So I've been busily merging stuff, and just wanted to send out a quick
reminder that I warned people in the 39 announcement that this might
be a slightly shorter merge window than usual, so that I can avoid
having to make the -rc1 release from Japan using my slow laptop (doing
"allyesconfig" builds on that thing really isn't in the cards, and I
like to do those to verify things - even if we've already had a few
cases where arch include differences made it less than effective in
finding problems).
And judging by the merge window so far, that early close (probably
Sunday - I'll be on airplanes next Monday) looks rather likely. I
already seem to have a fairly sizable portion of linux-next in my
tree, and there haven't been any huge upsets.
So anybody who was planning a last-minute "please pull" - this is a
heads-up. Don't do it, you might miss the window entirely.
Did I miss any major development mailing lists with stuff pending?
Linus
PS. The voices in my head also tell me that the numbers are getting
too big. I may just call the thing 2.8.0. And I almost guarantee that
this PS is going to result in more discussion than the rest, but when
the voices tell me to do things, I listen.
* Linus Torvalds <[email protected]> wrote:
> PS. The voices in my head also tell me that the numbers are getting too big.
> I may just call the thing 2.8.0. And I almost guarantee that this PS is going
> to result in more discussion than the rest, but when the voices tell me to do
> things, I listen.
I really hope there's also a voice that tells you to wait until .42 before
cutting 3.0.0! :-)
Ingo
On Mon, May 23, 2011 at 12:13:29PM -0700, Linus Torvalds wrote:
> PS. The voices in my head also tell me that the numbers are getting
> too big. I may just call the thing 2.8.0. And I almost guarantee that
> this PS is going to result in more discussion than the rest, but when
> the voices tell me to do things, I listen.
If you do this, I will buy you a bottle of whatever whiskey you want
that I can get my hands on in Tokyo next week.
{crosses fingers}
greg k-h
On Mon, 23 May 2011, Linus Torvalds wrote:
> PS. The voices in my head also tell me that the numbers are getting
> too big. I may just call the thing 2.8.0. And I almost guarantee that
> this PS is going to result in more discussion than the rest, but when
> the voices tell me to do things, I listen.
So the voices tell you to avoid .42 ?
Thanks,
tglx
On Mon, 2011-05-23 at 12:22 -0700, Greg KH wrote:
> On Mon, May 23, 2011 at 12:13:29PM -0700, Linus Torvalds wrote:
> > PS. The voices in my head also tell me that the numbers are getting
> > too big. I may just call the thing 2.8.0. And I almost guarantee that
> > this PS is going to result in more discussion than the rest, but when
> > the voices tell me to do things, I listen.
>
> If you do this, I will buy you a bottle of whatever whiskey you want
> that I can get my hands on in Tokyo next week.
I can recommend Hanyu Ace of Spades ... I can even arrange to be on
hand just to make sure it's as good as it should be ...
James
On Mon, 23 May 2011 21:25:25 +0200 (CEST) Thomas Gleixner wrote:
> On Mon, 23 May 2011, Linus Torvalds wrote:
> > PS. The voices in my head also tell me that the numbers are getting
> > too big. I may just call the thing 2.8.0. And I almost guarantee that
> > this PS is going to result in more discussion than the rest, but when
> > the voices tell me to do things, I listen.
>
> So the voices tell you to avoid .42 ?
They tell him to avoid the question to which 42 is the answer.
---
~Randy
*** Remember to use Documentation/SubmitChecklist when testing your code ***
On Mon, May 23, 2011 at 12:20 PM, Ingo Molnar <[email protected]> wrote:
>
> I really hope there's also a voice that tells you to wait until .42 before
> cutting 3.0.0! :-)
So I'm toying with 3.0 (and in that case, it really would be "3.0",
not "3.0.0" - the stable team would get the third digit rather than
the fourth one.
But no, it wouldn't be for 42. Despite THHGTTG, I think "40" is a
fairly nice round number.
There's also the timing issue - since we no longer do version numbers
based on features, but based on time, just saying "we're about to
start the third decade" works as well as any other excuse.
But we'll see.
Linus
On Tue, May 24, 2011 at 00:33, Linus Torvalds
<[email protected]> wrote:
> On Mon, May 23, 2011 at 12:20 PM, Ingo Molnar <[email protected]> wrote:
>>
>> I really hope there's also a voice that tells you to wait until .42 before
>> cutting 3.0.0! :-)
>
> So I'm toying with 3.0 (and in that case, it really would be "3.0",
> not "3.0.0" - the stable team would get the third digit rather than
> the fourth one.
>
> But no, it wouldn't be for 42. Despite THHGTTG, I think "40" is a
> fairly nice round number.
>
> There's also the timing issue - since we no longer do version numbers
> based on features, but based on time, just saying "we're about to
> start the third decade" works as well as any other excuse.
>
> But we'll see.
Maybe, 2011.x, or 11.x, x increasing for every merge window started this year?
This would better reflect the steady nature of the releases, but would
certainly break a lot of scripts. ;)
On Mon, May 23, 2011 at 01:21:26PM -0700, Randy Dunlap wrote:
>
> They tell him to avoid the question to which 42 is the answer.
What 2.6 Linux kernel version was the last before 3.0?
-- Steve
On 5/23/11, Linus Torvalds <[email protected]> wrote:
> On Mon, May 23, 2011 at 12:20 PM, Ingo Molnar <[email protected]> wrote:
>>
>> I really hope there's also a voice that tells you to wait until .42 before
>> cutting 3.0.0! :-)
I think, the best time for this, after reorganize the ARM arch folder / tree.
>
> So I'm toying with 3.0 (and in that case, it really would be "3.0",
> not "3.0.0" - the stable team would get the third digit rather than
> the fourth one.
>
> But no, it wouldn't be for 42. Despite THHGTTG, I think "40" is a
> fairly nice round number.
>
> There's also the timing issue - since we no longer do version numbers
> based on features, but based on time, just saying "we're about to
> start the third decade" works as well as any other excuse.
>
> But we'll see.
>
> Linus
> --
> 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/
>
On Mon, May 23, 2011 at 01:33:48PM -0700, Linus Torvalds wrote:
> On Mon, May 23, 2011 at 12:20 PM, Ingo Molnar <[email protected]> wrote:
> >
> > I really hope there's also a voice that tells you to wait until .42 before
> > cutting 3.0.0! :-)
>
> So I'm toying with 3.0 (and in that case, it really would be "3.0",
> not "3.0.0" - the stable team would get the third digit rather than
> the fourth one.
I like that, it would make things much easier for me to keep track of
stuff.
> But no, it wouldn't be for 42. Despite THHGTTG, I think "40" is a
> fairly nice round number.
>
> There's also the timing issue - since we no longer do version numbers
> based on features, but based on time, just saying "we're about to
> start the third decade" works as well as any other excuse.
That sounds reasonable as well.
greg k-h
On Mon, May 23, 2011 at 4:33 PM, Linus Torvalds
<[email protected]> wrote:
> On Mon, May 23, 2011 at 12:20 PM, Ingo Molnar <[email protected]> wrote:
>>
>> I really hope there's also a voice that tells you to wait until .42 before
>> cutting 3.0.0! :-)
>
> So I'm toying with 3.0 (and in that case, it really would be "3.0",
> not "3.0.0" - the stable team would get the third digit rather than
> the fourth one.
Could we set a goal of having 3.0 be the first release with a totally
cleaned up ARM arch? That would give everyone a good target to work
towards.
--
Jon Smirl
[email protected]
On Mon, May 23, 2011 at 01:33:48PM -0700, Linus Torvalds wrote:
> On Mon, May 23, 2011 at 12:20 PM, Ingo Molnar <[email protected]> wrote:
> >
> > I really hope there's also a voice that tells you to wait until .42 before
> > cutting 3.0.0! :-)
>
> So I'm toying with 3.0 (and in that case, it really would be "3.0",
> not "3.0.0" - the stable team would get the third digit rather than
> the fourth one.
If we change from 2.6.X to 3.X, then if we don't change anything else,
then successive stable release will cause the LINUX_VERSION_CODE to be
incremented. This isn't necessary bad, but it would be a different
from what we have now.
- Ted
On Mon, 23 May 2011 19:17:21 -0400 Ted Ts'o wrote:
> On Mon, May 23, 2011 at 01:33:48PM -0700, Linus Torvalds wrote:
> > On Mon, May 23, 2011 at 12:20 PM, Ingo Molnar <[email protected]> wrote:
> > >
> > > I really hope there's also a voice that tells you to wait until .42 before
> > > cutting 3.0.0! :-)
> >
> > So I'm toying with 3.0 (and in that case, it really would be "3.0",
> > not "3.0.0" - the stable team would get the third digit rather than
> > the fourth one.
>
> If we change from 2.6.X to 3.X, then if we don't change anything else,
> then successive stable release will cause the LINUX_VERSION_CODE to be
> incremented. This isn't necessary bad, but it would be a different
> from what we have now.
It's just another little thing to break several scripts...
---
~Randy
*** Remember to use Documentation/SubmitChecklist when testing your code ***
On 05/23/2011 04:17 PM, Ted Ts'o wrote:
> On Mon, May 23, 2011 at 01:33:48PM -0700, Linus Torvalds wrote:
>> On Mon, May 23, 2011 at 12:20 PM, Ingo Molnar <[email protected]> wrote:
>>>
>>> I really hope there's also a voice that tells you to wait until .42 before
>>> cutting 3.0.0! :-)
>>
>> So I'm toying with 3.0 (and in that case, it really would be "3.0",
>> not "3.0.0" - the stable team would get the third digit rather than
>> the fourth one.
>
> If we change from 2.6.X to 3.X, then if we don't change anything else,
> then successive stable release will cause the LINUX_VERSION_CODE to be
> incremented. This isn't necessary bad, but it would be a different
> from what we have now.
>
That sounds like a good thing.
-hpa
Another advantage of switching numbering models (ie 3.0 instead of
2.8.x) would be that it would also make the "odd numbers are also
numbers" transition much more natural.
Because of our historical even/odd model, I wouldn't do a 2.7.x -
there's just too much history of 2.1, 2.3, 2.5 being development
trees. But if I do 3.0, then I'd be chucking that whole thing out the
window, and the next release would be 3.1, 3.2, etc..
And then in another few years (probably before getting close to 3.40,
so I'm not going to make a big deal of 3 = "third decade"), I'd just
do 4.0 etc.
Because all our releases are supposed to be stable releases these
days, and if we get rid of one level of numbering, I feel perfectly
fine with getting rid of the even/odd history too.
Linus
On Mon, May 23, 2011 at 03:21:21PM -0700, Greg KH wrote:
> On Mon, May 23, 2011 at 01:33:48PM -0700, Linus Torvalds wrote:
> > On Mon, May 23, 2011 at 12:20 PM, Ingo Molnar <[email protected]> wrote:
> > > I really hope there's also a voice that tells you to wait until .42 before
> > > cutting 3.0.0! :-)
> >
> > So I'm toying with 3.0 (and in that case, it really would be "3.0",
> > not "3.0.0" - the stable team would get the third digit rather than
> > the fourth one.
>
> I like that, it would make things much easier for me to keep track of
> stuff.
As long as 3.14 turns into a long-term support kernel and gets up to 159 ...
In all serious, I'm very supportive of this move. I'm heartily sick
of people claiming "we have version 2.6 support" when they really mean
they haven't updated since version 2.6.9. Yeah, congratulations, you're
seven years out of date.
--
Matthew Wilcox Intel Open Source Technology Centre
"Bill, look, we understand that you're interested in selling us this
operating system, but compare it to ours. We can't possibly take such
a retrograde step."
Hi Linus,
On 05/23/2011 04:33 PM, Linus Torvalds wrote:
> On Mon, May 23, 2011 at 12:20 PM, Ingo Molnar <[email protected]> wrote:
>>
>> I really hope there's also a voice that tells you to wait until .42 before
>> cutting 3.0.0! :-)
>
> So I'm toying with 3.0 (and in that case, it really would be "3.0",
> not "3.0.0" - the stable team would get the third digit rather than
> the fourth one.
A few months ago, I briefly considered suggesting that the demise of the BKL
would be a suitable milestone for the numbering shakeup.
But I am a mere mortal lurker, and I remember past flame-fests this topic
spawned. So I chickened out.
As a small-scale linux evangelist, I would sure like to skip the explanation
of the version numbers.
Phil
* Linus Torvalds <[email protected]> wrote:
> Another advantage of switching numbering models (ie 3.0 instead of
> 2.8.x) would be that it would also make the "odd numbers are also
> numbers" transition much more natural.
Yeah, it sounds really good to get rid of the (meanwhile) meaningless
"2.6." prefix from our version code and iterate it in a more
meaningful way.
I suspect the stable team and distros will enjoy the more meaningful
third digit as well: it will raise the perceived importance of
stabilization and packaging work.
Btw., we should probably remove the fourth (patch) level, otherwise
distros might feel tempted to fill it in with their own patch-stack
version number, which would result in confusing "3.3.1.5" meaning
different things on different distros - while 3.3.1-5.rpm style of
distro kernel package naming denotes the distro patch level more
clearly.
I don't think the odd/even history will linger too long: in practice
we'll iterate through 3.1, 3.2, 3.3 and 3.4 rather quickly, in the first
year, so any residual notion of stable/unstable will be gone within a
few iterations.
> Because of our historical even/odd model, I wouldn't do a 2.7.x -
> there's just too much history of 2.1, 2.3, 2.5 being development
> trees. But if I do 3.0, then I'd be chucking that whole thing out the
> window, and the next release would be 3.1, 3.2, etc..
>
> And then in another few years (probably before getting close to 3.40,
> so I'm not going to make a big deal of 3 = "third decade"), I'd just
> do 4.0 etc.
Perhaps we could do 4.0 once the last bit of -rt hits upstream? /me ducks
> Because all our releases are supposed to be stable releases these
> days, and if we get rid of one level of numbering, I feel perfectly
> fine with getting rid of the even/odd history too.
They are very stable releases as far as i'm concerned - i can pretty
confidently run and use -rc2 and better kernels on my boxes these days
and could do so for the past few years.
Thanks,
Ingo
* Linus Torvalds <[email protected]> wrote:
> On Mon, May 23, 2011 at 12:20 PM, Ingo Molnar <[email protected]> wrote:
> >
> > I really hope there's also a voice that tells you to wait until .42 before
> > cutting 3.0.0! :-)
>
> So I'm toying with 3.0 (and in that case, it really would be "3.0",
> not "3.0.0" - the stable team would get the third digit rather than
> the fourth one.
>
> But no, it wouldn't be for 42. Despite THHGTTG, I think "40" is a
> fairly nice round number.
Also, in all fairness, we should probably display a certain amount of humility:
while Linux has certainly reached milestones such as world domination (as far
as large and small computers are concerned), so calling it 3.0 is a fair deal,
we probably have to wait for version 42.0 before we can consider the Linux
kernel to be the ultimate answer to life, universe and everything.
Thanks,
Ingo
On Tuesday 24 May 2011, Linus Torvalds wrote:
> Another advantage of switching numbering models (ie 3.0 instead of
> 2.8.x) would be that it would also make the "odd numbers are also
> numbers" transition much more natural.
>
> Because of our historical even/odd model, I wouldn't do a 2.7.x -
> there's just too much history of 2.1, 2.3, 2.5 being development
> trees. But if I do 3.0, then I'd be chucking that whole thing out the
> window, and the next release would be 3.1, 3.2, etc..
I like that. While I don't really care if you call it 2.7, 2.8 or 3.0
(or 4.0 even, if you want to keep continuity following .38 and .39),
the current 2.5/2.6 numbering cycle is almost 10 years old and has
obviously lost all significance.
The only reason I can see that would make it worthwhile waiting for
is if the enterprise and embedded people were to decide on a common
longterm kernel and call that e.g. 2.7.x or 2.8.x while you continue with
2.9.x or 3.0.x or 3.x. My impression is however that the next longterm
release is still one or two years away, so probably not worth waiting
for and hard to estimate in advance.
> Because all our releases are supposed to be stable releases these
> days, and if we get rid of one level of numbering, I feel perfectly
> fine with getting rid of the even/odd history too.
We still have stable and unstable releases, except that you call the
unstable ones -rcX and they are all nice and short, unlike the infamous
2.1.xxx series ;-)
IMHO simply changing the names from 2.6.40-rcX to 2.7.X and from
2.6.40.X to 2.6.8.X etc would be the most straightforward change
if you want to save the 3.0 release for a special moment.
Enough bike shedding from my side, please just make a decision.
Arnd
On Tuesday 2011-05-24 01:33, Linus Torvalds wrote:
>Another advantage of switching numbering models (ie 3.0 instead of
>2.8.x) would be that it would also make the "odd numbers are also
>numbers" transition much more natural.
>
>Because of our historical even/odd model, I wouldn't do a 2.7.x -
>there's just too much history of 2.1, 2.3, 2.5 being development
>trees.
.oO(Though once 2.{7 or more, odd} trickle into the distros, it would
become pretty much apparent that they are not devel.)
>And then in another few years (probably before getting close to 3.40,
>so I'm not going to make a big deal of 3 = "third decade"), I'd just
>do 4.0 etc.
While 2.6 has certainly worn out, already thinking of a 4.0 is highly
reminiscient of the version number arms race Firefox and ChromeBrowser
are doing currently.
>Because all our releases are supposed to be stable releases these
>days, and if we get rid of one level of numbering, I feel perfectly
>fine with getting rid of the even/odd history too.
If I remember past-time discussions right, ELF was the contributing
factor to bump the major number to 2.0 back then; ever since 2.0, no
similarly breakthrough-ing event has occurred.
2011/5/24 Jan Engelhardt <[email protected]>:
> On Tuesday 2011-05-24 01:33, Linus Torvalds wrote:
>
>>Another advantage of switching numbering models (ie 3.0 instead of
>>2.8.x) would be that it would also make the "odd numbers are also
>>numbers" transition much more natural.
>>
>>Because of our historical even/odd model, I wouldn't do a 2.7.x -
>>there's just too much history of 2.1, 2.3, 2.5 being development
>>trees.
>
> .oO(Though once 2.{7 or more, odd} trickle into the distros, it would
> become pretty much apparent that they are not devel.)
>
>>And then in another few years (probably before getting close to 3.40,
>>so I'm not going to make a big deal of 3 = "third decade"), I'd just
>>do 4.0 etc.
>
> While 2.6 has certainly worn out, already thinking of a 4.0 is highly
> reminiscient of the version number arms race Firefox and ChromeBrowser
> are doing currently.
>
>>Because all our releases are supposed to be stable releases these
>>days, and if we get rid of one level of numbering, I feel perfectly
>>fine with getting rid of the even/odd history too.
>
> If I remember past-time discussions right, ELF was the contributing
> factor to bump the major number to 2.0 back then; ever since 2.0, no
> similarly breakthrough-ing event has occurred.
What then about BKL removal? Nice place to celebrate with version jump
and heaving some beers.
-Jacek
On Tuesday 2011-05-24 14:30, Jacek Luczak wrote:
>2011/5/24 Jan Engelhardt <[email protected]>:
>> On Tuesday 2011-05-24 01:33, Linus Torvalds wrote:
>>
>>>Another advantage of switching numbering models (ie 3.0 instead of
>>>2.8.x) would be that it would also make the "odd numbers are also
>>>numbers" transition much more natural.
>>>
>>>Because of our historical even/odd model, I wouldn't do a 2.7.x -
>>>there's just too much history of 2.1, 2.3, 2.5 being development
>>>trees.
>>
>> .oO(Though once 2.{7 or more, odd} trickle into the distros, it would
>> become pretty much apparent that they are not devel.)
>>
>>>And then in another few years (probably before getting close to 3.40,
>>>so I'm not going to make a big deal of 3 = "third decade"), I'd just
>>>do 4.0 etc.
>>
>> While 2.6 has certainly worn out, already thinking of a 4.0 is highly
>> reminiscient of the version number arms race Firefox and ChromeBrowser
>> are doing currently.
>>
>>>Because all our releases are supposed to be stable releases these
>>>days, and if we get rid of one level of numbering, I feel perfectly
>>>fine with getting rid of the even/odd history too.
>>
>> If I remember past-time discussions right, ELF was the contributing
>> factor to bump the major number to 2.0 back then; ever since 2.0, no
>> similarly breakthrough-ing event has occurred.
>
>What then about BKL removal? Nice place to celebrate with version jump
>and heaving some beers.
The BKL going away was not a change that would require new
userspace programs.
2011/5/24 Jan Engelhardt <[email protected]>:
> On Tuesday 2011-05-24 14:30, Jacek Luczak wrote:
>
>>2011/5/24 Jan Engelhardt <[email protected]>:
>>> On Tuesday 2011-05-24 01:33, Linus Torvalds wrote:
>>>
>>>>Another advantage of switching numbering models (ie 3.0 instead of
>>>>2.8.x) would be that it would also make the "odd numbers are also
>>>>numbers" transition much more natural.
>>>>
>>>>Because of our historical even/odd model, I wouldn't do a 2.7.x -
>>>>there's just too much history of 2.1, 2.3, 2.5 being development
>>>>trees.
>>>
>>> .oO(Though once 2.{7 or more, odd} trickle into the distros, it would
>>> become pretty much apparent that they are not devel.)
>>>
>>>>And then in another few years (probably before getting close to 3.40,
>>>>so I'm not going to make a big deal of 3 = "third decade"), I'd just
>>>>do 4.0 etc.
>>>
>>> While 2.6 has certainly worn out, already thinking of a 4.0 is highly
>>> reminiscient of the version number arms race Firefox and ChromeBrowser
>>> are doing currently.
>>>
>>>>Because all our releases are supposed to be stable releases these
>>>>days, and if we get rid of one level of numbering, I feel perfectly
>>>>fine with getting rid of the even/odd history too.
>>>
>>> If I remember past-time discussions right, ELF was the contributing
>>> factor to bump the major number to 2.0 back then; ever since 2.0, no
>>> similarly breakthrough-ing event has occurred.
>>
>>What then about BKL removal? Nice place to celebrate with version jump
>>and heaving some beers.
>
> The BKL going away was not a change that would require new
> userspace programs.
True but as you I guess - kind off - notice there's no such event that
would launch fireworks and we get features smoothly. By that then we
should celebrate killing old nightmares aka BKL. It's more like - lets
not find the reason but include one just to feel better. At the end
the simplified version convention is the best reason to do this cut
off. I even plan to send a truck full of chickens to Linus if this
will convince him :) Then, while describing new kernel deployment, I
won't need to pronounce the cool sounding - ,,Mika so I've now
installed two(dot)six(dot)thirty-five(dot)forty-one(dash)one
version.''
Cheers,
-Jacek
Version numbers should not be only numerology, but
connected to quality or progress. Comparing the situation
of the 1, 2.0, 2.2, 2.4, 2.6 versions, since starting 2.6
, the progress of the kernel, and also the situation of
Linux as a whole and of its use changed very. Exactly in
the last subversions with many new hardware
modules/drivers the quality improved very. At the same
time, the subversions becomes too much since the last new
version. Thus, it's time for 2.8 or even 3.0 in order to
mark this very improved quality and use of Linux since the
start of the 2 and 2.6 versions. wl
---
Professional hosting for everyone - http://www.host.ru
On Tue, May 24, 2011 at 9:40 AM, werner <[email protected]> wrote:
> Version numbers should not be only numerology, but connected to quality or
> progress. ?Comparing the situation of the 1, 2.0, 2.2, 2.4, 2.6 versions,
> since starting 2.6 , the progress of the kernel, and also the situation of
> Linux as a whole and of its use changed very. ?Exactly in the last
> subversions with many new hardware modules/drivers the quality improved
> very. ?At the same time, the subversions becomes too much since the last new
> version. ?Thus, it's time for 2.8 or even 3.0 in order to mark this very
> improved quality and use of Linux since the start of the 2 and 2.6 versions.
> ? ?wl
> ---
If i were to give my feeling, and i am doing it right now, i wish that
such version change to be backed with things like removal of dead code
for instance in the GPU side we would like at one point to drop non
KMS path and delete a whole bunch of files and dead API in the process
(at least that's on my bad santa list). Non KMS is in survival mode,
at this point we try to not break it but we don't give shit about it.
Bug against it are redirected to the closest black hole.
Cheers,
Jerome
> If we change from 2.6.X to 3.X, then if we don't change anything else,
> then successive stable release will cause the LINUX_VERSION_CODE to be
> incremented. This isn't necessary bad, but it would be a different
> from what we have now.
I think I prefer 3 digits. Otherwise we will have to pass 3.0, 3.1 and
3.11 all of which numbers still give older sysadmins flashbacks and will
have them waking screaming in the middle of the night.
Also saves breaking all the tools and assumptions people have been used
to for some many years
Alan
Can we drop most of MCA, EISA and ISA bus if we are going to have a big
version change ? A driver spring clean is much overdue and it's all in
git in case someone wishes to sneak out at midnight and bring some crawly
horror back from the dead.
Alan
On Mon, May 23, 2011 at 07:17:21PM -0400, Ted Ts'o wrote:
> > So I'm toying with 3.0 (and in that case, it really would be "3.0",
> > not "3.0.0" - the stable team would get the third digit rather than
> > the fourth one.
>
> If we change from 2.6.X to 3.X, then if we don't change anything else,
> then successive stable release will cause the LINUX_VERSION_CODE to be
> incremented. This isn't necessary bad, but it would be a different
> from what we have now.
It will require another bunch of changes to scripts that try to make sense
out of kernel Linux version numbers. It's a minor issue and we might be
better off doing something else than version number magic. Not last a
new major version number raises expectations - whatever those might be.
Ralf
On Tue, May 24, 2011 at 10:43 AM, Alan Cox <[email protected]> wrote:
> Can we drop most of MCA, EISA and ISA bus if we are going to have a big
> version change ? A driver spring clean is much overdue and it's all in
> git in case someone wishes to sneak out at midnight and bring some crawly
> horror back from the dead.
2.8 could mark the beginning of the great cleanup
--- work out the details of what needs to be cleaned and set a goal
--- remove old buses/driver, switch to device tree, graphics, 32/64
merges, etc
3.0 would mark its completion
--
Jon Smirl
[email protected]
A new major version number will reach high expectations. I think it's
better to wait till the power issues of 2.6.38/39 are fixed and proved
to be absent, before naming it Linux 3.0.
Anyway, wouldn't it be a great date to release Linux 3 on August 21,
2011? I know it's hard to achieve this exact date with a stable 3.0. If
you don't think a fixed date for a stable version is a good idea, you
could release the first RC of 3.0 on that date.
About the scripts: if they don't like the version number to be 1 digit
shorter, why not append an extra .0 to uname? This digit can be used for
bugfix releases, like 2.6.38.7. In the current system, an extra digit is
added for those releases. But if that extra digit will always be there,
and is 0 by default, scripts won't care about the lack of a 3rd digit.
Then, the 3.0 will be 3.0.0, with new releases (current 3rd digit) being
3.1.0, 3.2.0, ..., and bugfix releases (current 4th digit) being 3.0.1,
3.0.2, .... It might be a bit confusing after the switch, but if we
change to a 2-digit number, it's confusing too.
Scripts recognizing bugfixes as new releases is a smaller disaster than
scripts crashing or returning errors due to the lack of a 3rd digit.
I agree with Jon Smirl that it could be probably better to release 2.8
first, but there's also a good side on switching to 3.0. We are skipping
2.7, that means our version numbering system already changes a bit. Why
not change it completely then?
Sorry if I've missed out some discussion about this, I am not
subscribed. I have only read some messages in the LKML archive.
Albert Pool
On Tue, May 24, 2011 at 15:48:39 +0100, Ralf Baechle wrote:
On Mon, May 23, 2011 at 07:17:21PM -0400, Ted Ts'o wrote:
> > So I'm toying with 3.0 (and in that case, it really would be "3.0",
> > not "3.0.0" - the stable team would get the third digit rather than
> > the fourth one.
>
> If we change from 2.6.X to 3.X, then if we don't change anything else,
> then successive stable release will cause the LINUX_VERSION_CODE to be
> incremented. This isn't necessary bad, but it would be a different
> from what we have now.
It will require another bunch of changes to scripts that try to make sense
out of kernel Linux version numbers. It's a minor issue and we might be
better off doing something else than version number magic. Not last a
new major version number raises expectations - whatever those might be.
Ralf
On Tue, May 24, 2011 at 03:43:48PM +0100, Alan Cox wrote:
> Can we drop most of MCA, EISA and ISA bus if we are going to have a big
> version change ? A driver spring clean is much overdue and it's all in
> git in case someone wishes to sneak out at midnight and bring some crawly
> horror back from the dead.
Dunno about MCA but I doubt we can kill all of (E)ISA. i8253, i8259 and a
few others still refuse hard to die.
Is it worth to setup a system to track success / failure reports for
drivers and ditch drivers once there are no success reports for a driver
for too long? It may not be a good idea - people tend not report success
much more rarely than failure.
(On that matter, I wonder if there are 5.25" USB floppy drives ...)
Ralf
On Tuesday 2011-05-24 17:46, Ralf Baechle wrote:
>On Tue, May 24, 2011 at 03:43:48PM +0100, Alan Cox wrote:
>
>> Can we drop most of MCA, EISA and ISA bus if we are going to have a big
>> version change ? A driver spring clean is much overdue and it's all in
>> git in case someone wishes to sneak out at midnight and bring some crawly
>> horror back from the dead.
>
>Dunno about MCA but I doubt we can kill all of (E)ISA. i8253, i8259 and a
>few others still refuse hard to die.
>
>Is it worth to setup a system to track success / failure reports for
>drivers and ditch drivers once there are no success reports for a driver
>for too long? It may not be a good idea - people tend not report success
>much more rarely than failure.
>
>(On that matter, I wonder if there are 5.25" USB floppy drives ...)
If there were, they would appear as Mass Storage devices (at least
the 3.5" USB floppy gadgets do), and as such, don't depend on ISA
or the classic floppy driver at all.
On 05/24/2011 08:07 AM, [email protected] wrote:
> On Tue, May 24, 2011 at 10:43 AM, Alan Cox <[email protected]> wrote:
>> Can we drop most of MCA, EISA and ISA bus if we are going to have a big
>> version change ? A driver spring clean is much overdue and it's all in
>> git in case someone wishes to sneak out at midnight and bring some crawly
>> horror back from the dead.
>
> 2.8 could mark the beginning of the great cleanup
> --- work out the details of what needs to be cleaned and set a goal
> --- remove old buses/driver, switch to device tree, graphics, 32/64
> merges, etc
> 3.0 would mark its completion
>
I think this whole discussion misses the essence of the new development
model, which is that we no longer do these kinds of feature-based major
milestones. If we want to to deprecate lots of drivers (which I
personally would advocate against -- I have built systems specifically
to run a real floppy drive since the Linux floppy driver is amazingly
flexible and can read/write a lot of formats that nothing else can,
including USB floppies) then we should do that in the normal course of
action, incrementally, and listed in feature-removal-schedule.txt, not
all at once due to some arbitrary milestone.
We have found it works better this way.
-hpa
--
H. Peter Anvin, Intel Open Source Technology Center
I work for Intel. I don't speak on their behalf.
On Tue, May 24, 2011 at 10:36 AM, H. Peter Anvin <[email protected]> wrote:
>
> I think this whole discussion misses the essence of the new development
> model, which is that we no longer do these kinds of feature-based major
> milestones.
Indeed.
It's not about features. It hasn't been about features for forever.
So a renumbering would be purely about dropping the numbers to
something smaller and more easily recognized. The ABI wouldn't change.
The API wouldn't change. There wouldn't be any big "because we finally
did xyz".
Linus
> So I'm toying with 3.0 (and in that case, it really would be "3.0",
> not "3.0.0" - the stable team would get the third digit rather than
> the fourth one.
How about stardates? That'd make a release made now 64860.8
I really should sleep more...
--
Lisa Milne <[email protected]>
On 23.05.2011 13:33, Linus Torvalds wrote:
> On Mon, May 23, 2011 at 12:20 PM, Ingo Molnar <[email protected]> wrote:
> >
> > I really hope there's also a voice that tells you to wait until .42 before
> > cutting 3.0.0! :-)
>
> So I'm toying with 3.0 (and in that case, it really would be "3.0",
> not "3.0.0" - the stable team would get the third digit rather than
> the fourth one.
What about strictly 3 part versions? Just add a .0.
3.0.0 - Release Kernel 3.0
3.0.1 - Stable 1
3.0.2 - Stable 2
3.1.0 - Release Kernel 3.1
3.1.1 - Stable 1
...
Biggest problem is likely version phobics that get pimples when they see
trailing zeros. ;-)
Bis denn
--
Real Programmers consider "what you see is what you get" to be just as
bad a concept in Text Editors as it is in women. No, the Real Programmer
wants a "you asked for it, you got it" text editor -- complicated,
cryptic, powerful, unforgiving, dangerous.
On Tue, May 24, 2011 at 1:41 PM, Linus Torvalds
<[email protected]> wrote:
> On Tue, May 24, 2011 at 10:36 AM, H. Peter Anvin <[email protected]> wrote:
>>
>> I think this whole discussion misses the essence of the new development
>> model, which is that we no longer do these kinds of feature-based major
>> milestones.
>
> Indeed.
>
> It's not about features. It hasn't been about features for forever.
>
> So a renumbering would be purely about dropping the numbers to
> something smaller and more easily recognized. The ABI wouldn't change.
> The API wouldn't change. There wouldn't be any big "because we finally
> did xyz".
>
Me, a nobody end user, would prefer a version number that corresponded
to the date. Something like:
%y.%m.<stable patch>
%Y.%m.<stable patch>
Then users would know the significance of the number and when a vendor
says they support Linux 11.09 the user will immediately know if they
are up to date.
Using the date also clearly communicates it is not about features.
When there is a 3.0 (4.0) release people expect big new features and
API/ABI breakage.
My 2 cents.
On Tue, 24 May 2011, Matthias Schniedermeyer wrote:
> On 23.05.2011 13:33, Linus Torvalds wrote:
>> On Mon, May 23, 2011 at 12:20 PM, Ingo Molnar <[email protected]> wrote:
>>>
>>> I really hope there's also a voice that tells you to wait until .42 before
>>> cutting 3.0.0! :-)
>>
>> So I'm toying with 3.0 (and in that case, it really would be "3.0",
>> not "3.0.0" - the stable team would get the third digit rather than
>> the fourth one.
>
> What about strictly 3 part versions? Just add a .0.
>
> 3.0.0 - Release Kernel 3.0
> 3.0.1 - Stable 1
> 3.0.2 - Stable 2
> 3.1.0 - Release Kernel 3.1
> 3.1.1 - Stable 1
> ...
>
> Biggest problem is likely version phobics that get pimples when they see
> trailing zeros. ;-)
since there are always issues discovered with a new kernel is released
(which is why the -stable kernels exist), being wary of .0 kernels is not
neccessarily a bad thing.
I still think a date based approach would be the best.
since people are worried about not knowing when a final release will
happen, base the date on when the merge window opened or closed (always
known at the time of the first -rc kernel)
in the thread on lwn, people pointed out that the latest 2.6.32 kernel
would still be a 2009.12.X which doesn't reflect the fact that it was
released this month. My suggestion for that is to make the X be the number
of months (or years.months if you don't like large month values) between
the merge window and the release of the -stable release. This would lead
to a small problem when there are multiple -stable releases in a month,
but since that doesn't last very long I don't see a real problem with just
incramenting the month into the future in those cases.
David Lang
Linus Torvalds wrote:
> PS. The voices in my head also tell me that the numbers are getting
> too big. I may just call the thing 2.8.0. And I almost guarantee that
> this PS is going to result in more discussion than the rest, but when
> the voices tell me to do things, I listen.
Correct :)
I would still prefer the version number change to something like 2011.0 -
already proposed at http://kerneltrap.org/Linux/Kernel_Release_Numbering_Redux
I don't think that it is reasonable to say that it is bad because third party
scripts would break - they would break anyway (I would bet that many of them
don't expect to see 3.x anyway). And changing now to 3.0 and then incrementing
the second one everytime for 10 years will also lead to something like 3.56.7.
I would also say that defining the release number using the time of the merge
window start/end is easy understandable. "2.6.40" would be the third
development cycle this year aka v2011.2 or v2011.2.0 when the patchlevel
should always be included.
--
Emil Langrock
Hi,
2011/5/24 Lisa Milne <[email protected]>:
>> So I'm toying with 3.0 (and in that case, it really would be "3.0",
>> not "3.0.0" - the stable team would get the third digit rather than
>> the fourth one.
>
> How about stardates?
This is a wonderful idea! :)
> That'd make a release made now 64860.8
>
> I really should sleep more...
>
> --
> Lisa Milne <[email protected]>
> --
> 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/
>
--
Slawa!
Zimny Lech z Wawelu
On Tuesday 2011-05-24 20:48, eschvoca wrote:
>On Tue, May 24, 2011 at 1:41 PM, Linus Torvalds wrote:
>>
>> It's not about features. It hasn't been about features for forever.
>
>Using the date also clearly communicates it is not about features.
On the contrary: Whenever a 2.6.x release was set out the door, there
was at least one new feature in the cycle. Given the endless manpower
that seems to trail Linux, that seems unlikely to change in the near
term.
On "It is not about features" - it is not /just/ about features - it
is also about fixes, which are equally important, and they also
warrant a version bump of some sort on their own. Pointing out the
obvious, the stable serieses.
"Fleeing" to date-based version numbering seems like an excuse
for making way for releases without any change whatsoever.
(IOW, if there were features/fixes, a non-date based scheme of
incremental numbers could indicate them already.)
>Me, a nobody end user, would prefer a version number that corresponded
>to the date. Something like:
>
>%y.%m.<stable patch>
>%Y.%m.<stable patch>
Except that LINUX_KERNEL_VERSION has only 8 bits for each,
so 2011 is out of range, which would need to kept in mind.
And a 2-digit spec.. nah that does not feel very 100-year
proof. (Just look at struct tm.tm_year for the mess 2-digits
made technically.)
>Then users would know the significance of the number and when a vendor
>says they support Linux 11.09 the user will immediately know if they
>are up to date.
An added issue with that would be that numbers would not increase in
same-sized steps anymore and leave gaps. (There would probably be no
11.06 between 11.05 aka 2.6.39 and 11.07-or-so aka 2.6.40)
My 2円.
On 05/23/2011 04:33 PM, Linus Torvalds wrote:
> On Mon, May 23, 2011 at 12:20 PM, Ingo Molnar<[email protected]> wrote:
>>
>> I really hope there's also a voice that tells you to wait until .42 before
>> cutting 3.0.0! :-)
>
> So I'm toying with 3.0 (and in that case, it really would be "3.0",
> not "3.0.0" - the stable team would get the third digit rather than
> the fourth one.
>
> But no, it wouldn't be for 42. Despite THHGTTG, I think "40" is a
> fairly nice round number.
>
> There's also the timing issue - since we no longer do version numbers
> based on features, but based on time, just saying "we're about to
> start the third decade" works as well as any other excuse.
>
I don't think year-based versions (like 2011.0 for the first 2011
release, or maybe 2011.5 for May 2011) are pretty, but I'll make an
argument for them anyway: it makes it easier to figure out when hardware
ought to be supported.
So if I buy a 2014-model laptop and the coffee-making button doesn't
work, and my favorite distro is running the 2013 kernel, then I know I
shouldn't expect to it to work. (Graphics drivers are probably a more
realistic example.)
Also, when someone in my lab installs <insert ancient enterprise distro
here> on a box that's running software I wrote that needs to support
modern high-speed peripherals, then I can say "What? You seriously
expect this stuff to work on Linux 2007? Let's install a slightly less
stable distro from at least 2010." This sounds a lot less nerdy than
"What? You seriously expect this stuff to work on Linux 2.6.27? Let's
install a slightly less stable distro that uses at least 2.6.36."
--Andy
On Monday 23 May 2011, 22:33:48 Linus Torvalds wrote:
> On Mon, May 23, 2011 at 12:20 PM, Ingo Molnar <[email protected]> wrote:
> > I really hope there's also a voice that tells you to wait until .42
> > before cutting 3.0.0! :-)
>
> So I'm toying with 3.0 (and in that case, it really would be "3.0",
> not "3.0.0" - the stable team would get the third digit rather than
> the fourth one.
>
> But no, it wouldn't be for 42. Despite THHGTTG, I think "40" is a
> fairly nice round number.
>
> There's also the timing issue - since we no longer do version numbers
> based on features, but based on time, just saying "we're about to
> start the third decade" works as well as any other excuse.
But hey, do you really want to release a Linux 3.0 kernel without
serious layered filesystem functionality?
Shame on you,
Pete
PS.: Sorry for being such a pest in this regard, but filesystem layering
is one of the most important missing bits to excel out of the box in
* live distros
* diskless computing
* flash based systems
Even the linux based commercial PBX solution (mobydick), I bought, ships
with it.
PPS.: Bad timing, I know, but I'm glad, that Al is back to life again..
On Tue, 24 May 2011 14:30:59 +0200, Jacek Luczak said:
> 2011/5/24 Jan Engelhardt <[email protected]>:
> > On Tuesday 2011-05-24 01:33, Linus Torvalds wrote:
> >
> >>Another advantage of switching numbering models (ie 3.0 instead of
> >>2.8.x) would be that it would also make the "odd numbers are also
> >>numbers" transition much more natural.
> >>
> >>Because of our historical even/odd model, I wouldn't do a 2.7.x -
> >>there's just too much history of 2.1, 2.3, 2.5 being development
> >>trees.
> >
> > .oO(Though once 2.{7 or more, odd} trickle into the distros, it would
> > become pretty much apparent that they are not devel.)
> >
> >>And then in another few years (probably before getting close to 3.40,
> >>so I'm not going to make a big deal of 3 = "third decade"), I'd just
> >>do 4.0 etc.
> >
> > While 2.6 has certainly worn out, already thinking of a 4.0 is highly
> > reminiscient of the version number arms race Firefox and ChromeBrowser
> > are doing currently.
> >
> >>Because all our releases are supposed to be stable releases these
> >>days, and if we get rid of one level of numbering, I feel perfectly
> >>fine with getting rid of the even/odd history too.
> >
> > If I remember past-time discussions right, ELF was the contributing
> > factor to bump the major number to 2.0 back then; ever since 2.0, no
> > similarly breakthrough-ing event has occurred.
>
> What then about BKL removal? Nice place to celebrate with version jump
> and heaving some beers.
Well, if we're looking at ELF-sized ABI changes, how about 3.0 be the
release where we re-sync the syscall numbers on all the archs? ;)
Hi,
On Mon, May 23, 2011 at 12:13:29PM -0700, Linus Torvalds wrote:
> PS. The voices in my head also tell me that the numbers are getting
> too big. I may just call the thing 2.8.0. And I almost guarantee that
> this PS is going to result in more discussion than the rest, but when
> the voices tell me to do things, I listen.
With all the discussion about matching the year, keeping the numbers smallish and breaking build/test scripts...
Why not match the year this way: v2.11 = 2011
So, to get there, how about:
v2.6.39 current
v2.6.40
v2.7.? (see below)
v2.11.0 sometime this year maybe or skipped.
v2.12.? early 2012.
v2.13.? early 2013.
and so on...
To hell with making the months match up to anything. IMHO, that's micromanaging BS that will cause recurring troubles forever.
Advantages:
-Build & test scripts won't break for a while. (see note about this below)
-Major and Minor numbers together approximately match the year of release.
-Subreleases don't get too big. e.g. with 2 week releases, v2.12.x might reach .26 or .27 and resets with v2.13.0
Disadvantages:
-in the year 2100 the version will have to leap to v21.0 and that numbering will last until v29.99 or so.
-most of us won't live long enough to see v31.14 or v42.0
The build scripts would finally break in the year 25500 with the step after kernel v254.99.?. I suspect one of our decendants in the next 23.5 thousand years can figure out how to fix that.
IMHO, this may be a least effort compromise between the keep-numbers-small camp, the match-the-year camp and the don't-break-scripts camp.
Of course, math junkies like me would love to see v2.7.18.28 at some point in the process. Could we jump straight to that just before we leap to v2.11.0 or whatever numbering is chosen please?
-phil
On Tuesday 24 May 2011 23:05:30 Jan Engelhardt wrote:
> On Tuesday 2011-05-24 20:48, eschvoca wrote:
> >On Tue, May 24, 2011 at 1:41 PM, Linus Torvalds wrote:
> >> It's not about features. It hasn't been about features for forever.
> >
> >Using the date also clearly communicates it is not about features.
>
> On the contrary: Whenever a 2.6.x release was set out the door, there
> was at least one new feature in the cycle. Given the endless manpower
> that seems to trail Linux, that seems unlikely to change in the near
> term.
>
> On "It is not about features" - it is not /just/ about features - it
> is also about fixes, which are equally important, and they also
> warrant a version bump of some sort on their own. Pointing out the
> obvious, the stable serieses.
You are mixing up features based versioning and identifier for versions. Linux
has no feature based concept for most parts of their version number (only the
patch part clearly states: fixes, fixes, fixes). We only need something that
is easily readable and maybe has no extreme weird meaning that leads to false
conclusions. And yes, that is what eschvoca meant and not something like
"linux is stagnating".
> "Fleeing" to date-based version numbering seems like an excuse
> for making way for releases without any change whatsoever.
> (IOW, if there were features/fixes, a non-date based scheme of
> incremental numbers could indicate them already.)
Yes, that is usally the case... release the same source tarball again and
again. I always had that feeling when looking at those wine, ubuntu, gentoo,
ms, texlive, iasl, hugin, u-boot, ... developers. They are doing nothing the
whole day and the marketing department does everything.
> >Me, a nobody end user, would prefer a version number that corresponded
> >to the date. Something like:
> >
> >%y.%m.<stable patch>
> >%Y.%m.<stable patch>
>
> Except that LINUX_KERNEL_VERSION has only 8 bits for each,
> so 2011 is out of range, which would need to kept in mind.
> And a 2-digit spec.. nah that does not feel very 100-year
> proof. (Just look at struct tm.tm_year for the mess 2-digits
> made technically.)
What is LINUX_KERNEL_VERSION? I only know LINUX_VERSION_CODE and
KERNEL_VERSION
And the calculation behind it is following:
(((a) << 16) + ((b) << 8) + (c))
So for KERNEL_VERSION(2,6,40) we would get 0x20628 and for
KERNEL_VERSION(2011,2,0) we would get 0x07DB0200. Of course our
grandgrandgrand...grand children would die in agony in the year 65536.
And maybe (probably the module version check guys will kill me) could use a
compressed version of that without hurding the comparison function in out of
kernel modules. KERNEL_VERSION_Y(a,b) would be defined as
#define KERNEL_VERSION_Y(a,b) ({typeof (a) _a = a; \
typeof (b) _b = b; \
KERNEL_VERSION(_a >> 8, _a & 0xff, _b); })
This would bring us to the year 16777216 before everybody gets punched in his
private parts by the versioning scheme. It is also possible to get more out of
32 bits when we can assume that Linus or his grandgrand...grand children will
never do more than 128 releases a year.
But yes, I aggree not to use 2 digit numbers for years.... unless we want to
start the y2k+100 problem here.
> >Then users would know the significance of the number and when a vendor
> >says they support Linux 11.09 the user will immediately know if they
> >are up to date.
>
> An added issue with that would be that numbers would not increase in
> same-sized steps anymore and leave gaps. (There would probably be no
> 11.06 between 11.05 aka 2.6.39 and 11.07-or-so aka 2.6.40)
Ok, this is really a good example why we should not use the month for
releases, but an ever increasing number until the first number is also
increased which resets the second number to 0.
Kind regards,
Emil
On 24/05/11 21.13, [email protected] wrote:
> Well, if we're looking at ELF-sized ABI changes, how about 3.0 be the
> release where we re-sync the syscall numbers on all the archs? ;)
If you want to do that I think the best way to do it is to
have both the old and the new numbers co-exist through the
3.x series and the old ones go away in 4.0.
You'd first have to find the highest number currently
assigned and then add a bit of safety margin to decide on
a starting point for the new numbers.
Should architecture dependent system calls be assigned
from a separate interval where they could overlap between
architectures? Or should they be assigned from the same
sequence as other calls and return -ENOSYS on other
architectures than the one they were targeted for?
Or was it all a joke, and you don't actually want that
cleanup to happen because of too much breakage?
--
Kasper Dupont -- Rigtige m?nd skriver deres egne backupprogrammer
#define _(_)"d.%.4s%."_"2s" /* This is my email address */
char*_="@2kaspner"_()"%03"_("4s%.")"t\n";printf(_+11,_+6,_,11,_+2,_+7,_+6);
On Tue, 24 May 2011, Andy Lutomirski wrote:
> Also, when someone in my lab installs <insert ancient enterprise distro
> here> on a box that's running software I wrote that needs to support
> modern high-speed peripherals, then I can say "What? You seriously
> expect this stuff to work on Linux 2007? Let's install a slightly less
> stable distro from at least 2010." This sounds a lot less nerdy than
> "What? You seriously expect this stuff to work on Linux 2.6.27? Let's
> install a slightly less stable distro that uses at least 2.6.36."
I hate to jump into this excellent example of bike-shedding discussion,
but anyway ...
Your example doesn't really reflect reality.
It's common for older enterprise distributions to gradually incorporate a
lot of backported code (and most importantly new hardware support
code/drivers) while not upgrading the kernel major version. So yes, you
will in reality get 2.6.16 kernel (at least according to uname) with
libata with newer service packs of SLES 10, for example (and you could
find many of those across distributions).
--
Jiri Kosina
SUSE Labs
On 05/23/2011 11:52 PM, Alexey Zaytsev wrote:
> On Tue, May 24, 2011 at 00:33, Linus Torvalds
> <[email protected]> wrote:
>> On Mon, May 23, 2011 at 12:20 PM, Ingo Molnar <[email protected]> wrote:
>>>
>>> I really hope there's also a voice that tells you to wait until .42 before
>>> cutting 3.0.0! :-)
>>
>> So I'm toying with 3.0 (and in that case, it really would be "3.0",
>> not "3.0.0" - the stable team would get the third digit rather than
>> the fourth one.
>>
>> But no, it wouldn't be for 42. Despite THHGTTG, I think "40" is a
>> fairly nice round number.
>>
>> There's also the timing issue - since we no longer do version numbers
>> based on features, but based on time, just saying "we're about to
>> start the third decade" works as well as any other excuse.
>>
>> But we'll see.
>
> Maybe, 2011.x, or 11.x, x increasing for every merge window started this year?
> This would better reflect the steady nature of the releases, but would
> certainly break a lot of scripts. ;)
My $0.017 on this. Clearly current process is time based. People have said.
* Keep Three digit numbers to retain script compatibility
* Make it clear from the version when it was released.
* Linus said 3 as for 3rd decade
* Nice single increment number progression
* Please make it look like a nice version number sys-admins will feel
at home with
So if you combine all the above:
D. Y. N
D - Is the decade since birth (1991 not 1990)
Y - is the year in the decade so you have 3.1.x, 3.2.x, .. 3.10.x, 4.1.X and so on
Nice incremental number.
N - The Linus release of this Year. So this 3rd one goes up to 4 most probably.
Linus always likes, and feels very poetic about the Christmas version release.
He hates it when once it slipped into the next year. So now he gets to increment
the second digit as a bonus.
The 2nd digit gets to start on a *one*, never zero and goes up to *10*, to symbolize
the 1991 birth. And we never have .zero quality, right?
The first Digit gets incremented on decade from 1991 so on 2011 and not 2010
So here you have it, who said we need to compromise?
Free life
Boaz
On Tue, 24 May 2011, Zimny Lech wrote:
> Hi,
>
> 2011/5/24 Lisa Milne <[email protected]>:
>>> So I'm toying with 3.0 (and in that case, it really would be "3.0",
>>> not "3.0.0" - the stable team would get the third digit rather than
>>> the fourth one.
>>
>> How about stardates?
>
> This is a wonderful idea! :)
I'd rather go for a gardensheduler, which can tell me the optimal color
for any given moment *and* do the paint job. If it eventually ends this
discussion, it could be renamed "completely fair gardensheduler".
>
>> That'd make a release made now 64860.8
>>
>> I really should sleep more...
Or drink less coffee ... ;)
--
Regards,
Martin
On Wed, 25 May 2011 14:26:18 +0200, Kasper Dupont said:
> On 24/05/11 21.13, [email protected] wrote:
> > Well, if we're looking at ELF-sized ABI changes, how about 3.0 be the
> > release where we re-sync the syscall numbers on all the archs? ;)
>
> If you want to do that I think the best way to do it is to
> have both the old and the new numbers co-exist through the
> 3.x series and the old ones go away in 4.0.
>
> You'd first have to find the highest number currently
> assigned and then add a bit of safety margin to decide on
> a starting point for the new numbers.
Most archs are sitting around 300-340. Alpha is the apparent
winner with 501. So start at 512 which is a nice round number.
> Or was it all a joke, and you don't actually want that
> cleanup to happen because of too much breakage?
Well, with 2.0, we had some ABI cratering due to ELF. I figured
*if* we're willing to do ABI cratering for a 3.0, cleaning up the
syscalls would be a nice drastic way to start. ;)
(Though to be honest, I think moving the a copy syscalls into one consistent
table starting at 512 or so for 3.0, and then nuking the low numbers for 4.0
would be better. We would however have to take into account architectures
that have limited number of syscalls available - any archs unable to handle
syscall numbers up to 1024?)
On Wed, May 25, 2011 at 7:12 AM, Boaz Harrosh <[email protected]> wrote:
> So if you combine all the above:
>
> D. Y. N
> D - Is the decade since birth (1991 not 1990)
> Y - is the year in the decade so you have 3.1.x, 3.2.x, .. 3.10.x, 4.1.X and so on
> ? ?Nice incremental number.
> N - The Linus release of this Year. So this 3rd one goes up to 4 most probably.
>
> Linus always likes, and feels very poetic about the Christmas version release.
> He hates it when once it slipped into the next year. So now he gets to increment
> the second digit as a bonus.
>
> The 2nd digit gets to start on a *one*, never zero and goes up to *10*, to symbolize
> the 1991 birth. And we never have .zero quality, right?
>
> The first Digit gets incremented on decade from 1991 so on 2011 and not 2010
This is clearly the best suggestion so far - small numbers, somewhat
date related (but without stuffing a "2011." on the front). No ".0"
releases, ever.
But best of all it defines now when we will switch to 4.x.y and 5.x.y
so we don't have to keep having this discussion whenever someone thinks
that the numbers are getting "too big" (well perhaps when we get to the
tenth decade or so :-)
So the only thing left to argue is whether the upcoming release should
be numbered "3.1.1" as the first release in the first year of the 3rd
decade ... or whether we should count 2.6.37 .. 2.6.39 as the first
three releases this year and thus we ought to start with "3.1.4" (so we
start with "pi"!).
Linus: If you go with this, you should let Boaz set the new "NAME"
as a prize for such an inspired solution.
-Tony
On Tue, 2011-05-24 at 11:07 -0400, [email protected] wrote:
> On Tue, May 24, 2011 at 10:43 AM, Alan Cox <[email protected]>
> wrote:
> > Can we drop most of MCA, EISA and ISA bus if we are going to have a
> big
> > version change ? A driver spring clean is much overdue and it's all
> in
> > git in case someone wishes to sneak out at midnight and bring some
> crawly
> > horror back from the dead.
>
> 2.8 could mark the beginning of the great cleanup
> --- work out the details of what needs to be cleaned and set a goal
> --- remove old buses/driver, switch to device tree, graphics, 32/64
> merges, etc
> 3.0 would mark its completion
Here it go my opinion,
Many people ask for beginning of 2.7 kernel series which will end on
2.8, by old numeration.
Kernel 2.8 will mainly a major clean up, of support of the very old
hardware, like "math co-processor" at only exist in 386 and before
Pentium. If some one want put Linux on this very old hardware should use
kernel 2.2.
However a new numeration of kernel is independent of this, and I agree
with new numeration of kernel on drop a number.
Last but not least, I would like to see marked a hiper stable kernel ,
which will be used by Debian guys. Debian guys tend to stop in a kernel
which is not the best one, so let we choose for them what is the stable
of stables .
Best regards,
--
Sérgio M. B.
On 05/26/2011 01:21 AM, Tony Luck wrote:
> On Wed, May 25, 2011 at 7:12 AM, Boaz Harrosh <[email protected]> wrote:
>> So if you combine all the above:
>>
>> D. Y. N
>> D - Is the decade since birth (1991 not 1990)
>> Y - is the year in the decade so you have 3.1.x, 3.2.x, .. 3.10.x, 4.1.X and so on
>> Nice incremental number.
>> N - The Linus release of this Year. So this 3rd one goes up to 4 most probably.
>>
>> Linus always likes, and feels very poetic about the Christmas version release.
>> He hates it when once it slipped into the next year. So now he gets to increment
>> the second digit as a bonus.
>>
>> The 2nd digit gets to start on a *one*, never zero and goes up to *10*, to symbolize
>> the 1991 birth. And we never have .zero quality, right?
>>
>> The first Digit gets incremented on decade from 1991 so on 2011 and not 2010
>
> This is clearly the best suggestion so far - small numbers, somewhat
> date related (but without stuffing a "2011." on the front). No ".0"
> releases, ever.
>
> But best of all it defines now when we will switch to 4.x.y and 5.x.y
> so we don't have to keep having this discussion whenever someone thinks
> that the numbers are getting "too big" (well perhaps when we get to the
> tenth decade or so :-)
>
> So the only thing left to argue is whether the upcoming release should
> be numbered "3.1.1" as the first release in the first year of the 3rd
> decade ... or whether we should count 2.6.37 .. 2.6.39 as the first
> three releases this year and thus we ought to start with "3.1.4" (so we
> start with "pi"!).
>
Yes, Yes I like this a lot. I love pi, thanks.
Boaz
> Linus: If you go with this, you should let Boaz set the new "NAME"
> as a prize for such an inspired solution.
>
> -Tony
Many interesting ideas on version numbering schemes. I like 2.11.X
because it maps to years easily in people's mind, but I look forward
to seeing what is chosen. You guys break many of the rules for
software development, so why not going backwards in version numbers
;-)
While you are talking about arbitrary numbers and new goals, I want to
offer that you could consider a push towards zero bugs. In general, as
long as your reliability monotonically increases (no regressions) that
is an acceptable minimum approach because it means that you will never
have a customer go from being happy to unhappy.
However, it is common in companies to make an effort to get towards
zero bugs. Zero bugs is impossible, and that is a philosophical
discussion. If you look through your current list of bugs, nearly
every one looks scary to me and important to someone. You currently
have 2,800 active bugs (http://bit.ly/LinuxBugs) The last time I
looked, I found the median age was 10 months. In general, bugs should
be fixed in the next release and so therefore 3 months.
Zero bug bounces is hard for the others because they don't have
sufficient resources. However, I believe you easily do. I can't say
that anything magical technically will happen if you work on your bugs
faster, but I can say that people I respect as much as you taught me
this. My salary was based on my ability to promptly respond to my
bugs, and zero was everyone's goal. Hitting zero, even for a minute,
could be a newsworthy event, as another way Linux is better than the
others. It also shows leadership to user mode. I sometimes get the
feeling that many in the FOSS community look at bugs as something they
could work on when they get bored of adding new features, instead of:
"Holy poop, there is someone unhappy out there."
Warm regards,
-Keith
http://keithcu.com/
On Thu, 26 May 2011 22:44:24 PDT, Keith Curtis said:
> However, it is common in companies to make an effort to get towards
> zero bugs. Zero bugs is impossible, and that is a philosophical
> discussion. If you look through your current list of bugs, nearly
> every one looks scary to me and important to someone. You currently
> have 2,800 active bugs (http://bit.ly/LinuxBugs) The last time I
> looked, I found the median age was 10 months. In general, bugs should
> be fixed in the next release and so therefore 3 months.
You may want to look at what percentage of those bugs were reported on
one hardware platform by one reporter, and the reporter has since evaporated.
I myself started a thread back on April 26 regarding a wonky PS2->USB adapter.
Within 24 hours I had a bunch of good suggestions for further debugging, none
of which I've had a chance to actually follow up on (Hmm.. Monday is a holiday
but nobody will be in the office, maybe I'll have a chance to get in the 4-5
reboots it will take.. :) So if I had opened a bug, how old is it, and who's fault
is it that it's that old?
> Hitting zero, even for a minute, could be a newsworthy event, as another way
> Linux is better than the others. It also shows leadership to user mode.
Never happen, as at least one of those 2,800 bugs will involve testing a fix on
hardware the reporter no longer has, or the reporter is no longer available, or
similar issues.
You also need to look at the *severity* of the bugs - my USB issue merely
causes me literally 5 second's inconvenience every morning (part of why I
haven't chased it further - it's hard to justify spending 45 minutes fixing a
5-second issue). If the reporter can't be bothered to help, what are we
supposed to do? Then there's the oops and panic reports that should count for
a lot more. How severe are most of those 2,800 bugs?
Probably a lot more productive than "zero bugs" would be "swat the top 10
entries in the kerneloops database", as we know by measurement that they're
both pervasive and high-impact.
On Fri, 27 May 2011 10:29:32 -0400
[email protected] wrote:
> On Thu, 26 May 2011 22:44:24 PDT, Keith Curtis said:
>
> > However, it is common in companies to make an effort to get towards
> > zero bugs. Zero bugs is impossible, and that is a philosophical
> > discussion. If you look through your current list of bugs, nearly
> > every one looks scary to me and important to someone. You currently
> > have 2,800 active bugs (http://bit.ly/LinuxBugs) The last time I
> > looked, I found the median age was 10 months. In general, bugs should
> > be fixed in the next release and so therefore 3 months.
>
Besides the valid points Valdis gave, also many of the bugs (if they
are severe/real world bugs) are already fixed. That is because the
bugzilla is not the only place where bugs are reported and resolved and
only the original submitter and some other people are able to close bugs.
Most of the development happens on the mailinglist. So this
is also where most of the bugs get solved.
( Mailinglists also have the advantage of automatically garbage
collecting old bug reports nobody has any interest in solving. This is
a big drawback with bugzilla. If there is too much garbage in the
bugzilla because nobody closes bugs, then developers will not take old
bugzilla reports seriously. )
So if you want to make the bugzilla more useful, just go for it and put
relevant people on the cc list of ignored bug reports, ping people that have not
answered to a specific question for months (the bugzilla looses
notification mails sometimes).. And if you determine bugs that can be
closed, drop me a line, I will be happy to do that, if it is
appropriate.
Regards,
Flo
Hello;
I wrote two responses, which eventually allowed me to boil it down to
one sentence: You can do anything you want in addition to perfection
(i.e. zero bugs).
Long answer:
I was baptized into the zero bugs religion about 20 years ago. This
was before the web and time-based releases, but these only add
complications, and are yet compatible with zero bugs all the time. As
a beginner, you can cheat such as zero bugs older than 6 months / 2
releases, or give yourself 1,000. Over time, you can pick new goals
and improve your numbers. Reaching enlightenment may require multiple
steps. You don't have to meet your goals, but you will become better
for trying.
Like other religions, zero bugs requires faith and diligence, but
provides benefits at the end. In this religion, members spend regular
time reading their bible / bug list. They find it a useful source of
knowledge on how they can make themselves and the world better today.
It also allowed them to figure out when they had too much work or
needed help, and when they were done with their duties and could rest.
Many will tell you that the more time they spend with their bible, the
better it becomes. It is also a source of Numbers that the wise men
would scratch their heads over.
I have been watching your bug count for years and I realize now that
you have the benefits of the religion, but not the faith and habits.
When zero becomes an additional goal, the codebase will get better
faster, and eventually the people happier. You may want better tools
or release managers. It can also involve seeking a priest who can
answer any questions of faith. I am not even ordained, but there are
others. When your belief is tested, it can help to think of Knuth,
Boeing, Debian, etc. It may be efficient to perform a mass exorcism /
conversion. Please repent you amazing heathens.
Zero bugs is like biking: I wouldn't try if weren't a goal, and I
didn't believe it would be beneficial. I would have gone biking today
but I had to get drunk to have the courage to send this. Mission
accomplished & go Barcelona FC.
I apologize if my mails come off as 0bvious rants, etc. I am in awe of
what already exists, so part of me wants to be quiet rather than risk
pissing off people that I write about and who inspire me. However,
I've decided to share my work also, to speed the fun along. You can
thank me later ;-)
So, cheers. Happy to be here for the start of Linux 3.
Kind regards,
-Keith
http://keithcu.com/