back to article Linux 3.19 released for your computing pleasure

Version 3.19 of the Linux kernel has been signed off by Linus Torvalds. News of the release emerged in a typically economical Sunday evening post to the Linux Kernel Mailing List, in which Torvalds noted there are still a couple of bugs in this release but they were pretty obscure so “... while I was tempted a couple of times …

  1. Anonymous Coward
    Anonymous Coward

    I miss the old days

    When version numbers were not just arbitrarily assigned by developers. We used to be able to work out how big the changes were from the numbers. If you're not going to use major, minor and build numbers properly then why not just use a single number and increment by one each time?

    1. Robert E A Harvey

      Re: I miss the old days

      >why not just use a single number

      Firefox does that, and I still havn't adjusted

    2. Nigel 11

      Re: I miss the old days

      We used to be able to work out how big the changes were from the numbers

      That may work OK for an app. It won't work for a general-purpose operating system. The thing that is most likely to be immediately visible to a non-technical user is a change in the user API causing existing programs to break (which is something that Linux tries very, very hard to avoid). The next most likely is a new bug in a facility that you are using, but that's hardly something that they wanted to ship!

      Apart from this, who decides what is a big change? A complete re-working of the code for massively parallel SMP systems may be scarcely visible to a person with a single 4-core CPU (and even less visible to someone working with a single-core embedded peabrain). A new filesystem ( for example Btrfs) may be of huge interest to some, and of no relevance whatsoever to others that aren't intending to use it. And so on.

      The switch from Linux 2.x to Linux 3.x was supposedly arbitrary, but did in fact coincide with a major architectural change that the kernel debelopers had been working towards for the best part of a decade. What do you mean, you didn't notice the final demise of the Big Kernel Lock? Well, actually, you weren't supposed to. Its removal was a success. Cause for celebration by kernel developers (and maybe, the reason for the big version number change), but a big yawn for everyone else.

      So, is there any reason for Linux kernel going from 3.x to 4.0 other than (maybe) the next release after 3.99? Well, just maybe ...

      1.x ... a developer / enthusiast system

      2.x ... production-ready, large scale SMP handicapped by big Kernel lock

      3.x ... big Kernel lock finally gone, scales from embedded peabrains up to huge datacenters.

      What next? I'm hoping for

      4.x ... Microsoft abandons its own OS kernel, adopts Linux.

      1. Oninoshiko

        Re: I miss the old days

        That's okay, Linux isn't a general purpose operating system. It's a kernel.

        In order to qualify as a OS it would have to include more. At least as much as solaris' OS/Net consolidation.

        1. wayward4now

          Re: I miss the old days

          "That's okay, Linux isn't a general purpose operating system. It's a kernel."

          Depends on who you listen to. Here's a quote from "The design of the unix operating system", Maurice J. Bach, Prentice/Hall, 1986, page 4:

          "The operating system interacts directly with the hardware, providing common services to programs and insulating them from hardware idiosyncrasies. Viewing the system as a set of layers, the operating system is commonly called the system kernel, or just the kernel, emphasizing its isolation from user programs. Because programs are independent of the underlying hardware, it is easy to move them between UNIX systems running on different hardware if the programs do not make assumptions about the underlying hardware."

          Ergo, Linux is an OS.

          Stolen shamelessly from "http://linux.topology.org/lingl.html"

      2. Fatman

        Re: I miss the old days

        What next? I'm hoping for

        4.x ... Microsoft abandons its own OS kernel, adopts Linux.

        Choose any common expression of something that has a probability of less than 1 in 100 billion, and you have the chance of that happening. Just being realistic, here.

        1. JamesTQuirk

          Re: I miss the old days

          Yeah, they will just use it for Azurb Clod Servers & Business, & SELL/Rent YOU Windows ....

        2. admiraljkb

          Re: I miss the old days

          ">>4.x ... Microsoft abandons its own OS kernel, adopts Linux.

          >Choose any common expression of something that has a probability of less than 1 in 100 billion, and you have the chance of that happening. Just being realistic, here."

          People don't buy Windows Servers for the kernel specifically, they buy it for the technologies that run on it. Same with Windows desktops. Remember, the kernel portion of Windows (and related OS areas) is a major cost center for them, not a profit center. MS sells applications, development tools, server software, and gui's across servers and desktops, not kernels. :)

          I agree that MS wouldn't adopt the Linux kernel (licensing issues), but I do suspect they will adopt BSD and follow in Apple's path, and probably sooner than later. This would NEVER happen with Ballmer, but with Satya Nadella in charge it becomes a real possibility. After having done OEM Windows OS development work I can tell you underpinnings of the OS (ala interactions with several craptons of various hardware) is expensive to develop for and even more expensive to test for. In my case that was with MS already having done the most expensive heavy lifting under what my team was doing. Why not "outsource" the lowest levels which allows Microsoft to then concentrate on the MS technologies that actually make their money - AD, Exchange, SQL, IIS, S!@#epoint, etc?

    3. bazza Silver badge

      Re: I miss the old days

      I think that the PREEMPT_RT patch set should be folded in to the mainstream kernel. That'd be worthy of a 4.0 version number.

  2. frank ly

    Good enough - ship it

    "... there are still a couple of bugs in this release but they were pretty obscure so “... while I was tempted a couple of times to do an rc8, there really wasn't any reason for it.” "

    Is there a desperate need to meet a deadline? Does a major contract demand delivery?

    1. fishman

      Re: Good enough - ship it

      If you waited to get all the obscure bugs out, it would never get released.

      1. admiraljkb

        @fishman Re: Good enough - ship it

        "If you waited to get all the obscure bugs out, it would never get released."

        Software never releases - it ESCAPES!

        (line I've used for a while in final bug review meetings before a software release when evaluating defect severity's)

    2. Anonymous Coward
      Anonymous Coward

      Re: Good enough - ship it

      Most of the remaining bugs won't affect the majority of users so there's no point holding it up.

      Would you hold up the release for something like a bug in the m68k nommu port that is used by probably 10 people max for example? I don't think so.

      1. JamesTQuirk

        Re: Good enough - ship it

        As somebody who still runs a older Debian on a 68060 amiga 2000 w/ 130Meg ram, it may be ...

        (Still my Best Internet Box, not many Virus Care ...)

        But probably better for my RayTracing Habit's, if it improves CPU threading Etc, as nowdays I have a "renderfarm" in a box and not the 8 Amiga 2000's I used to use...

  3. h4rm0ny

    If there's no reason to go to 4.0...

    ...then don't go to 4.0. Major version numbers are for significant changes. If this is a collection of further small refinements, 3.20 is fine. It's actually a sign of a mature product.

    1. Lusty

      Re: If there's no reason to go to 4.0...

      It was the sign of a mature product, but considering the move to 3.0 was pure whimsy on the part of Torvalds there is no way to tell any more. This is especially true when he says he released knowing there were bugs which he didn't deem important. At least Microsoft have the decency to pretend they don't think there are bugs!

      1. Anonymous Coward
        Anonymous Coward

        Re: If there's no reason to go to 4.0...

        > he says he released knowing there were bugs which he didn't deem important.

        Lusty, I do not know what the extent and responsibility of your project management experience is, but I know Torvald's and his team.

  4. Dazed and Confused

    KVM on IA64

    I didn't think there were any drivers for the more recent IA64 systems anyway, so IA64 Linux boxes are probably still stuff at the Montvale level. I've not seen it on any of the 4 or 8 core chips.

  5. Steve Graham

    "Never install x.0 of anything." Or in this case, x.x.0

    My current kernel builds are 3.18.1 (3.18.6 is supposed to be the most recent "stable" version).

  6. Anonymous Coward
    Anonymous Coward

    device drivers

    Just a few obscure bugs? Does he count device drivers or just the pure kernel? After all, the device drivers are shipped as part of the kernel tar file. Yet there are persistent and not obscure bugs which never get fixed.

    1. gerryg

      Re: device drivers

      Good to get some retro trolling every now and again, reminds me of the old days. Is your analogue Wintel modem still giving you problems?

    2. Anonymous Coward
      Anonymous Coward

      Re: device drivers

      > Yet there are persistent and not obscure bugs which never get fixed.

      Is your IDE of choice out of order? Not enough funds to pay someone to do the job? Send it back and apply for a refund then.

  7. dogged
    IT Angle

    Needs an article by Gavin Clarke telling us that it's shit on the basis of, er, nothing.

  8. Anonymous Coward
    Anonymous Coward

    Support for keyboard backlights needs to be in the kernel? Hmm, principle of least privilege seems to have passed the Linux boys by!

    1. Anonymous Coward
      Anonymous Coward

      > Support for keyboard backlights needs to be in the kernel? Hmm, principle of least privilege seems to have passed the Linux boys by!

      This way to the LKML. Send a patch and show them how it's supposed to be done.

      Incidentally, as a basic Unix user I only ever wrote one device driver and in my case the device in question had been engineered really well, so that I could do a decent job myself on the driver side (and would have had no excuse for not doing so). Looking at many other devices though, they do not exactly leave the driver writers with a lot of good choices.

    2. Charles 9

      Trouble is that a device driver by necessity has to go to the metal (as they are by definition the interface between the software and the hardware). Least privilege in this case IS the kernel, which should be the only thing able to go to the metal. This is why device drivers have always been a sticking point regardless of the OS.

  9. ricardian

    Linux 3.19? Linux 4.4 surely

    http://www.theregister.co.uk/2016/01/11/linux_four_point_four_released/

    1. gazthejourno (Written by Reg staff)

      The story you're commenting on is almost a year old. Hope this helps.

POST COMMENT House rules

Not a member of The Register? Create a new account here.

  • Enter your comment

  • Add an icon

Anonymous cowards cannot choose their icon

Other stories you might like