Linux 2.4.18 Released

Kourino writes: "Marcelo announced the release of 2.4.18 a couple hours ago after 4 release candidates, but the tree marked 2.4.18 on kernel.org is missing the -rc4 patch that finally made the kernel releasable. Basically, what's marked as 2.4.18 is really -rc3, and what's marked as -rc4 is what should have become 2.4.18. According to Marcelo on #kernelnewbies, most users won't be affected, but people on SPARC systems should definitely grab 2.4.18-rc4. Your best bet is probably just to get 2.4.17 and patch to 2.4.18-rc4. Seems 2.4 is destined to be an "interesting" release branch ^_^; For the new release, head over to your favorite kernel.org mirror. (Marcelo will set things straight in 2.4.19-pre1.)"

Comments

NEW UPDATE

[8bahl]

I want the new MS OS killer... 2.4.666

Re:this is an enterprise ready os?

[Archie Steel]

Come on, it's been enterprise-ready for a while, now. Businesses don't need the "latest kernel available", they want the most stable. There are a couple of extremely stable kernels out there...

Sience this patch seems mostly for the Sparc.

[jellomizer]

It seems that Sun is now 20 years old. Happy Birthday Sun. So on its birtday it get a nice new Linux patch for its platform. Ok everyone Dig out your old Sparc 2's and install you Linux kernel.

Re:Sience this patch seems mostly for the Sparc.

[bconway]

The patch in -rc4 causes major problems on all non-x86 platforms. This includes Alpha, SPARC, and PPC, just to name a few.

Re:Sience this patch seems mostly for the Sparc.

[FredGray]

The patch in -rc4 causes major problems on all non-x86 platforms.

I think you mean that it fixes major problems on these architectures.

Version numbers.

[saintlupus]

Basically, what's marked as 2.4.18 is really -rc3, and what's marked as -rc4 is what should have become 2.4.18. According to Marcelo on #kernelnewbies, most users won't be affected, but people on SPARC systems should definitely grab 2.4.18-rc4.

Wow. Now that's professionalism, eh? Good thing that this whole Open Source badge makes it all okay.

Would the fifteen second delay to rename a couple files before release really have killed anyone?

--saint

Re:Version numbers.

[markj02]

Wow. Now that's professionalism, eh? Good thing that this whole Open Source badge makes it all okay.

No. What makes it OK is that the fix is out within 24 hours, that even 2.4.18-rc3 is very usable, and that people who run anything on Linux shouldn't be upgrading to a kernel that has just been released, even in the "stable" series.

Microsoft, just to pick one commercial example, releases a new version of Windows once every few years, and major service packs fairly infrequently. They also invest hundreds of millions of dollars in each release. And, you pay hundreds of dollars for Microsoft's software. That's what makes it not OK when Microsoft breaks a kernel release and users end up being stuck with it for months. And Microsoft releases packages with major flaws constantly, much bigger flaws than a forgotten rc4 patch.

Re:Version numbers.

[CentrX]

More like: no patch needed. One can download 2.2.18-rc4, which is identical to what the -final release was supposed to be. So, really, someone can just use that instead of the real -rc3/-final.

Anyway, not reuploading the release without a peep is a responsibility thing to do. He could have just reuploaded the release, with the possibility that a few people would get burned. Keep it all under wraps and no one would even know about it.

Re:Version numbers.

[qta]

I don't know why you drag MS into the picture, not that I am anti-[MS basher] or anything, I just think you missed the point. It is a good thing that Marcelo and the kernel guy managed to put out fixes quickly. But rushing to release some fixes and mislabel the source tree along the way is - as the original comment put it - not very professional. We are talking about the *stable* Linux kernel source tree here. We all make mistake sometimes, and we need to learn from mistake not denying it. I am glad Marcelo is open about it on #kernelnewbies

Re:Version numbers.

[Skjellifetti]

You are using the wrong word. A mistake like this does not imply a lack of professionalism. The best professionals make mistakes. It implies only a lack of perfection. And if Linux and everything about it were perfect, then why would it need patching at all?

Re:Version numbers.

[Arandir]

That's not professionalism, it's a complete lack of any plan. Sometimes I suspect Linux doesn't even have a release manager. What? They don't? I knew it!

Open Source means that such bonehead blunders get fixed quickly and efficiently. In the meantime, this is the stable branch of the poster boy for Open Source. This raw egg everywhere certainly demonstrates the openness, but it doesn't do jack to demonstrate any professionalism.

Of course, nothing in life is perfect. But the whole 2.4 branch has been plagued by crap like this from day one. Frankly, Linux is starting to get a reputation, and it's not a pretty one.

Re:Version numbers.

[jelle]

Making a mistake is human, being direct and clear about it is being professional. Summarizing, Marcelo is a professional human being. We can all rest assured that the next time he'll doublecheck his releases. What's all the fuss about?

Re:Version numbers.

[MentlFlos]

Kinda like how service pack 2 for NT on alpha "fixed" so much stuff, like the DHCP server. It fixed it so well that it just didn't work at all.

The MS solution was "don't install SP2, just wait for SP3 to come out." That was many months down the road.

Re:Version numbers.

[markj02]

I dragged in MS because the article I was responding to implicitly compared open source software with "professionally developed" software. Microsoft is simply the most prominent example of such software (in the sense of "professional" as "being done as a livelihood", not as in "being done with skill and experience").

Re:Version numbers.

[CentrX]

No one's "keeping it all under wraps", it says right in the changelog that the patch is missing.

Re:Version numbers.

[Trejus]

*yawn*

I've heard this all before, linus sucks, the kernel isn't stable, 2.4.x is horrible. How do you know 2.2 series wasn't plagued by crap like this before. When that series was in it's infancy, most of us didn't use Linux and weren't reading slashdot to discover all the mistakes that were made.

Btw, Linus didn't manage this release, Marcello did. People make mistakes and the best part of this open process is that it doesn't take long for someone to notice something isn't right and warn the community about it.

Ok so yeah, the kernel roll out for the 2.4 series hasn't been as smooth as we all would like it to be, but does that really matter? The casual geek, like me, updates his kernel whenever get gets a few free hours and there's one that's been around for a while. I used 2.4.5 and had no problems. I upgraded a few days ago. I don't go out looking for the newest kernel because it's not that important to me. You seem to be getting personally offended by the fact that things aren't perfect. It's a feeling that's pretty common and one i just don't understand. Yes we like things to be professional, but professionallism is for companies. The linux kernel has been and always will be the domain of a rag-tag group of developers, people who do it just for fun and they deserve credit for putting forth the effort at all. If you want professional, go to red hat, suse, or mandrake which are the professional front ends for linux, and are what most people see.

Re:Version numbers.

[Arandir]

How do you know 2.2 series wasn't plagued by crap like this before.

Because I've been using Linux since the 2.0 days. Remember, not everyone in Slashdot is still in kindergarten. Some of us actually have histories longer than six weeks.

I'm not blaming any individuals for the mess of 2.4. From my perspective, Linux has outgrown its development structure and is in desperate need of a new one.

Ok so yeah, the kernel roll out for the 2.4 series hasn't been as smooth as we all would like it to be, but does that really matter?

It matters because the Linux reputation is at stake, During 2.0 and 2.2, Linux had a reputation for utmost reliability and stability. That reputation is now eroding. People looking in at the Linux world see "whoops" releases. That's not good.

Yes we like things to be professional, but professionallism is for companies.

That's the whole problem now, isn't it? You all have this alien notion that professionalism is bad. People look at Linux and they see a bunch of developers PROUD that they don't do professional quality work. They see people proud that there is no design. Proud that there aren't any specs. Proud that there isn't any formal testing.

If that's the way you want Linux to be, then fine. But don't expect Linux to have any solid reputation outside of your own little clique.

Re:Version numbers.

[xanadu-xtroot.com]

No, no no, man. What CentrX was trying to say is that Marcelo could've easily uploded a diffrent tarball then he had origanilly released to "cover his ass". He didn't. He admited an "over sight" and submitted a patch. THAT is the attraction of open source. The mistake was caught, fixed and made available in less than 24 hours. Try to get *ANY* of the comercial companies to do that.

RANT Src="OffTopic"
Not even OS companies. How long did it take for the AIM hack to get fixed? XP had a full blown service pack a week after it's release? Then there's Universal Plug-n-Pray...

less than 24 hours ain't bad at all. I think I personally have "gotten" the best attention from Open Source projects. Even something like the IRC client I use. I've e-talked with a couple of the developers a couple times now. Somethings got fixed and submitted to CVS while on with them. My other (real) e-mail address can be found in a good few of Mandrake's RPM's because of things I've reported and helped fix.

Try that with a comercial company.
/RANT

Re:Version numbers.

[CentrX]

The fix was available /immediately/. One can just use the rc4, which was available before "-final" was released, and it was well-publicized or somesuch less than 6 hours later. So this is much better than even 24 hours.

Re:Version numbers.

[Trejus]

Ok, sorry, I had no idea. I wasn't around for the early kernels.

The point I was trying to make is that we should understand the difference between Linux the business and linux the operating system. You want professionalism, but i don't think you can ever really get that from a bunch of developers. Most of the people that do linux do it for fun, and fixing obscure bugs is not fun. But that said, I don't think it's fair to critisize the kernel maintainers for a complete lack of professionalism. Atleast when they make a mistake, they come out and admit it. There have probably been hundreds of "professional" or commercial "whoops" releases, but we never know about them because thier developement process is closed.

I'm not exactly sure that linux is getting a bad reputation outside of the core of linux die-hard geeks. Who else reads /. and the kernel mailing lists to find out what the "whoops" releases are. Most people i know (yeah ok so that's about three people) equate linux with "Red Hat" or "Mandrake" and those two companies do a relatively good job of keeping thier noses clean. Those who are intersted in the developement process do find out about all the mistakes and stupid stuff that goes on, but on the whole, those just looking for finished product end up getting a polished and professional one.

Re:Version numbers.

[jo42]

Ok, so he forked up this time. We'll forgive him. But next time, he better get it right.

Re:Version numbers.

[jo42]

> People looking in at the Linux world see "whoops" releases. That's not good.

That's because the Linux kernel architecture and development methodology is a hack that can only be carried so far before it collapses.

Which "archs"?

[srw]

> Load code did not set personality for
> binaries without an interpreter: This was
> breaking static apps on several archs

Okay, but Which architectures does this affect? It might be nice to know.

I can't trust this release...

[Anonymous Coward]

the tree marked 2.4.18 on kernel.org is missing the -rc4 patch that finally made the kernel releasable. Basically, what's marked as 2.4.18 is really -rc3, and what's marked as -rc4 is what should have become 2.4.18. According to Marcelo on #kernelnewbies, most users won't be affected, but people on SPARC systems should definitely grab 2.4.18-rc4. Your best bet is probably just to get 2.4.17 and patch to 2.4.18-rc4. Seems 2.4 is destined to be an "interesting" release branch ^_^;

I've had enough kernel problems in the past. The degree of uncertainty presented around this latest Kernel doesn't inspire a whole lot of confidence.

Think I'll hold off for a while, thanks.

What a silly thing to do

[talonyx]

Wouldn't it have been smarter to create a proper Linux-2.4.18.tar.gz (and bz2), and update the kernel.org and mirror sites, and THEN post to Slashdot and other such places?

This is just going to confuse people!

Re:What a silly thing to do

[llamalicious]

It's called trial-by-fire.

Once the flames die down, and the dust settles, he'll probably not do that again.

Re:What a silly thing to do

[felipeal]

Unfortunately, it's not that simple. Once he released the new version and posted about it, he couldn't change the binaries. That would means 2 different binaries were released for the same file (even if in a short period of time), and that could cause even worse problems (how could someone know, in the future, which version a file really is? What if a mirror didn't get the new files properly?)

Re:What a silly thing to do

[Gaccm]

simple, replace the old tar balls (those arn't binaries) with the new ones AND make a show of the md5sum of the new kernel. that way people could download from a mirror (most mirrors show the md5sum anyway) and md5 it and see if they get the same value.

Re:What a silly thing to do

[Anonymous Coward]

right, and then brace for the zillion emails coming from all over the world of people complaining about how the md5sum suddenly changed on a kernel image they already downloaded AND verified...

Re:What a silly thing to do

[eggnet]

But it is that simple. Release the originally intended 2.4.18 as 2.4.19.

Few things you should know :)

[Skuld-Chan]

Unless this kernel patched it - the only way I could get any 2.4.x kernels running on older sun hardware (like in my case sun4m systems) was to enable smp support - even if you only have one cpu. I found that if it wasn't enabled the kernel would hang on startup.

2.4.17 is actually quite stable on my old SS20's - one of which is doing firewall stuff right now on att broadband.

Re:Few things you should know :)

[uzi]

Actually, sparc32 as a whole (sun4, sun4m, sun4d and sun4c) lacks a maintainer at the moment. They're all a part of the same arch tree (arch/sparc/)... the only sparcs being actively maintained are the 64-bit UltraSPARC machines in the sparc64 port. This was mentioned by Dave Miller on sparclinux@vger.kernel.org the other day.

But yeah, sun4c is broken, but not too badly broken. I've tracked it to the point where it's trying to start init... it does a page fault, but stops after just the one and goes into the cpu_idle() infinite loop. Probably won't be too hard to fix.

2.4.19-pre1 is out now.

[dietz]

Here's the ChangeLog.

That was quick. (It includes a lot more stuff than just the patch missing from rc4).

Re:2.4.19-pre1 is out now.

[felipeal]

It includes a lot more stuff than just the patch missing from rc4

Ironically, it doesn't mention the missing patch, so I just hope he didn't forget it again :)

(just kidding, he probably had the pre1 ready already, and decided to release it quickly to fix the rc4 problem, but then he just forgot to add this new 'change' to the log).

Re:2.4.19-pre1 is out now.

[dietz]

It's there:

- Apply the SET_PERSONALITY patch missing from 2.4.18 (me)

Re:2.4.19-pre1 is out now.

[bconway]

The problem is that it doesn't only include the missing patch from -rc4. A lot of stuff considered too experimental for the 2.4.18 kernel was saved until 2.4.19-pre1, but that version is now required for most non-x86 platforms. Suddenly people are locked into testing experimental code if they're using a non-x86 architecture and require updates in the 2.4.18-pre's, and this is a BAD THING.

Re:2.4.19-pre1 is out now.

[gmack]

But wait... just use 2.4.18-rc4!

Re:2.4.19-pre1 is out now.

[felipeal]

Hmm, my bad :(
I swear I double-checked before posting, but I must got confused with the lower/upper cases, as the original changelog was:

- Load code did not set personality for
binaries without an interpreter: This was
breaking static apps on several archs (Tom Gall)

Re:2.4.19-pre1 is out now.

[_ECC_]

The reason it includes more than just the SPARC fix is because Marcelo queues up major additions/fixes for early pre-patches, to allow for more testing. So most of these fixes were probably approved but not added while 2.4.18 was in the 'rc' stage.

--
ecc

Re:2.4.19-pre1 is out now.

[germanbirdman]

Don't forget it is a -pre. Marcelo was keeping a lot of patches out of the late 18-pres because he wanted to get the damn thing stable.

We should also note the time that has gone past between 2.4.17 and 2.4.18 - more than two whole months. This is Marcelo's first real own kernel in my opinion. 2.4.16 was a bug fix for 2.4.15 - 2.4.17 came out only a few weeks afterwards, but this baby really had time to mature.

This is also why I don't mind reading this (commenting on all the "This is not freshmeat" discussions) on slashdot. This is a stable kernel, the first for a long time. It is not like in the times when a new "stable" kernel came out like every other day and people got annoyed.

I have 2.4.18-rc4 running for almost 9 hours now (basically since it came out) with setiathome, dnetc, tftpd, nfs, smbd, cups, pppoe, bind9, dhcpd3, tftpd (for remote booting clients) using huge reiserfs partitions and I like what I see. It is just my busybox DSL router, firewall and file server, so not really a production system, but it is in a server case, running a dual pentium II so hardware that while not fully up to date resembles that of production servers in medium sized companies. I don't normally pull a new kernel everytime one comes out, but I suddenly needed NFS support which I didn't have compiled in before, so I decided to upgrade to either .18 or the most current release candidate. Before that it was running for at least a month on .17.

Re:2.4.19-pre1 is out now.

[Eil]

Or you could simply use the current stable kernel that's running on your machine right now.

My production machines are all still on 2.4.5, perhaps the most stable of the 2.4 series. (In my experience.)

Re:2.4.19-pre1 is out now.

[Cro Magnon]

I'm sure 2.4.19-pre1 is better than some release kernels *cough*greased turkey*cough*

Re:2.4.19-pre1 is out now.

[Spoke]

Or you could leave yourself open to the security holes which existed in previous kernels to local users.

If you don't have anyone that uses your system, then staying on 2.4.5 might not be a bad idea, but at the very least you should be testing the latest releases in case you find a problem that forces you to upgrade.

Re:2.4.19-pre1 is out now.

[Eil]

It all made perfect sense to me. And I'm well aware of 2.4's history. But you'd have to actually read my message to see that because you missed the part where I said I was still running 2.4.5, one of the more stable kernels of the series when we're up to 2.4.18.

Re:2.4.19-pre1 is out now.

[Eil]

More like I should not be bothering with new releases when the current version I'm running has no problems after performing superbly for the better part of 7 or 8 months.

If you know of any serious (read: remote) exploits for 2.4.5 then I'll gladly upgrade but I currently know of none.

Re:this is an enterprise ready os?

[J'raxis]

Yeah, in the 2.2.* branch

Re:this is an enterprise ready os?

[brer_rabbit]

There are a couple of extremely stable kernels out there...

Yeah, and we'd like to know which ones!

Please seperate Linux kernel from Linux OS topic

[Nailer]

You've all heard this before, but that way people who aren't particularly interested in minor kernel revisions but are interested in general Linux stories can filter away the linux kernel topic.

amd cache coherency

[theEdgeSMAK]

Has anybody addressed the amd cache coherency bug? I'm still booting with mem=nopentium. That makes me sad.

jeremy

Re:amd cache coherency

[strredwolf]

I belive this was fixed with 2.4.18-pre1, which was the first patch in.

But I'm still scared over 2.4.18 missing the -rc4 patch, and both 2.4.18 and 2.4.19-pre1 are fresh. I'm not going to compile it even on my own system until 2.4.19 is released.

Oh Alan, where are you...

Re:amd cache coherency

[strredwolf]

You idiot. You're changing the subject. All this guy would need is the -pre1 patches.

Besides, you're not even reading the changelogs. Guess what rc4 corrects? I'll tell you, no, I'll quote the freakin' log:

rc4:

- Load code did not set personality for binaries without an interpreter: This was breaking static apps on several archs (Tom Gall)

I consider this a Major, showstopping bug for a production server online *NOW*. If it's this new and broken out of the box, no way I'm going to suffer it, even on my own home hardware.

Do you have hardware that spits out 1-2 gigs of pics and data a day? Are you going to stand by your words and stop being an Anonymous Coward, kook?

"Go ahead, punk. Make my day." --Hal, in Stone Trek, "20001 BC: A Space Oddity - Part 1."

Re:this is an enterprise ready os?

[Anonymous Coward]

Hey if Windows can claim to be enterprise ready with the csrss.exe backspace bug present I think Linux can claim to be enterprise ready.

Re:this is an enterprise ready os?

[reaper20]

Redhat and Suse provide some damn good kernels. I've had no problems whatsoever. They really stress test those things.

Boy, everyone sure loves to jump on the "2.4.x sucks" bandwagon. Sure, there were some issues in the past, but I would like to know how many people reading slashdot right now are really seeing all of these problems.

90% of you who got burned and will "go back to 2.2.x" were probably being stupid and tried it on a production server and got properly burned.

Test your shit before you deploy, if you're not doing that then you're an idiot.

Re:this is an enterprise ready os?

[Pope Slackman]

There are a couple of extremely stable kernels out there...

There sure are. Here's one and here's another.

C-X C-S

Linux/SPARC

[uzi]

I believe the problem is on sparc64 (and not sparc32). It's a trivial patch... just a few lines, so it's easy to have been overlooked.

Anyone using a SPARC/Linux machine that wants the latest and greatest should always turn to the
vger.samba.org sparc/linux kernel cvs tree. It's always got the latest stuff for both 64-bit and 32-bit SPARCs (and networking as well).

That said, 64-bit SPARC machines should run fine with the recent kernels. For the 32-bit SPARC machines, I can only comment on the sun4m and sun4c machine. Currently, the sun4m machines should boot and be ok... the sun4c machines do not.

sun4m machines: ss20, ss10, ss5, ss4, lx, classic, javastations
sun4c machines: ss2, ss1+, ss1, ipx, ipc, slc, elc

Re:Linux/SPARC

[Bert64]

What exactly is the problem with sun4c machines right now? It gets to the "freeing kernel memory, 96k freed" and freezes, no errors or anything. just stops dead And more importantly, when is it expected to be fixed?

How about a new Slashdot feature.

[jellomizer]

On the top right of the page a little box giving the current kernel version and the date/time of the change when you click on it you can get to its message board. That way you guys can get your Kernel Fix and read about your problems and us people who dont update the kernel in the main messages. I like getting news about Linux and Also Major Kernel versions. But all these little ones gets annoying. It seems like if anyone really want to post a story with there name on it they keep two browsers open and one refreshes kernel.org untill a version apears and then they post a message stating that. Say I can make a script to do that. (But I wont)

Re:How about a new Slashdot feature.

[gmhowell]

There is a slashbox available with this information. Granted, it won't be in the top right, but it will be on a logged-in user's home page.

Re:How about a new Slashdot feature.

[jdavidb]

It can be in the top right if you want. ... Or at least mine can. Am I the only one with movable slashboxes? Or were you only saying it wouldn't be at the extreme top right (like up against the browser scroll bar)?

Re:How about a new Slashdot feature.

[gmhowell]

Both:)

I forgot that slashboxes are movable.

Re:How about a new Slashdot feature.

[duckie13]

Not to be a pest, but I must be blind..after searching through all the Slashboxes, I found nothing like that. Could anyone tell me which it is so I can stop feeling like a dumb newbie? :/

Re:How about a new Slashdot feature.

[gmhowell]

Sorry. I think I screwed that one up. I know it used to be available...

Just checked. I can't find it. Maybe they no longer have it. I must apologize. I used to have it on my page, but dropped it when 2.3 went to 2.4.

It was really nifty. Maybe someone else can find it?

Re:So. uH what?

[Karn]

It's news, b/c all the 'know-the-only-way' types will come out of the woodworking, denouncing methods and procedures of which they know nothing about.

Only in the patch

[awptic]

Appearantly, the missing piece from -rc4 is only missing from the patch, NOT the full tarball.

2.4.18 IS OK?

[beezly]

There's a report on linux-kernel that the 2.4.18 tarball is actually ok.

Has anyone had an opportunity to check?

Here's a link to the chap saying it's ok:
http://marc.theaimsgroup.com/?l=linux-kernel& m=101 467401412021&w=2

Re:2.4.18 IS OK?

[Anonymous Coward]

nope, it needs this:

--- linux-2.4.18/fs/binfmt_elf.c.orig Mon Feb 25 14:56:59 2002
+++ linux-2.4.18/fs/binfmt_elf.c Mon Feb 25 14:57:17 2002
@@ -564,6 +564,9 @@
// printk(KERN_WARNING "ELF: Ambiguous type, using ELF\n");
interpreter_type = INTERPRETER_ELF;
}
+ } else {
+ /* Executables without an interpreter also need a personality */
+ SET_PERSONALITY(elf_ex, ibcs2_interpreter);
}

/* OK, we are done with that, now set up the arg stuff,

Re:2.4.18 IS OK?

[beezly]

Hmm, I'm not convinced.

It looks like the tarball is ok, but the patches haven't been created properly.

Re:2.4.18 IS OK?

[AirLace]

Just checked. The patch is NOT applied in the tarball. Damn.

Re:2.4.18 IS OK?

[beezly]

It may only be in the gz tarball, not the bz2 one.

linux-kernel is slowly working out what's what :)

Re:2.4.18 IS OK?

[Uncle Warthog]

It may only be in the gz tarball, not the bz2 one.

I just checked the .gz tarball. The lines from the patch are not in the file.

Re:2.4.18 IS OK?

[beezly]

looks like I was wrong ;)

2.4.18 is missing the patch in every format.

Re:2.4.18 IS OK?

[ChadN]

Which is actually a good thing. Now patching from 2.4.18 is not predicated on how you got to 2.4.18 (although those running 2.4.18-rc4 have to patch to 2.4.18 proper, before patching to 2.4.19).

Re:Allow me

[Ironfist_ironmined]

...because people dont care when windows is updated?

But what *I* want to know is...

[Aexia]

whether the Linux underwear for this version has been released yet.

Re:But what *I* want to know is...

[pete-classic]

I think that:

#chgrp fine-honeys
#chmod 770 *

would be better.

-Peter

Release frequency, release engineering thoughts.

[seantrue]

There is a natural rhythm to the frequency of a successfully released software system, which is an indescribably complex function of the size of the system, the size of the releasing team, the maturity of the software, the sophistication of the end user, and the cost of making mistakes.

What is an ideal release frequency for one point in this space, is not at another.

At one point I worked at a DOS extender company (Rational Systems, not related to Rational of California), and we released the software every week. The system was small, the team was small, the customers were very sophisticated, and the value of adding new features was very high. We were praised for being responsive. Three years later, the software was much bigger, the release cycle was down to 2 times a year, and the value of not adding new bugs to the old features was very high. We still got good marks for technical support.

Unlike most commercial software, it's hard to point at revenue streams for Linux that justify the midlife software development expenses like full-time, paid-for, this-isn't-fun-but-it-has-to-get-done release engineering. Although there is a large virtual software team for this OS, I strongly suspect that there is less infrastructural support than you get with old fashioned, iron vendor supported systems like Solaris, HP-UX, et al. TANSTAAFL, folks.

Don't get me wrong, I use Linux daily, my servers run on it, and I depend on a variety of other open source software (particularly Python!). I even buy RedHat/KRUD releases just so that some value flows back into the release process from a happy recipient. But I sometimes feel like holding my breath while installing that next kernel release!

TANSTAAFL -- There ain't no such thing as a free lunch. Thanks, RAH, wherever you are!

Why not release 2.4.19 immediately?

[markj02]

2.4.18 with rc4 applied seems to be a stable, well-tested kernel that has gone through all the testing needed for a release. It deserves its own kernel version number. Since 2.4.18 is not-quite-right, why not just make what was going to be 2.4.18 into 2.4.19?

Re:Why not release 2.4.19 immediately?

[GigsVT]

Then everyone would bitch and whine, like they did when Linux released a kernel three days after the last.

But you can't, really...

[Lethyos]

If you must post to get self-validation, try making up a story or something.

I'd agree with you if the editors didn't reject every damn story submitted. Lots of interesting stuff gets let go, but editors ALWAYS seem to make sure that a patch level kernel release gets posted. It's silliness. The editors here are just too pigheaded for users to submit interesting stories and get them known. Disagree? Stupid crap like this Linux-newbie question gets posted while intelligent topics get thrown out.

The only option is to whine about the stories that do get posted.

Re:But you can't, really...

[nomadic]

The thing that annoys me is that they don't even seem to read the stuff that I submit anymore. My last story submission was rejected in about 2 minutes (I was keeping track). What happened to all these hundreds of story submissions that they claim overwhelm them? Can't be that many if you have a 2 minute turnaround time...

I could understand it if I constantly spammed them and they were just sick of dealing with me, but I average maybe a single story submission every 2 months or so. If they aren't even reading them, then just let me know so I won't even waste that 5 minutes every 2 months...

Re:But you can't, really...

[leviramsey]

With regards to your .sig: [I'd post this in response to his journal, but the discussion has been archived, and you don't have any email listed...], I agree with you. While I do not metamod any down mod as Unfair, I have very high standards for qualifying as a neutral down mod.

At this moment I am working on a Journal post that will propose a simple means by which Over-rated and Under-rated mods can be metamodded.

Why can't they fix it now?

[irq]

Why can't they just rename the files?
I mean, these people make operating systems, right?

Re:Why can't they fix it now?

[Phork]

Because it would be bad to have two seperate versions of 2.4.18 out there. Because once they put it into the mirror system, it starts to propogate, and it would proably cause problems if a second version of 2.4.18 was put up.

Re:Why can't they fix it now?

[Bronster]

Honestly, what they _should_ have done is put out 2.4.19, with nothing but that _one_ patch, marked 2.4.18 DONTUSE, and started with 2.4.20-pre1. Who cares about an extra number?

Re:Why can't they fix it now?

[BlowCat]

Care to explain? What's exactly wrong with 2.4.18? It seems to me that you are basing your opinion on comments of those who don't know what they are talking about.

I don't see anything in 2.4.19-pre1 that would justify immediate release of 2.4.19 and marking 2.4.18 as DONTUSE.

2.4.18 was released only once, with the personality bug fixed. There was no "other" 2.4.18.

If a bug is found in a release candidate, it is the right thing to do to make another release candidate with this bug fixed and no other changes, and that's exactly what has been done. I see no point in releasing 2.4.18 with a known bug and marking it DONTUSE right away.

Re:Why can't they fix it now?

[SmittyTheBold]

What I just read is this:
rc3 was released with a bug.
rc4 was released with the bug fixed
rc3 was marked final anyway.

so now we have a final relaese with a bug that a *later* rc did not have.

Re:Why can't they fix it now?

[Bronster]

rc3 was released with a bug.
rc4 was released with the bug fixed
rc3 was marked final anyway.

More than that, it's a show-stopper bug on non-Intel architectures. That's sort-of bad and stuff - there's no reason to leave something with a known bug as the most recent release any longer than necessary.. there's an infinite number of version numbers available, it's not like anything would be wasted (and anyone downloading the patch would only have to get a couple of Kb).

Re:Allow me

[Zapdos]

Logon to slashdot
Click on preferences
Click on homepage
check the stories you dont want to read
scroll down to the bottom of the page
click on "save"

Was that hard?

Re:Kernels-R-US

[DocSnyder]

You forgot something:

• The preemptive patch
  
• The 'ac1' patch
  
• The XFS patch
  

Re:Kernels-R-US

[NotoriousQ]

You forgot the one guy who posts the predictions for the posts that will be posted.

^_^

[billybob]

hehe... its funny to see the ^_^ smiley becoming mainstream.

^_^

Re:Allow me

[donutz]

"Have you ever seen the back of a twenty-dollar bill...ON WEED?"

Wait, who's on weed? You, or the twenty dollar bill? Your .sig isn't clear on that point....

Re:Please seperate Linux kernel from Linux OS topi

[hawkbug]

I strongly disagree - I think the most interesting linux news is when the new kernel releases come out. I think this should be a slash poll to see what everyone else thinks.

VIA and Cyrix III must have rc-4

[Dark Coder]

If you have a Cyrix III 1Ghz and a VIA southbridge controller, you must have the -rc4 .

Re:VIA and Cyrix III must have rc-4

[einer]

Cyrix still makes chips? I thought they died a long time ago!

Re:VIA and Cyrix III must have rc-4

[Asdex]

Why?

Re:Linux Kernel 2.4.18 Changelog

[jjccss]

How long does it take for a site to run out of bandwidth after the news has been posted to /.?

IT'S CALLED KNOWLEDGE. It's nice to be able to read a quick reply that tells me w/o going to an archive whether or not I am going to use the kernel on the servers. Especially when the following link is omitted from the article.

Kernel 2.4.18 Changlog

Re:Please seperate Linux kernel from Linux OS topi

[DocSnyder]

I know it's redundant, but the so-called "Linux OS" has its own topic - GNU is Not Unix.

IANAS (I Am Not A Stallmanist), but without his work you would have to write your /. postings with Micky$hit Internet Exploder on W1nd0z3... that is, if /. existed at all.

Re:Please seperate Linux kernel from Linux OS topi

[LordNimon]

Then you can just subscribe to both Linux and Linux-Kernel stories.

WinXP vs. Win2000

[David Jao]

It's interesting that you picked Windows 2000 instead of Windows XP. If you're going to compare Linux with Windows and you choose the version of Windows that's one release back, you have to be fair and also choose the Linux version that's one release back, which is Linux 2.2. And no one's complained about the stability of Linux 2.2.

Perhaps the reason you picked Windows 2000 is that Windows XP has serious problems with security and reliability. If the admins quoted in that EWeek article had deployed Windows XP into production instead of on their testing systems, they would have had major issues.

Conclusion: Test your shit before you deploy, if you're not doing that then you're an idiot.

Re:WinXP vs. Win2000

[Strog]

You could have mentioned some of XP's other problems too. I have a test partition with XP pro on it and have not run into this particular issue yet. I hope I never do. Makes a crash a little more serious.

Re:WinXP vs. Win2000

[York the Mysterious]

I've just deployed XP to 250 workstations in a technology school. The machines here get hit and hit hard and I can say that they have performed much better than 2000 did. The management features are nice, the performance is better than 2k once your turn the gui down (shit OS X rip), and it's stable (except for IE since IE 6.0 sucks). -Tim

Re:WinXP vs. Win2000

[posmon]

Howabout... because 2000 IS the latest SERVER version, fuckmunch.

Re:Please seperate Linux kernel from Linux OS topi

[g1zmo]

I only play single-player games, so I shouldn't have to put up with stories about multi player games, so I need to be able to filter those out.

I don't really care about other people's opinions, so I need to be able to mask interviews too.

I'm not too bright, so stories about genetic research go right over my head, and I need a checkbox so I don't have to see those.

I don't have a laptop, I don't travel anywhere, and I only use ethernet networks, so I really don't want to see all those annoying stories about wireless protocols either!

Re:Allow me

[dead_penguin]

"Why is this news? This isn't freshmeat! Why don't you post every time Windows has a tiny update?"

We're already one step ahead of you! If you look around, you'll see most of the comparable windows stories are about bugs, problems and vulnerabilities. Windoes news here on /. seems to be so current that it gets posted long before there even is a fix. Linux bug-type news seems to have to wait until a whole new kernel gets released. I want equality, damnit!

Mirrors? huh

[buserror]

At this time, the 2.4.18 is neither on UK nor FRA mirrors. And the german one doesn't have the .bz2

So why announce it and whinne about the mirrors when they aren't available?

Result is, I'm downloading the .bz2 from... kernel.org direct.

Not for lack of trying...

Re:Mirrors? huh

[buserror]

You idiot, if it's not ready, why announce it?

I go have a look at the changelog, then I'll download it "because I'm here and it's there".

You can save your shift key syndrom for "why is the news posted when the host is not ready" and blame someone else.

The responsability of slashdoting kernel.org is not mine, it's.... slashdot's

Re:Allow me

[jbailey999]

More like "Who can keep up with the daily patches and updates"...

Re:what to expect now

[phoxix]

how about expecting useless posts such as the above??

sure, when I first saw posts like these, I thought they too were cool

but now all the lusers are posting posts like the above, and the fun no longer exists, so just stop it

Sunny Dubey

Re:this is an enterprise ready os?

[Woko]

The business I worked for needed a stable VM, stable NFS, stable journalling filesystem (ext3 is not the answer for all situations), high speed ethernet, hardware RAID (on newer hardware) and a fast TCP/IP stack.

With reiser & NFS having incompatabilities until about 2.4.10, the VM not being settled at all, various problems with intel ethernet drivers in earlier 2.4.x kernels, I think "stable" is far too generous a term.

Re:Please seperate Linux kernel from Linux OS topi

[Strog]

Kernel Poll
1: seperate kernel and linux stories
2: leave it alone
3: seperate Katz from his mouth
4: more coyboyneal update stories instead

Re:this is an enterprise ready os?

[OpenGLFan]

You really shouldn't have picked the backspace bug as an example. Is there anyone who hasn't found The Backspace Bug?

Re:what to expect now

[Hiro Antagonist]

You forgot:

11 Cutesy posts making predictions about other posts.

Re:what to expect now

[GigsVT]

You forgot:

At least 3 posts with a list of predictions about what posts will be posted

and:

At least one post replying to the prediction posts correcting the missing prediction of prediction posts.

Oh, and I hear Alan Thicke is dead, or something. :)

Re:this is an enterprise ready os?

[phoxix]

openbsd sure is stable, there is no doubt in the world that Theo de Raadt would produce anything else

however OpenBSD suffers quiet a bit from lack of cpu optimization. OpenBSD doesn't even currently support the MMX extensions for it x86 platform. Additionally, OpenBSD has a history of having a slower performing file system.

While OpenBSD is the world's most secure OS, and it sure is pretty damn stable, it would fail on the enterprise level. There is simply no way, Company XYZ would be able to roll out their production database system (or whatever requires loads of CPU and usage) on OpenBSD.

However, OpenBSD would kick the crap out of just about everything else for firewall usage. For those of you who haven't looked into it, check out OpenBSD's *AMAZING* ability to be an invisible bridge (not only can it invisibly bridge, but it can filter the bridge as well, which is something other OS'es have yet to be able to do).

Sunny Dubey

Also Available At:

[Un1v4c]

linuxupdate.kernel.org
; )

Re:Linux on the desktop: very convincing...NOT

[GreenHell]

2.4.18 isn't 2.4.18, it's 2.4.18-rc3, unless you're using an x86 in which case 2.4.18 is 2.4.18. Now, if you aren't using an x86 then 2.4.18 isn't 2.4.18, that's 2.4.18-rc4. See, it's really as clear as day. :-)

Re:this is an enterprise ready os?

[Zoopee]

Actually, I would say I feel the recent 2.4 kernels are more stable then the 2.2 ones.

Does anyone have any point of fact, or reason for claiming that the 2.4 branch is less stable? I mean, the first few were... But the late ones seem pretty fine to me.

Of course you could point out a few unsuccessful ones, like 2.4.9 with VM performance problems, 2.4.13 - I think - not even compiling with the parallel port driver, a filesystem corruption bug in 2.4.15-final... But those bugs were pinpointed in a few hours after the release... I wouldn't install a kernel right after it releases, but with 2.4, in the long run, the kernels are quite stable.

So right, sometimes the maintainers are drunk while releasing... But when they're sober, it's fine...

Well, now I know

[GSV NegotiableEthics]

Yep, now I know what kernel I'll be running on Debian Stable in 2005...

Re:Well, now I know

[Da Masta]

Yeah man, along with Mozilla 0.6, and _maybe_ Gnome 1.4, that should be one smoking system!

Re:Well, now I know

[GSV NegotiableEthics]

Lighten up, it was just a joke. I use Debian and I no longer run any other Linux, but sometimes I like to poke fun at the long lead times to stable.

Re:Please seperate Linux kernel from Linux OS topi

[PD]

Imagine that it's 1983, and I decided that I wanted to filter out everything that wasn't of particular interest to me at that time.

Now, jump ahead to 2002. I would have missed all the great leaps of the past 20 years. I wouldn't even have a clue that I would want a beowulf cluster of Commodore 64's! I might even post a comment like this to Ask Slashdot: "Hello Slashdot, I was in a restaurant the other day and I overheard some people laughing hysterically about something called Amiga. It sounded really fun. Doesn anyone know if I can buy one?" Oh, I would have been just like Rip Van Winkle if I had managed to filter out everything that wasn't of interest to me.

Please seperate whiny Linux kernel comments...

[festers]

You've all read this before, but that way people who aren't particularly interested in petty whining but are interested in insightful comments can filter away these time wasters.

Re:Please seperate whiny Linux kernel comments...

[Nailer]

Please seperate whiny Linux kernel comments... (Score:1)
by festers

Yeah, people who whine suck ;)

Re:this is an enterprise ready os?

[emir]

emir@tvrtko:~$ uname -a
Linux tvrtko 2.4.14 #1 Tue Nov 20 00:54:59 CET 2001 i586 unknown

emir@tvrtko:~$ uptime 00:15:34 up 51 days, 16:54, 1 user, load average: 0.00, 0.00, 0.00

seems pretty stable to me, offcourse this machine doesnt have that high load, we are like 4-5 ppl using it....

Re:Linux is dying

[jrnchimera]

Last time I checked Linux supports a much wider range of hardware than FreeBSD. To me that makes Linux very appealing.

Re:Block anonymous Cowards?

[GigsVT]

Oh well, watch that karma fly!

Anyway, if you want to be anonymous and be able to post at 1, just follow this following easy as shit recipe:

1. Go over to satan, uh I mean hotmail.com. Create a fake passport with total bullshit info.

2. Create an account on Slashdot using your new-super-cool hotmail email address.

3. Log in.

4. Tell everyone to shut up about the whole AC thing, AC isn't needed to be anonymous, this is the Internet for fuck's sake.

This is just as anonymous as the AC option, for all intents and purposes. It takes like 5 minutes to set up.

Re:Please seperate Linux kernel from Linux OS topi

[msaavedra]

without his work you would have to write your /. postings with Micky$hit Internet Exploder on W1nd0z3

You could have picked a better example there, because as far as I know, there is no GNU web browser. I guess galeon (which I use, btw) could qualify although it is not an FSF project and the rendering engine is from Mozilla. I would guess that most Linux people use Netscape 4.7x, Mozilla or Konqueror, which are definitely not GNU projects. And of course X11 is certainly not GNU.

I agree with your general point though. Splitting Linux topics into kernel and OS would be a bad idea, mainly because I'm of the opinion that the kernel is the OS. All the other stuff can go in the X section, the GNU section, the KDE section, etc.

True reason for missing a patch

[Jacek Poplawski]

Say what you want, but looks like true reason for missing SPARC patch is just a revenge on Sun for this .

Just give me a kernel that turns off my computer

[Bradee-oh!]

The only kernel that's ever actually powered down my machine on shutdown or halt is the special 2.4.8 included with my Mandrake distro.

Has anyone else had this problem and actually fixed it??

Re:Please seperate Linux kernel from Linux OS topi

[Lozzer]

The FSF web browser is called wget.

ducks...

Re:Please seperate Linux kernel from Linux OS topi

[Nailer]

This is off topic, but actually, I think I'd be using the BSD tools (with the Linux kernel). Yes the FSF deserves credit for their work, but so do many other projects, like KDE, GNOME, BSD, etc.

Re:... or deploy Win2k

[yintercept]

You have a good point. If you wait a year or so, all of the hardware and software companies will redesign their products around the new version so that you can get a somewhat stable server out of the mess.

Re:Just give me a kernel that turns off my compute

[N8w8]

I use APM and supply "apm=power-off" as a kernel parameter. Good luck :)

Re:Please seperate Linux kernel from Linux OS topi

[Eil]

This reminds of those people who complain that there's nothing good on TV but spend 6-8 hours watching it anyway.

Take the Linux kernel updates out of the Linux OS topic? If you do that, then you'd have to rename the Linux OS topic to "GNU-based OS" topic because the only thing that makes Linux Linux is the kernel.

If you don't want to read about "minor" version releases then for god's sake use your scroll bar and read something else. Betteryet, read a book

Re:Please seperate Linux kernel from Linux OS topi

[greenrd]

No, it's called emacs. Well, it's part of emacs, anyway.

Re:Just give me a kernel that turns off my compute

[Lumpy]

redhat 7.2 has for 5 different laptops and 3 different pc's this is with kernels ranging from 2.4.7-pre 10 to 2.4.17

be sure that the APM is compiled in, and if you have SMP it will never ever work. you cant do apm and SMP at the same time.

Re:this is an enterprise ready os?

[Liam]

check out OpenBSD's *AMAZING* ability to be an invisible bridge (not only can it invisibly bridge, but it can filter the bridge as well, which is something other OS'es have yet to be able to do).

You mean like Linux?

Re:this is an enterprise ready os?

[J'raxis]

An RPM of 2.4.17 that I tried on this junk box (486DX2) of mine randomly rebooted it. Enough said.

Re:this is an enterprise ready os?

[elbles]

Well, then try a 2.4.16+ kernel with SGI's XFS patches; works great for me. ;-)

Re:this is an enterprise ready os?

[Zoopee]

Uh... well, I never encountered that... Are you sure that box isn't faulty?

Scheduling patches

[TeknoHog]

I'm not compiling until this patch is available for 2.4.18. It combines Ingo's O(1) scheduler with RML's preemptible kernel code patch.

Re:Scheduling patches

[AYEq]

It doesn't combine the two, rather you first apply the O(1) sched. then this patch. I am running this combo on 2.4.18-rc3. I would bet that it would apply cleanly to rc4. (you can test this with "patch --dry-run") So grab the last O(1) here and then the rml preempt patch.

Re:Scheduling patches

[YellowBook]

The O(1) scheduler and preempt patches for 2.4.18-pre-bignum apply cleanly to 2.4.18-final. Have fnu.

Re:Scheduling patches

[TeknoHog]

Thanks! It's running alright now.

BZZZT* WronG!

[TeknoHog]

The sun is about five billion years old. ;-)

Re:Just give me a kernel that turns off my compute

[MrZaius]

I've had the same issues from mandrake 7.1 on, but I figured out what was wrong.

In every new kernel I'd compile, for some inexplicable reason, I would enable ACPI and disable APM. Enabling APM BIOS support (under General Setup in the kernel's config)

This worked just find with 2.4.18-rc3 on my VIA Apollo Pro 133A chipset motherboard.

Re:Just give me a kernel that turns off my compute

[sparkz]

General Setup -> APM BIOS Support -> Use Real Mode APM BIOS Call To Power Off

Works for me (unheard of laptop, 2.4.7)

New SPARC kit? Move along there's nothing to see.

[spinlocked]

Linux was fantastic when Sun released Solaris 7 - their first 64 bit OS. It meant that you could give a new lease of life to those 'cherished' old SPARCstation 1/2/Classic/LX/10 etc. The reason? Solaris bloat. Solaris has grown up with the Sun hardware range. Versions 7 and 8 have a great deal of code which supports later generation hardware. It's getting pretty difficult to fit it on a 1 gig disk - even a custom install, let alone the 420MB drive which came with my old sparcy2.

The non-pagable kernel memory used to fit, just about, in 32MB with some to spare for buffer cache (well, 2.5.1 did) . Nowadays it just swaps horribly. Why you ask? the old SPARC workstations don't have much of the hardware which new versions of Solaris provide support for (much of it installed even if you don't have the hardware grr.). Solaris has a mature multithreaded kernel, it has amazingly well tuned, truly scalable, kernel synchronisation primitives (check out the book "Solaris Internels" - Mauro, Mc Dougall) it has in-kernel support for Sun's hardware enterprise features; dynamic reconfiguration (the ability to tell Solaris to stop using memory, CPU or IO devices on a certain system board, drain the memory to swap, re-dispatch the active processes to other CPUs, remap the IP addresses to other cards, detach the board, replace, reattach - start using the new hardware - no reboot), hotplug PCI, processor sets, dynamic system domains etc. etc.

Decent Sun boxes (by that I mean anything with more that 4MB L2 cache and SCSI disks - a curse on Ultra 5/10/X1/SunBlade 100s), will run Solaris 8 very well, plus you get a tier one Oracle/Sybase/Java platform, with all of your favourite window managers/web browsers/IRC clients etc. available for download.

Mark my words, once Linux starts making real inroads on the sort of Enterprise server kit (i.e. more than 8 SMP CPUs, and much more than 4gb RAM) that you need for serious financial/HR/government/pharms. type applications , it too will be bloated. You could argue it already is - my 486SX/8MB of RAM gave very good service as a firewall, using ipchains and kernel 2.2. Kernel 2.4 and iptables (and I suppose my new stateful filter) make it rather too slow to survive my next hardware cull. Ah well, out with the old...

In related news...

[_Neurotic]

I got my haircut last week. This brings my haircut to verion 8.2.3.1.7.2.5.4.312.5.6.43243.3

Oh boy!

Re:Please seperate Linux kernel from Linux OS topi

[Nailer]

Take the Linux kernel updates out of the Linux OS topic? If you do that, then you'd have to rename the Linux OS topic to "GNU-based OS" topic because the only thing that makes Linux Linux is the kernel.

Yeah, the kernel. And all the drivers within that kernel that only Linxu supports, like filesytems and packet filters. And the FHS, and the LSB and the standards they entail, like SysV and RPM. And all those the distributions that don't distribute other OSs (everyone except Debian). And all the software ports that are for Linux and not for any other Unix. And the various political issues which surround Linux and not BSD and other OSes.

Yeah, there's no Linux OS or Linxu specific issues. If that were the case, you'd have crazy stuff for each individual Unix OS, like a BSD section or apple.slashdot.org, and we all know that would never happen, right?

Its about signal noise ratio. You don't seem to understand that.

Re:Just give me a kernel that turns off my compute

[saintlupus]

The only kernel that's ever actually powered down my machine on shutdown or halt is the special 2.4.8 included with my Mandrake distro.

Has anyone else had this problem and actually fixed it??

Hasn't ever been a problem for me, using 2.2.19 or 2.4.10 - but then, I'm using a PPC system, so YMMV.

--saint

Re:this is an enterprise ready os?

[SirNAOF]

It is somewhat stable, but how can you replace a major part of the OS in the middle of a "stable" release tree? Replacing the VM in the middle of a tree is something that happens in development, not in production. The only real exception is if there is a major flaw in the VM that cannot be corrected without major rewrites. And in that case, they shouldn't release new kernels with major bugs in the first place.

But that's just one man's opinion.

Very upset about the RC4 error.

[Zeio]

I tried to reason on an IRC channel where marcelo (and other kernel 'hackers') hang out. I was kiboshed. I tried to convince them that fixing the tarball called 2.4.18.tar.gz would be a good thing to do.

Here is some of my reasoning, musings and retorts to those who 'know' more than I do.

"Is there a plan to fix the 2.4.18.tar.gz or will I have to patch it. It is really annoying if this isn't going to be fixed to rc4/final, instead I have to patch 2.4.17 with the RC4 patch. This makes it difficult to use kernel.org as a "library". Pretend in some number of months some Joe Schmoe says 'Gee, 2.4.18 has been out for a while and is considered stable,' downloads it, and misses the RC4 patch."

This was rejected as reasonable. I was told that assuming a release is stable is bad practice, particularly based on how long its been out. My impression was this was the stable branch. I'm sure that, for example, RedHat picked 2.4.7 and 2.4.9 and hacked them for their own distributions for some reason or another. They, like the rest of use, should be ensured that what is fixed in the changelog should be included in a given release. I don't like being shunned for being closer to correct.

"I appreciate the need for a releases in software. The line is drawn, certain things are in, other are out. Its just that what was determined to be final and what is being masqueraded as such are two different things. so, the gatekeeper in this case should be able to rectify the mistake."

From the group came no response. The conversation had turned to more pressing things, such as people bragging about compiling XFS into highly experimental versions of the linux kernel. Proper release procedure is not nearly as important as strutting about having XFS working in a situation where it probably shouldn't.

Here is a reply, which was well stated and polite, but I vehemently disagree with:
"Zeio: You and Marcelo both :) everyone would have liked this to be a better release than it was, but ... mistakes happen. and, once published, one must live with them. :)"

So, with this reasoning, if I published a book. For the sake of argument this book is supposed to have 10,000 copies printed. I catch a typo after shipping 1,000. Wow, the rest of the 9,000 people have to eat the typo because once something is released it shouldn't be changed.

I also state this:
"I'm suggesting a viable way of dealing with it[the mis-release], to fix the problem by putting what was supposed to be final in place of the tarball which masquerades itself as a release, or rename it to DONT USE like 2.4.11. I would expect higher standards from the linux maintainer.

Finally, to my dismay, I realize that there is denigration concerning the theory that and open community should be attempting to mold the linux kernel tree into a pillar of perfection. Lines have to be drawn, periodic shortcomings have to be accepted until fixed, but gross errors which are easily fixable should be ignored because 2.4.19 is on the way. I'll lower my expectations of the "stable" 2.4 linux tree for the time being. I'll put 2.4.18 in the same category as 2.4.11 and the "greased turkey" release. Seems this is becoming a norm. I strongly appose nonchalant and half assed attitudes towards maintain something of this importance.

Another joke was made that this only gravely affected SPARC users. This reinforces the wholly incorrect attitude that x86 should come before others. I'd bet that if this 2.4.18 wouldn't boot on x86, they would have re-released it.

Sadly, I had to point out that even Mickey$oft was forced to re-release service pack 6 to 6A. The claim was that 2.4.19pre1 is already out, and that 2.4.19 will likely be out in less time that it took Mickeysoft to put out 6A. These to me are excuses. Inferior ones. I expect more from linux than Microsoft. I expect a group project to put its best foot forward. I'd hate to have to write code for a project where FINAL is a line that is arbitrarily drawn. I know I'm over reacting, but I tend to like testing the latest stable release when they come out, and wouldn't you know it I have a SPARC. Guess I'll wait for 2.4.19, like I had to wait after 2.4.11 (2.4.12 was out soon, albeit with a broken LPT module, but that is when Linus maintained 2.4) and greased turkey. Sorry, I don't like to patch a previous major release, I just don't like doing it, I don't get off on it, I don't want the hassle, even though it is easy and have done it for things like the AIC driver when it was taking them forever to integrate the changes into the stable tree.

Linus, show this kid how to rectify an error and do it quickly.

Re:Very upset about the RC4 error.

[the Atomic Rabbit]

I'm suggesting a viable way of dealing with it[the mis-release], to fix the problem by putting what was supposed to be final in place of the tarball which masquerades itself as a release, or rename it to DONT USE like 2.4.11.

The trouble with changing the 2.4.18 tarball on kernel.org is that having two different releases called 2.4.18 out there will lead to more confusion than it's worth. This is particularly important because kernel prepatches are distributed as diffs against the releases.

Linus, show this kid how to rectify an error and do it quickly.

Hmm... you know what? I think I see the reason you were "kiboshed". (Hint: it's not because of Marcelo's immaturity.)

Re:Very upset about the RC4 error.

[bankman]

So, with this reasoning, if I published a book. For the sake of argument this book is supposed to have 10,000 copies printed. I catch a typo after shipping 1,000. Wow, the rest of the 9,000 people have to eat the typo because once something is released it shouldn't be changed.

Well, actually you are proving his point with this statement: Just because there is a typo in the first edition of a book doesn't mean that the whole edition gets recycled (we don't burn or throw them anymore). The typo will be fixed in the next edition, like is being done here.

You definitely don't want two different versions with the same name, especially not in software. This would really confuse users.

BTW, people installing a brand new (i.e. just released) version of any software on production systems should be sacked right away. Who, with any interest in Linux by now doesn't know about this feature missing in 2.4.18. Everybody, especially users with non-x86 systems, should read the ChangeLogs and/or release notes very carefully. If they don't, well yet another reason to sack them right away for irresponsible behaviour.

Alright, now one might argue that Linux will never succeed in the home user segment with kernel developers being that careless. How many home users are there with SPARC systems? No, the average Linux home user doesn't run these kind of machines, it's mostly geeks who do read the ChangeLog and/or follow /. and maybe even the kernel mailing list.

People desperately in need for the latest, greatest kernel features on SPARC systems will just have to download 2.4.19pre1 or 2.4.18pre4. So, what exactly is problem? Damn, you're just a troll and I fell for it. Stupid me....

Support for announcing new kernels

[herk]

Yeah that's right, I said it. I LIKE having Slashdot announce new kernels. This is where I always hear about new releases. I don't want to check a kernel site every day of the year, I'd rather see it on a site I'm reading anyhow.

Too many people bitching about such pointless dribble; 2.4 sucks, BSD owns Linux, stop posting these kernel releases.(Despite the fact that it's clearly geek news, and being posted on a geek news site) And then we add capability to exclude topics from your slashdot homepage, and people still bitch.

This is a tech news site, Linux kernels are a perfectly viable news item. 2.4 does not suck. If you think it does, move on to something else. Ignore the topics. Stop ripping up people doing a perfectly good job.

Re:Allow me

[SirNAOF]

Maybe it's because Windows has Windows Update. It pops up that nice little reminder that there are critical upates that you should have for your machine. So everyone knows about them.

But as for linux news, specifically kernel releases, slashdot seems to be the place to find out. Perhaps someone should write a quick script to check for kernel updates and have it mail root@localhost or something. In fact, that sounds kinda fun.

Re:New SPARC kit? Move along there's nothing to se

[wysoft]

The non-pagable kernel memory used to fit, just about, in 32MB with some to spare for buffer cache (well, 2.5.1 did) . Nowadays it just swaps horribly.

I have an older SS5-110 with 64MB of memory and a couple of 2GB disks installed. Solaris 8 runs quite well on this old machine after some trimming of services, and it is not too hard to fit a basic installation with the necessary tools to install gcc, gmake, and start compiling your own software. The Solaris 9 beta (yes, version 9 STILL supports sun4m) runs even faster than Solaris 8. To be honest, I have experienced more swapping in Sparc Linux than in Solaris, however Linux is still a bit faster than Solaris on this machine.

I would probably be using NetBSD on it if they would finally get 24-bit color support for the TCX framebuffer.

Re:Please seperate Linux kernel from Linux OS topi

[herk]

God forbid anyone interested in Linux but not in a kernel updates be forced to glance at a news heading they didn't care about. Maybe we could subdivide even further, so people who are interested in Linux 2.5 kernel releases won't have to hear about 2.4 releases. What about those 2.2 releases?

Nitpicking sucks.

Re:this is an enterprise ready os?

[Bert64]

OpenBSD is not the world`s most secure OS, rather OpenVMS is, since when did you see an advisory for it on bugtraq?

Secondly, stability of openbsd is poor on some cheaper hardware (the same hardware works correctly with linux and freebsd) and proved itself highly unstable on a sparc machine (which since has been running solaris 7 reliably) openbsd also has no support for SMP, afaik on any architecture.

and finally, the supposed security is over hyped, the "no remote exploits in default install" statement comes from a minimalist default install, i`m sure there are linux distributions like this, windows 9x has no remote exploits by default either. Plus the claims of strong security give people a false sense of security, "ohh the os is secure so it doesnt matter if we install this vulnerable version of bind"

Re:Please seperate Linux kernel from Linux OS topi

[Nailer]

Difference being that there's enough of the former people to care about it, which is why it keeps coming up. Ranting about how other people dare ask for a choice sucks.

Re:Please seperate Linux kernel from Linux OS topi

[herk]

If developers caved to everyone requesting further subdivisions of topics that really have no business being subdivided, then they soon find themselves with every article being a topic itself. It's foolishness. Linux kernels are quite clearly a perfect candidate for posting under a Linux topic. A further Linux / Linux kernel release subdivision is just stupid, especially to appease a select few people only out to be difficult. Glancing over a kernel release you don't care about is NO BIG DEAL.

Re:this is an enterprise ready os?

[Woko]

Exactly. 4-5 people is hardly "Enterprise". When you've got your boss screaming and systems supporting tens to hundreds of thousands of users, you might be taken seriously.

Re:Please seperate Linux kernel from Linux OS topi

[Nailer]

Linux kernels are quite clearly a perfect candidate for posting under a Linux topic.

I disagree, as installing (and perhaps compiling) new kernels as soon as they released is clearly not a part of running a Linux system. Certainly not for its main use, servers. Most tech enthusiasts also prefer the stability of well tested stable kernels and have better things to do than install new kernels every two weeks (which is when Slashdot reports them). Newbies don't know or care how and will stick to distro kernels, thank you very much. Don't get so antsy about offering people a choice. I DON'T REALLT THINK THE CAPS ARE NECESSARY EITHER.

Re:Please seperate Linux kernel from Linux OS topi

[herk]

(a) Compiling kernels is a part of running a Linux system, depending on how we chose to run one.
(b) Linux's main use being servers is certainly open for debate, I for one am an avid desktop user.
(c) Some of us enjoy participating in the testing and development of new kernels.
(d) We're not all newbies, and enthusiasts who have better things to do than install new kernels every two weeks hardly covers all the non-newbies.
(e) Don't get so antsy about glancing over a posting you don't like.
(f) Your thoughts on my caps are subjective and irrelevant.

Re:Please seperate Linux kernel from Linux OS topi

[Nailer]

(a) Compiling kernels is a part of running a Linux system, depending on how we chose to run one.

It hasn't been necessary for many users since modutils came into existence. More to the point, grabbing and installing the latest release of Linux kernels, even in the stable tree, isn't recommended for most users apart from tech folk who don't genrally care about stability.

(b) Linux's main use being servers is certainly open for debate, I for one am an avid desktop user.

That's great. Although hard numbers are hard to come by, most statistics show otherwise.

(c) Some of us enjoy participating in the testing and development of new kernels.

Thats great. You should even get your own topic for it!

(d) We're not all newbies, and enthusiasts who have better things to do than install new kernels every two weeks hardly covers all the non-newbies.

No it doesn't, thank god. Thanks for making my point for me. People who care about uptime (which most experiienced users do) don't chase the bleeding edge.

(e) Don't get so antsy about glancing over a posting you don't like.

I'm not antsy. I suggested a choice. I said please. I didn't sarcastically rant. I didn't say your option sucks, or that your idea were stupid, or OVERUSE CAPS. Some else did. Guess who?

(f) Your thoughts on my caps are subjective and irrelevant.

No, they're evidence of you being antsy. Chill out.

Is it really that bad?

[entrylevel]

I've been upgrading to the latest 2.4.18 kernel pretty much every time a release comes out. Now my architecture isn't the wierdest, it's a beige G3 with a USB/FireWire card, Rage Orion, serial camera, USB scanner, FireWire CD-RW, and a serial Palm V with a Keyspan USB adapter.

Guess what? Through every kernel upgrade, all this hardware just works...better than under OS X (some of it doesn't wok at all under OS X). And I've never had a problem with the VM, or anything else. I don't see how a little mistake like this amounts to a hill of beans. At least they told everyone.

Remember how MS was telling security experts not to talk about or even disclose security exploits until _after_ a patch was released? And how long do that take? Well, if MS gets their way, you have know way of knowing until your NT box gets hacked. And when will the patch come out? Better check that Windows Update every 6 hours! What's that? You have better things to do with your time? Then why did you update your kernel the moment it was released???

Slashdotters are really beginning to sound a lot like Mac users.

Re:Release frequency, release engineering thoughts

[BlowCat]

Don't get me wrong, I use Linux daily, my servers run on it, and I depend on a variety of other open source software (particularly Python!). I even buy RedHat/KRUD releases just so that some value flows back into the release process from a happy recipient.

It's very sad that you have to say it. It reminds me of communist regimes where every, say, philosopher had to mention in every article that his theory is consistent with teachings of Marx and Lenin.

Moderators are supposed to rate your comment, the way how you explain opinion, but not your position in the so-called OS War. Unfortunately, they do the opposite quite often. Now you have to be with one of the camps to be heard and understood.

I used to like this site.

Re:New SPARC kit? Move along there's nothing to se

[slashdot_commentator]

If the bloat of the newer Solaris OSs makes your old hardware perform like a beached whale, why not just run Solaris 2.6? Its still maintained and will run packages of its era. So what that the latest Oracle won't run (or run well) on it, or that your old machine won't execute 64 bit code? It wasn't meant to run the latest Oracle monster, and the hardware was never meant to run 64 bit code.

(Granted, running SPARC Linux may be more entertaining, and I still suspect it would run faster than any version of Solaris...)

Re:Release frequency, release engineering thoughts

[seantrue]

I was actually more saying "thanks". I don't pay for most of this software, and what I do pay is a laughable fraction of the cost of development and delivery. I figure that at at least I can say nice things.

This is a marketplace of ideas and technologies -- not a battlefield. And markets require the flow of capital, whether it's based on gold, the faith and credit of the government, on reputation, or on intellectual property.

For what it's worth, I'm not a fan of C++. (Whew, there, I said it!)

Re:New SPARC kit? Move along there's nothing to se

[faboo]

pardon me for not actually running a server, but linux supports (I believe) reasonably large number of processors (>32 IIRC) and arbitrarily large amounts of memory. Granted, I have no experience using these extensions (my system is a work station), but they _are_ there, and they at least injure me none (as there is no requirement that I use them).
but that is that.

Well, if you're going to be that way about it...

[jdavidb]

... then you could've mentioned the OS that first brought UNIX to my home - starting on a Zip drive in a Mac LCIII: NetBSD.

I left the BSD world for Linux over a year ago, but I still get tired of seeing the world's most portable OS left out.

For some strange reason this message will probably be followed by trolls that sound a lot like {Net,*}BSD is dying. Ignore them.

modprobe apm?

[fferreres]

If it fails, make sure you are compiling it, for APM works perfectly. Play with the APM kernel options for buggy motherboards if nothing works.

Re: plz mod the troll down

[fferreres]

"instead I have to patch 2.4.17 with the RC4 patch"

Just download the .rc3 or 2.4.19-pre1

"I was told that assuming a release is stable ... based on how long its been out."

Take the advice, it's a good one. You never know which kernel is best until time has passed. After some time, you will know which kernel is best. That doesn't mean EACH release is stable (ie: perfect).

"I expect more from linux than Microsoft"

Don't expect, contribute. You sure contribute to Microsoft but expect from Linux.

"Linus, show this kid how to rectify an error and do it quickly."

WHAT?

Not fixed

[Sits]

Are you sure it is fixed at all? How could this have been fixed in 2.4.18-pre1 when the bug was first discovered when 2.4.18-pre7 was out?

I think I'll keep mem=nopentium until someone can point me to a changelog entry that mentions this directly.

Re:Just give me a kernel that turns off my compute

[jhanson]

Halt isn't supposed to power down your machine. Try calling 'poweroff' of using halt -p.

Re:Just give me a kernel that turns off my compute

[Adrian Voinea]

Yep... just turn on APM this way:

make menuconfig
General Setup -> Power Management support -> Advanced Power Management BIOS support -> Use real mode APM BIOS call to power off

This should fix it :)

Re:Please seperate Linux kernel from Linux OS topi

[herk]

I'm not saying compiling kernels is necessary, but it's of interest to ALOT of users. I'm not calling you or anyone else stupid. I'm not OVERUSING CAPS. I'm talking about excessive whining, which has to be discouraging for those running these forums, and is completely unproductive to the community as a whole.

If people want Linux news, there's no reason to ask them again if they want Linux KERNEL news. You seem to continuously glaze over that point and focus more on my use of caps.

If you're interested in Linux, subscribe to the topic. If you're not, don't. If you're interested in Linux, but can't possibly bear to glance over Linux kernel release news, well then you'll have to make this ever so difficult decision on your own.

Re:Please seperate Linux kernel from Linux OS topi

[madenosine]

Yes; have the penguin mean GNU/Linux, and then have a seperate image/topic for the Linux kernel.

Obviously, many people are tired of hearing about Linux kernel news, and wish they could filter it out (myself not incuded.) Is that not what a large point of seperate topics is?

Doing something like this is very simple for the /. editors.

a little excercise

[glwtta]

I thought I'd help out the /. community by providing a little practice for those people who are not interested in reading kernel release announcements but seem to still have trouble not clicking on the links.

Below are four links - three are of interest to most people here, one probably is not. Try going through the list and ONLY following the links you like (repeat until you can do it almost every time), have fun!

1. Slashdot
2. [H]ard|OCP
3. G0atse Guy
4. Kernel.org

How did you do? Just keep practicing with this list, and you'll be able to enjoy slashdot to the fullest in no time at all!

Re:this is an enterprise ready os?

[Dikarika]

That is so damn funny, it made my day...

Especially when you think of the BSOD reflecting off Picard's shiny bald head... :)

Re:this is an enterprise ready os?

[Svenne]

> For those of you who haven't looked into it, check out OpenBSD's *AMAZING* ability to be an invisible bridge (not only can it invisibly bridge, but it can filter the bridge as well, which is something other OS'es have yet to be able to do).

I used linux to set up a bridge and then used ethereal to sniff the virtual bridge interface several months ago. Worked like a charm, yet the machine was impossible to detect. I have no idea how long bridging has been in the Linux kernel, but OpenBSD certainly isn't the only OS that's able to sniff traffic while bridging.

Blast!!!

[phagstrom]

Does this mean that I have to buy another t-shirt with the Linux kernel code printed on it.

Re:this is an enterprise ready os?

[J'raxis]

Because it ran 100% stably under 2.4.2, and had some problems (but at least did not randomly reboot) under 2.4.9. I upgrade the kernel and it starts acting odd. I reboot with an older kernel and it stops acting odd.

Re:this is an enterprise ready os?

[J'raxis]

Well, it never did this under other kernels; in fact it stayed up for 2½ months under 2.4.2 it was finally taken out by a power failure. Im sure the box has problems, as it is a hodgepodge of old hardware (still has a 5¼ drive!), but the random-reboot problem only began under 2.4.17. Rebooting under an older 2.4.* brought it under control again. I didnt take the time to investigate exactly what was wrong with it.

In general, I dont like using packages due to the opaqueness of everything (type RPM -U and just hope it quietly installs hundreds of files in different places?); and am currently working on building a system from the ground up on the machine.

Re:this is an enterprise ready os?

[Balp]

The Amiga OS is as secure as OpenVMS, or rater much more secure Bugtraq has 3 for VMS and only one mention of Amiga (and thats is in an unreleted comment part of the patch code). But thats not a good whay to compare systems at all.

The last part of your comment seams to be a loot uninformed. Yes not every service is activeted ion a default OpenBSD installation, but one common discusition of the OpenBSD mailing lista are why the **** is that service activeted for? It's not needed and the anser usally is that the system has to be usable to the most users.

When it comes to bind the 4.x version in OpenBSD probably is a loot more secure that any other bind version installed anyware, OpenBSD faq has a small note about bind and dns (6.8.3.2.1 on http://www.openbsd.org/faq/faq6.html#DNS).

/ Balp

Re:this is an enterprise ready os?

[iamriley]

If your mobo has a removable cache chip (very common in the time period that I'm guessing your board is from), try removing it. If it's not removable, check your BIOS for an option to diable it. This *might* fix the random rebooting problem.

From the tech experience that I have, a machine that reboots randomly generally has a bad cache. If this is the problem, I'm at a loss to explain why it would show up in kernel 2.4.17. Anyway, it's just an idea.

Oh, and I haven't tried Linux From Scratch, but I use Sorcerer GNU/Linux, and even though I think the spell metaphor is cheesy, my system (475 AMD K6-2, 512MB RAM) is much faster than it is with any of the several binary distos I've tried. I'd be interested to see an in-depth article comparing the "compile everything" distros.

Waaay friggin' off topic, but hell, its just karma

[drik00]

~ This ole porch is just a long time,

waitin' and forgettin',

Rememberin' the comin' back, not cryin' bout the leavin'

and rememberin' the fallin' down and the laughter,

and the curse of luck of all those sons'a'bitches,

who said we'd never get back up... ~

--- more REK, Jr. just funny to see another fan on /., of all places...

Re: plz Censor and Opinion, Mr Marx doesnt agree.

[Zeio]

Hardly a troll, it was my opinion. I'm sorry if my take on things doesn't match yours.

I am disgusted that my take on this situation was modded down as flamebait.

Re:Please seperate Linux kernel from Linux OS topi

[Nailer]

I'm not calling you or anyone else stupid. I'm not OVERUSING CAPS.

Yeah, you never did any of those thigns. While we're dreaming, I'd also like a pony.

Its something a lot of people want, its not much work, and you're just complaining about my `whining' and insulting me because you can. I won't bother wasting any more time.

Re:Please seperate Linux kernel from Linux OS topi

[herk]

I like how you speak in the plural sense, as if you have someone siding with you.

Re:Please seperate Linux kernel from Linux OS topi

[Nailer]

We = you and me, you dolt :)

Re:Block anonymous Cowards?

[dnoyeb]

Most of the time they post meaningless stuff. I dont want to cencor them, I personally dont want to read what they have to say. Either you stand behind what you say, or beat it. Im not going to argue with some anonymous coward. Thats as bad as "my contacts" or "unidentified sources."

Your free to post as much as you want. Im just trying to filter through the 200+ posts that seem to show up, and the easiest way seems to eliminate the AC first. Their cowards anyway.

Linux is not an Operating System

[infernalC]

There is nothing to Linux except for the Kernel, so what the hell are you talking about? Linux is the kernel. The operating system is GNU, and together they are called GNU/Linux. No, no, you can't just call GNU the Linux OS, because it runs on top of other kernels, too, like Mach. It is the fault of the distro companies that folks don't know this. It would be nice if RedHat, Caldera and JoBobDistroCo all put prominent literature in their distros and on their web pages pointing to the FSF, but the companies, being capitalist entitities, are to a larger extent in it for the dough and thus fail to amplify the communist spirit of open source.

Yes it is

[Nailer]

Whether people choose to call the OS Linux or GNU is a personal choice, but there is definitely something specific about Linux.

• Standards like the Linux Standard Base, including the standards it entails like the FHS, RPM, and SysV
  
• Distributions that don't exist for other Unixes
  
• Particular combinations of kernel, versions of compilers, c libraries, window systems, etc that aren't found on other Unixes and are generally tested as eachother.
  
• Applications which aren't universally available on other Unixlike OS, such as Freeswan, or Opera, or Lotus Domino.
  
• Politics that don't exist for other systems
  

Compare the contents of Linux.org and Kernel.org or simply watch most Linux users, including the experiences ones, speak to realize there's a lot of people who call that OS Linux (because there's far more people responsible for its success than the FSF, including the BSD folk, XFree86, etc).

Most Open Source advocactes are capitalists, including Eric Raymond. So are most Free Software advocates. Why would you think otherwise?

That's not what I meant

[infernalC]

I did not mean to imply that free software advocates are not capitalists... I only mean that open source projects like GNU require a spirit of communism to be successful. If nobody did any work for the good of the community, nothing would get done.

I think that those of us who do contribute of our time and coding skill rise above our capitalist selfishness in a way, because we do contribute for the benifit of a commune: that of the computer-using proletariat.

Re:Just give me a kernel that turns off my compute

[Frank T. Lofaro Jr.]

Power down? Why would you ever want to do that? ;)