:06:42PM -0400, John Richard Moser wrote:
|

[...]

| This is not useful to distinguish your "suggestion" from the status quo. |

Help me out here, what's the status quo?

|
| -- wli
|

- --
All content of all messages exchanged herein are left in the Public Domain, unless otherwise explicitly stated.

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.6 (GNU/Linux)
Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org

iD8DBQFBfpZkhDd4aOud5P8RAnMyAJ4nbHZB6+2bdCxlu3bF4VXcwYhcAwCggV/I iU6/kloB9EgiCzKVAGC4cSc=
=JuRT
-----END PGP SIGNATURE-----

#3 lore raw
Quoted 5 lines · John Richard Moser

| This is not useful to distinguish your "suggestion" from the status quo. |

Help me out here, what's the status quo?

Pretty much exactly what you described already. Instead of "Volatile" being name 2.7, it's essentially 2.6.N-rcX. (Or whatever Linus has chosen as the new naming convention for as yet to be released stuff.)

At least that's how your proposal reads to me.

josh

#4 lore raw

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Josh Boyer wrote:

| On Tue, 2004-10-26 at 13:24, John Richard Moser wrote:
|
|>| This is not useful to distinguish your "suggestion" from the status quo.
|>|
|>
|>Help me out here, what's the status quo?
|
|
| Pretty much exactly what you described already.  Instead of "Volatile"
| being name 2.7, it's essentially 2.6.N-rcX.  (Or whatever Linus has
| chosen as the new naming convention for as yet to be released stuff.)
|
| At least that's how your proposal reads to me.
|

That's the point. I'd like the current model to split off the development into a separated branch, and do periodic stable releases. I think this would allow a much cleaner operation than simply consistantly enhancing the Stable branch. It's not that 2.6 is unstable/hackish/crashy/etc, it's that development away from mainline is more difficult which worries me.

I'd like to make a quick note here that "bugfixes and security updates" should be trimmed to just "bugfixes," since usually "security updates" just means "bugfixes for security holes." I'd like to avoid confusion with security updates such as new LSM hooks and SELinux extensions etc etc.

It would still require Linus to make periodic intermediate releases on the Stable branch for bug fixes; and someone would have to backport such bugfixes from Volatile, where they'd probably be found. So there's some upkeep, but minimal.

The point of restricting this upkeep as much as possible-- to the point of allowing only bugfixes-- would be to keep pressure off the Stable line maintainers. If release periods are approximately every 6 months, new drivers and features such as schedulers will be available every 6 months. No need to waste time and effort backporting what's soon to be stable anyway.

As for heavier work such as full scheduler rewrites, they could chase Stable and go into Volatile; although it may be a good idea for such things to make themselves known so that Volatile (and subsiquent Stables) can avoid overlapping work, making it difficult for the side project to keep up.

Still, a month or two to adapt to a new task scheduler out of 6 months leaves 4-5 months per stable release if the Volatile branch decides to hack up the scheduler. This is still a better scenario then "VM and scheduler infrastructures may change on any given release."

So OK, that's what's good here; so what's wrong with it? We've already established that there will be a minimal level of added work for a maintainer to keep the Stable up. Are there any other drawbacks? If not, any objections to trying to sell this one to Linus and Andrew? :)

| josh
|
| -
| To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
| the body of a message to [email protected]
| More majordomo info at  http://vger.kernel.org/majordomo-info.html
| Please read the FAQ at  http://www.tux.org/lkml/
|

- --
All content of all messages exchanged herein are left in the Public Domain, unless otherwise explicitly stated.

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.2.6 (GNU/Linux)
Comment: Using GnuPG with Thunderbird - http://enigmail.mozdev.org

iD8DBQFBfqSEhDd4aOud5P8RAt6QAKCBdRm+e5yHhL7xmTZuAbIYIhzXXwCaAhxF DsAfQhC3R4gy87LaB5DMXnM=
=7KX/
-----END PGP SIGNATURE-----

#5 lore raw
Quoted 11 lines · John Richard Moser

Still, a month or two to adapt to a new task scheduler out of 6 months leaves 4-5 months per stable release if the Volatile branch decides to hack up the scheduler. This is still a better scenario then "VM and scheduler infrastructures may change on any given release."

So OK, that's what's good here; so what's wrong with it? We've already established that there will be a minimal level of added work for a maintainer to keep the Stable up. Are there any other drawbacks? If not, any objections to trying to sell this one to Linus and Andrew? :)

Knock yourself out, I would be happy if they would just agree not to take features out of a stable series or intentionally break them. If new features don't break the old ones I don't think there's a persuasive argument to keep them out.

Signature · 3 lines
   -bill davidsen ([email protected])
"The secret to procrastination is to put things off until the
 last possible moment - but no longer"  -me