Aug 04 2006
21:44 GMT
21:44 GMT
Why kernel 2.6.19 will break most distros. Or not. misc
Years ago, there was an old way parents were thinking about their kids in my country: «I have made them, I will kill them too.» Does the fact that you created something give you the right to be a dictator?
This is not the first time when I write about Greg KH. As one of the most important kernel developers (especially in the drivers area), he deserves all the appreciation he can get, along with our gratitude, as simple users of Linux.
Sometimes however, the eagerness of Greg Kroah-Hartman to make things move forward can be dangerous and counterproductive to (a large part of) the community.
Thanks to a LWN article, New kernels and old distributions, I could find out how Greg KH manages to be quite arrogant even with other respected developers, say Andrew Morton for one.
Facts:
If distros that are 10 months older will be broken by a new kernel, this is going to be a lame situation for Linux (maybe it's time to get used with FreeBSD/NetBSD/OpenBSD?).
The next message from Greg KH is ethically unacceptable: «What do you propose we do then? Wait a year? Two years? I've already stated that I'm going to wait 3 more months in order to be able to fix up the currently supported distros to handle the changes needed. ... No, but it does show that you are running on an unsupported distro. Why should the kernel developers have to support this? ... How long should the community have to care about a distro after the creators of it have abandoned it?»
"Abandoned" is a tough word. And the imperial atitude of "I'll leave you 3 months before you die" is not the right one.
There is an exaggeration in Andrew's opinion too, though: «Greg, once we put an interface into the kernel and expose it to userspace, that's it - we support it forever. At least, that's the model.» Greg KH already said at Linux Symposium 2006 that... the kernel doesn't need a stable API. This is not acceptable for many people (including myself), but this is the way Greg thinks.
For the current change that might break everything older that 10 months... we should be grateful to Alan Cox: «Greg, this changeset should get ripped out and thrown away. It's wrong. It's an API breakage and it is going to catch a lot of people. ... you said you could use symlinks and the like to work around it, or fix the code. How about proving it or doing the work to make it possible rather than continuing to break everything repeatedly.»
So far, it seems that people paid by Red Hat (Alan Cox) are more responsible (and less rushed to go forward frantically at any cost) than people paid by Novell (Greg Kroah-Hartman). Go figure.
P.S.: For those of you thinking that, as long as GNU/Linux is (usually) free (if you don't happen to pay for the support of a corporate distro), you must upgrade at least yearly, think again: if someone is running Windows XP, and an update brought by Windows Update breaks the system, is he entitled to complain? (Of course he is!) If you are running a GNU/Linux distro, why should you expect for less?!
[permalink]
This is not the first time when I write about Greg KH. As one of the most important kernel developers (especially in the drivers area), he deserves all the appreciation he can get, along with our gratitude, as simple users of Linux.
Sometimes however, the eagerness of Greg Kroah-Hartman to make things move forward can be dangerous and counterproductive to (a large part of) the community.
Thanks to a LWN article, New kernels and old distributions, I could find out how Greg KH manages to be quite arrogant even with other respected developers, say Andrew Morton for one.
Facts:
- Kubuntu's udev was reported to be broken by 2.6.18-rc2-mm1.
- Andrew Morton also complained that FC3 breaks (yes, you might be using FC5, but there are many servers still running FC3): «This stuff breaks my FC3 test box and there is, afaict, no clear way for users to upgrade udev to unbreak it. As a developer I could of course bang on things until it works, but that's not the point. The point is that these patches break Linux on a major release from a major vendor only two years after its release. That's not a minor problem, is it?»
- Greg KH, who was behind the changes, replied rather contemptuously: «That distro is unsupported now, right? How long do you expect the kernel to support unsupported, community based distros that thrive on the fact that they are quickly updated? ... I will revert the patch in mainline that causes people to have to upgrade to a udev that is in FC5, and wait till the next release for that to happen (the minimum will be 081, which was released in January, 2006, by the time 2.6.19 is out, that will be about 10 months old.) ... How long should the kernel be forced to lag behind?»
- Andrew Morton points again to the real situation: «Please stop going on about FC3. My (repeat) point is that we're proposing to break _all_ distros which are older than ten months. We don't play the "oh, that isn't supported any more" game.»
If distros that are 10 months older will be broken by a new kernel, this is going to be a lame situation for Linux (maybe it's time to get used with FreeBSD/NetBSD/OpenBSD?).
The next message from Greg KH is ethically unacceptable: «What do you propose we do then? Wait a year? Two years? I've already stated that I'm going to wait 3 more months in order to be able to fix up the currently supported distros to handle the changes needed. ... No, but it does show that you are running on an unsupported distro. Why should the kernel developers have to support this? ... How long should the community have to care about a distro after the creators of it have abandoned it?»
"Abandoned" is a tough word. And the imperial atitude of "I'll leave you 3 months before you die" is not the right one.
There is an exaggeration in Andrew's opinion too, though: «Greg, once we put an interface into the kernel and expose it to userspace, that's it - we support it forever. At least, that's the model.» Greg KH already said at Linux Symposium 2006 that... the kernel doesn't need a stable API. This is not acceptable for many people (including myself), but this is the way Greg thinks.
For the current change that might break everything older that 10 months... we should be grateful to Alan Cox: «Greg, this changeset should get ripped out and thrown away. It's wrong. It's an API breakage and it is going to catch a lot of people. ... you said you could use symlinks and the like to work around it, or fix the code. How about proving it or doing the work to make it possible rather than continuing to break everything repeatedly.»
So far, it seems that people paid by Red Hat (Alan Cox) are more responsible (and less rushed to go forward frantically at any cost) than people paid by Novell (Greg Kroah-Hartman). Go figure.
P.S.: For those of you thinking that, as long as GNU/Linux is (usually) free (if you don't happen to pay for the support of a corporate distro), you must upgrade at least yearly, think again: if someone is running Windows XP, and an update brought by Windows Update breaks the system, is he entitled to complain? (Of course he is!) If you are running a GNU/Linux distro, why should you expect for less?!
[permalink]
Patrick
@ Aug 04, 2006 (06:47 PM)
Béranger
@ Aug 04, 2006 (07:00 PM)
Thanks for the comment, Patrick! You are one of the rare readers that seem to agree with me :)
I might try some day NetBSD, as when I tried it... in Jan. 1995 I guess (it was at 1.0!!!), it seemed a good *BSD. If you try it before I do, please drop me a line with impressions. Thx.
I might try some day NetBSD, as when I tried it... in Jan. 1995 I guess (it was at 1.0!!!), it seemed a good *BSD. If you try it before I do, please drop me a line with impressions. Thx.
Patrick
@ Aug 04, 2006 (07:41 PM)
Have you ever heard of Blag, the GNU/Linux distribution based on FC? Anyways, the maintainer of the distribution has a lot of good things to say about Open BSD, it is really cleanly coded and such. For me, however, I would try Desktop BSD, as it does not try to mimick the Windows approach to installing programs (PBI's and PC-BSD).
As for mono and Gnome, I have switched my system to Kanotix and updated to the latest changes (KDE 3.5.4), and then changed the repositories to Etch, which will eventually catch up to my system. No matter how hard I try, I just can't get away from Debian. I'd like to see how KDE works, and 4.0 seems to really be promising (especially the reduced hardware requirements for performace). For all the talk about Gnome being "dumbed down", I really had to search to find out how to do some really basic things in KDE. I changed the the task bar and panel to mimick Gnome, and I can't stand how a single click is used instead of the double click. The more I use it, the more comfortable I become with it. The biggest selling point for me are K3B and Amarok. Gnome based programs have yet to match the ease of use, but Gnome just "feels" better than KDE. I know I can run QT applications under Gnome, but the idea of installing all those QT libraries under Gnome is just redundant and, quite frankly, stupid.
As for mono and Gnome, I have switched my system to Kanotix and updated to the latest changes (KDE 3.5.4), and then changed the repositories to Etch, which will eventually catch up to my system. No matter how hard I try, I just can't get away from Debian. I'd like to see how KDE works, and 4.0 seems to really be promising (especially the reduced hardware requirements for performace). For all the talk about Gnome being "dumbed down", I really had to search to find out how to do some really basic things in KDE. I changed the the task bar and panel to mimick Gnome, and I can't stand how a single click is used instead of the double click. The more I use it, the more comfortable I become with it. The biggest selling point for me are K3B and Amarok. Gnome based programs have yet to match the ease of use, but Gnome just "feels" better than KDE. I know I can run QT applications under Gnome, but the idea of installing all those QT libraries under Gnome is just redundant and, quite frankly, stupid.
Patrick
@ Aug 04, 2006 (08:12 PM)
http://www.desktoplinux.com/news/NS7778908329.html
The above link shows that Lenovo has now entered an agreement with Suse to offer SLED 10 on a Thinkpad notebook (I forget which one, but the article states the model). For all those people who think that OEM has kept GNU/Linux from being adopted by the masses, this is now your chance to be proven wrong or right.
Regardless of what you think of Suse, this is a big step for GNU/Linux. If sells don't increase, then GNU/Linux really needs to rethink how to break into the mainstream.
The above link shows that Lenovo has now entered an agreement with Suse to offer SLED 10 on a Thinkpad notebook (I forget which one, but the article states the model). For all those people who think that OEM has kept GNU/Linux from being adopted by the masses, this is now your chance to be proven wrong or right.
Regardless of what you think of Suse, this is a big step for GNU/Linux. If sells don't increase, then GNU/Linux really needs to rethink how to break into the mainstream.
Anonymous Coward
@ Aug 05, 2006 (10:06 AM)
What Greg does is perfectly acceptable. Distros are constant maintenance work, they're not a one-off effort. That is why not everyone should try to put together his/her own distro. If someone isn't prepared to properly maintain a distro then he/she should concentrate on other work.
The Lenovo/Novell deal really is a BIG thing. One of Lenovo's best hardware offerings coupled with the as-to-date most usable Linux desktop. Thumbs up.
The Lenovo/Novell deal really is a BIG thing. One of Lenovo's best hardware offerings coupled with the as-to-date most usable Linux desktop. Thumbs up.
Béranger
@ Aug 05, 2006 (10:26 AM)
Well, if you agree with Greg KH... Alan Cox doesn't.
On the other hand, preloading a Linux distro on a high-end notebook (Lenovo's ThinkPad T60p starts at about $1,800!) is a stupid decision!
The best decision would be to preload a Linux distro on an entry-level notebook, and Lenovo has its 3000 series: C100, N100, V100, with C100 starting at $600.
To have a market impact, people should be able to buy a laptop priced $600-$800, preloaded with Linux.
I'm afraid Lenovo is not smart enough. You know, they're from the P.R.C., not from "the other China" (R.O.C. Taiwan)...
On the other hand, preloading a Linux distro on a high-end notebook (Lenovo's ThinkPad T60p starts at about $1,800!) is a stupid decision!
The best decision would be to preload a Linux distro on an entry-level notebook, and Lenovo has its 3000 series: C100, N100, V100, with C100 starting at $600.
To have a market impact, people should be able to buy a laptop priced $600-$800, preloaded with Linux.
I'm afraid Lenovo is not smart enough. You know, they're from the P.R.C., not from "the other China" (R.O.C. Taiwan)...
Anonymous Coward
@ Aug 05, 2006 (01:34 PM)
Why would you think that going high-end is a bad decision for Lenovo. The attainable target market clearly is professional users (or companies) who know and dig high-class hardware and software alike.
Joe User wants to play games with his notebook and wants also wants Windows thrown in ("for free"...) simply for the reason that he won't have to shell out money for it. Linux is for free after all, so why pay for a notebook with Linux if you can get it with Windows which is actually pricy...? That's the consumer-market for you. As far as professional users (companies) are targeted Lenovo are actually likely to sell a couple of units... This is underlined by the choice of SLED which is an office-centred distro.
Maybe Xandros or something for budget notebooks is to follow but going broad-consumer-market would be a much more 'adventurous' decision in any case.
Joe User wants to play games with his notebook and wants also wants Windows thrown in ("for free"...) simply for the reason that he won't have to shell out money for it. Linux is for free after all, so why pay for a notebook with Linux if you can get it with Windows which is actually pricy...? That's the consumer-market for you. As far as professional users (companies) are targeted Lenovo are actually likely to sell a couple of units... This is underlined by the choice of SLED which is an office-centred distro.
Maybe Xandros or something for budget notebooks is to follow but going broad-consumer-market would be a much more 'adventurous' decision in any case.
Béranger
@ Aug 05, 2006 (01:45 PM)
Because I believe so! It's ignominious to disconsider the home users that do NOT play stupid games!
1. If you preload a laptop with SLED, this is for the business users, agreed. But the popularity of notebooks these days should make manufacturers to consider the home market too! Why is everybody ignoring the home user?!
2. And the price is also an aspect to consider: why should you pay $100 for the OEM Windows license in a $700 laptop?! If you don't need paid support, you can use a completely free distro.
3. Also with laptops: it's almost impossible to buy a notebook without paying the "Microsoft tax". This is why the low-end models NEED to be supported, and an agreement to preload a Linux distro is needed! (OK, you can buy some models with FreeDOS, but you can't be sure the hardware is supported very well in Linux.)
1. If you preload a laptop with SLED, this is for the business users, agreed. But the popularity of notebooks these days should make manufacturers to consider the home market too! Why is everybody ignoring the home user?!
2. And the price is also an aspect to consider: why should you pay $100 for the OEM Windows license in a $700 laptop?! If you don't need paid support, you can use a completely free distro.
3. Also with laptops: it's almost impossible to buy a notebook without paying the "Microsoft tax". This is why the low-end models NEED to be supported, and an agreement to preload a Linux distro is needed! (OK, you can buy some models with FreeDOS, but you can't be sure the hardware is supported very well in Linux.)
Anonymous Coward
@ Aug 05, 2006 (02:18 PM)
Ah; this is a misunderstanding. Let me rephrase: Why would you think that going high-end is a bad decision for _Lenovo_?
Of course it would be nice of them if they tried to evangelize Linux to the masses. But is it good _business_? I think not.
Why is everybody ignoring the user? Because everyone is afraid that what usually happens to those who make the first charge will happen to them, too. Early adopters in a market (that perhaps doesn't exist, yet) take a risk.
And the cheaper the laptop model, the more units have to sold to make a net profit. ATM there is IMHO few new users to attract in the home-desktop segment of the market. The software range that home-users use is much more diversified - read 'cluttered' - than in the business segment. Hard to make everyone happy there. Some may even miss drivers for their multi-purpose-super-watch (yes, I've actually heard someone complaining). Let's face it, ATM only the existing Linux user-base would buy a low-end Thinkpad with Linux preloaded. The rest of the users are tough to convice, if they've even heard of Linux.
Going home-user is risky. No-one (of he first-tier suppliers) wants to go first.
I'm personally very happy that Novell could convince Lenovo to preload Linux at all on any of their models. For now, let's regard this as the success it is and hope for more to follow. Perhaps if this becomes a huge success the idea will spread that Linux could actually do better than Windows in many respects for many (though not all) home-users, too.
Of course it would be nice of them if they tried to evangelize Linux to the masses. But is it good _business_? I think not.
Why is everybody ignoring the user? Because everyone is afraid that what usually happens to those who make the first charge will happen to them, too. Early adopters in a market (that perhaps doesn't exist, yet) take a risk.
And the cheaper the laptop model, the more units have to sold to make a net profit. ATM there is IMHO few new users to attract in the home-desktop segment of the market. The software range that home-users use is much more diversified - read 'cluttered' - than in the business segment. Hard to make everyone happy there. Some may even miss drivers for their multi-purpose-super-watch (yes, I've actually heard someone complaining). Let's face it, ATM only the existing Linux user-base would buy a low-end Thinkpad with Linux preloaded. The rest of the users are tough to convice, if they've even heard of Linux.
Going home-user is risky. No-one (of he first-tier suppliers) wants to go first.
I'm personally very happy that Novell could convince Lenovo to preload Linux at all on any of their models. For now, let's regard this as the success it is and hope for more to follow. Perhaps if this becomes a huge success the idea will spread that Linux could actually do better than Windows in many respects for many (though not all) home-users, too.
Béranger
@ Aug 05, 2006 (02:27 PM)
Oh, right: I forgot all the companies _despise_ the individual user, after all.
I will not buy a new laptop unless it will be free of the Microsoft tax. And I would like to have it preloaded with a completely free distro (SLED is _not_ free) just to see everything is supported and working -- I'll install my preferred distro later anyway.
But I am not the typical user...
I will not buy a new laptop unless it will be free of the Microsoft tax. And I would like to have it preloaded with a completely free distro (SLED is _not_ free) just to see everything is supported and working -- I'll install my preferred distro later anyway.
But I am not the typical user...
As for the *BSD's, it is undeniable that progress is being made, especially with Desktop BSD and PC-BSD (though there is no Gnome). This year's Google Summer of Code will work on Xen support and improvements to the Ports Collection among others. OSNews also linked to the "secret" development plans for FreeBSD; the notables include a new USB system, DTrace, and ports to MIPS and ARM. One of these days I migt try PC-BSD or Desktop BSD.
http://osnews.com/story.php?news_id=15136
http://code.google.com/soc/freebsd/about.html