Slashdot Mirror
Linus Torvalds Speaks Out on Future of Linux
SlinkySausage writes "Linus Torvalds has laid out his plans for the future of Linux, including the 3.0 kernel [there probably won't be one], problems with the Linux release cycles and which distro he personally runs on his home PC. '"Compile everything by hand" ones simply weren't interesting to me,' Torvalds says."
The development of something like a kernel NEEDS a dictator, and if Linus walks away, who is going to have the credibility and/or authority to keep a handle on it? I worry that an advisory 'board' or 'panel' would be the death by a thousand cuts that could really mess up kernel development for linux.
This is a sig. It is like every other sig in the world, except that it is mine, and it is different.
it's more like going out to dinner at a sit-down restraunt, rather than a fast food restraunt.
All it is is one command per app install (or less, if one app requires other apps)
ex, if you want to play boson:
Gentoo$ emerge boson (compiles and installs boson, with any cooking instructions you have in your make.conf)
Debian$ apt-get boson (installs precompiled boson, straight from the wrapper under the heat light)
Same amount of work really... You just have more options available in Gentoo
34486853790
Connection too slow for X forwarding? Try "ssh -CX user@host"
Maybe we could get Con Kolivas to take over kernel stewardship?
Do daemons dream of electric sleep()?
Please stop perpetuating the myth that Gentoo == Linux From Scratch
/usr/ports insipred Portage...
With Gentoo you can start from a stage 3 install, and you can also install binary packages if you so choose
Why doesn't FreeBSD have the stigma Gentoo does?
Here's to the crazy ones
Sounds like he is sticking with the programming model of doing a large number of releases with small changes type model. Glad to see it actually, as this is the approach that I have been using on all of the software I build for work. What this does bring up, though, is the unfilled need currently of having an auto-upgrader software package where new kernel packages can be auto-upgraded and then migrated too on the fly without requiring a reboot. This would be quite complex I would admit, and maybe not possible in all kernel releases, but this is definitely something that needs to be looked at... Just my 2 cents worth.. :-)
There is a lot of new stuff happening, but it's in the main not specific to the kernel. New things the kernels needs to do are thin on the ground now. Not to say it'll ever be finished as such, just that there aren't any needed big new features. It'll take a major new shift in computing to do that, I suspect. Something way bigger than extensions or tweaks to x86/SPARC/PPC/ARM etc. I'm not holding my breath.
I may be stating the obvious, but the site is slashdotted, so I can't see what Linus has apparently said.
Run a compiled KDE/OpenOffice system from gentoo with the appropriate flags for your CPU in make.conf
Compare the performance to the pre-compiled Gentoo, Fedora, or Ubuntu performance
The taste (err, performance) is a lot better with the compiled yourself. And you don't get asked 40-50 questions, or if you do, you forgot to set batch mode.
I use FreeBSD, with a build system similar to Gentoo, and I have two steps more than what I would get with an apt-get situation.
(1) add "CPUTYPE=[whatever-my-cpu-is-here]" to my make.conf file
(2) type "export BATCH=yes" if I am going to build anything with a given terminal (just once, not for each build), or add "--batch" to portupgrade. It builds automatically, and I don't have to answer any questions.
Now, if I want to change options, I can quite trivially do so, but beyond those two steps, and the time it takes your computer to complete the process, there is no difference between apt-getting a package, and "hand building" a package in a FreeBSD system. I know Gentoo has a parallel to the second step, but I don't remember it, the first step is the same I believe.
34486853790
Connection too slow for X forwarding? Try "ssh -CX user@host"
Personally, my favorite aspect of this interview is that Linus himself basically considers the core OS now stable enough that anything new is likely to be incremental instead of ground-shakingly different, at least on the x86 platforms.
Which would imply a stability that leads to dependability which leads to usability which leads to widespread use. At least that is my hope in the enterprise, that the combination of commodity hardware with a commodity, high powered and stable OS can be coupled with increasingly powerful database engines such as mySQL, Veritas, etc. Oracle on Linux is now considered stable as well.
At home? stability leading to dependability leading to integration leading to crossover applications that will no longer depend on a proprietary OS stack to function. The only thing missing from my desired tool set on Linux right now is basically an easy to use, high powered MIDI to music recording and notation system -- and the pieces for all of that is already there -- it's my time to research and integrate the pieces that is in short supply.
I guess my point is that stability and upgradeability cause me to buy (several Linuxes and Win2K). Give me yet a large bulkier OS that doesn't really do much but add coolness (Vista or even XP) and I yawn.
...Open Source isn't the only answer -- but it's almost always a better value than the alternatives...
> (1) add "CPUTYPE=[whatever-my-cpu-is-here]" to my make.conf file
The problem is that most of the people don't know what their CPUTYPE is. I don't know it either and I have actually build the pc from parts on my own. Is it really impossible to autodetect the CPUTYPE?
I agree with what you're saying... my point is that since Linus shares his name with the product in question, I just thought his role as a kernel only developer kind of stifles adoption because his voice is left out of it. When he speaks, people listen -- and I just thought that his using his position in the industry to spur development, reconcile GUI options, unify, etc... would be a good idea.
:)
As good as a kernel developer as he may be, I still think he'd be valuable reconciling the problems that the different distros have and producing a product that is not only good on technical merits, but also other merits
I figured I'd get flack for the initial comment I posted, but I stand by it.
The price is always right if someone else is paying.
Not at all. These days if you buy something with Intel chipset, even hot off the press before the docs are online, there's a bloody good chance Linux will just work out the box. No tracking down weird drivers, beta out of tree patches. It just works.
Why? Because Intel can be bothered to give the devs the specs and not get their tits in a knot like ATI.
Happy days indeed.
In fact, it could be argued that the personality type that makes him so great at being the "benevolent dictator" for the kernel (strong opinions, detail oriented, more concerned with pragmatics than ideology) make him a terrible choice for spokesperson (makes enemies with strong comments, cares more about coding than marketing, doesn't care about large-scale ideology issues).
The fact is that Linus prefers being a technical guy building awesome technology, and doesn't really want the responsibility of guiding large-scale direction. Moreover, like many FOSS coders, he considers freedom of choice to be really important, and so has a general attitude of "do whatever works for you." Like I said, there's truth to your statements... however at the end of the day I think we're better off looking to people like Mark Shuttleworth as leaders of this movement.
Yeah, a coworker once said the same thing.
I convinced him to try it with just OpenOffice
He sang a different tune. Call bullshit all you want, but I've tried both, and I know which I prefer.
And I never once said that this was a good idea for big companies. Please don't make assumptions that make you look idiotic, as a professional sysadmin whos avoided that mistake but seen others make it, it'll get you a pink slip just as easily.
It's the difference between 1/2 second to open OpenOffice 2.1 vs. 2+ seconds on one of my systems.
34486853790
Connection too slow for X forwarding? Try "ssh -CX user@host"
are you *sure* it was the compile flags that did that, and not some disk fragmenting or cache issues, or whatever else might have screwed your 'test'. Loading openoffice isn't exactly CPU intensive is it, and the compiled versions only have the CPU-specific extras enabled, so I can't see that having SSE3 etc enabled in your OO build v the pre-built one will cause that much of a slowdown.
So you create a bypass argument to compile for a different build than the current machine. It seems ass backwards to me. Default it to the current PC and if you need to change it, override it.
Every time I start to have faith in humanity, I ruin it by driving to work between 7 and 8 am.
I agree with the meat of your comment, but I don't entirely agree with the premise. The "smoothness" and relative performance increase I see on gentoo is primarily about CPU optimizations, which as you discovered are pretty negligible. I do however not have X support and Pango and OpenType and a million other unneeded libs linked to my applications as happens with the default packages on SUSE and Redhat. The result is applications which have a smaller memory footprint and don't require a shit ton of ancillary packages to be installed. This yields a real measurable benefit both in terms of performance and overall complexity of a given install. Just adding "USE= -ipv6 -X11 -multi-lib -doc" to a vanilla gentoo minimal build destined to be a server can shave a considerable amount of overhead from the finished product.
-*The above statement is printed entirely on recycled electrons*-
Seriously, 1.0 was considered "feature complete" at the time of release, and there are some major architectural changes which will be required in order to improve scalability across multi-core as well as SMP systems, not to mention some fairly major pieces of work that are still under development which will need to be merged fully at some point (DAPL being one of the bigger). With the growth in the cluster market, I would also expect some meta-structure to go in to support the basic concepts. Even PCI-e 2.1 support is going to have a serious impact, due to the changes introduced in it.
If I was in Linus' shoes, I'd be pushing for these big infrastructure components to be readied and maybe placed in the -mm tree at this point. Once they're ready for the big time - which might take a while - I'd migrate them into the main tree and wait three or four cycles for last-minute bugs to settle down, then flip to 3.0 to mark the first of a generation of kernels that are keeping pace with the curve. I'd reserve 4.0 for when Linux is not only stable for mainstream use but defines the curve for OS development. I think everyone on Slashdot is at least aware of the research into new hardware technologies, new OS technologies and so on, so I don't think anyone seriously believes that Linux won't undergo more fundamental changes in its life.
Obviously, I am not Linus and he gets to do what he wants, whether I - or anyone else - would agree with his beliefs or not. However, his reluctance to flip digits is not new - I remember when kernels had a letter at the end to mark the sub-sub-version and it had to go into 2 letters because Linus ran out of alphabet. I also remember the first time sub-sub-version numbers ran into the hundreds. On both occasions, there was gigantic frustration with the absurdity. I guess he's forgotten the problems caused, or something.
It's a small world and it smells funny; I'd buy another if it wasn't for the money; Take back what I paid (SoM)
Which would imply a stability that leads to dependability which leads to usability which leads to widespread use. At least that is my hope in the enterprise, that the combination of commodity hardware with a commodity, high powered and stable OS can be coupled with increasingly powerful database engines such as mySQL, Veritas, etc. Oracle on Linux is now considered stable as well.
Depending on what kind of infrastructure you needed, six or seven years ago, you were fairly likely to get funny looks if you announced you were running a significant chunk of your servers on Linux. If you were running a significant chunk of your desktop infrastructure on Linux, the funny looks were a dead cert.
Three or four years ago, the funny looks regarding Linux servers were long gone, replaced with genuine interest.
Today, nobody bats an eyelid about server infrastructure, and you'd be just as likely (if not more so) to get genuine interest as funny looks if you are seriously migrating desktops to Linux. About the only thing you can't easily replace is Exchange and the centralised configuration UI that Active Directory gives you (no, LDAP user authentication doesn't count). The centralised configuration isn't too difficult to work around, the full integration of Exchange is.
Make and gcc have both improved alot since 2003 when gentoo came out. Today alot of improvements are automatically included and linked during runtime where before you had to hand config files or use flags to have the compiler do it manually. Portage really is not needed anymore.
http://saveie6.com/
One of the most interesting things (to me, anyway) that Linus talked about in the interview as how proud he was of the technical merits of the kernel and of Linux as an OS in general. I thought that was fairly interesting.
I really don't want to try and turn this into a Linux vs BSD vs [something else] flamewar here, but since I'm not really qualified to comment on things like memory-management algorithms, I wondered if anyone wanted to weigh in on exactly what areas they think Linux really excels at -- from a purely technical perspective. I really like the idea that Linus is getting at, namely that the real confirmation of open source is technical excellence, but I'm curious exactly what areas the Linux kernel is "The Best."
In particular I've always been interested in how some of the different open-source OSes handle different technical problems. Is how Linux handles (just for an example) memory management quantifiably better than how BSD does it? And if so (or not) why?
"Ladies and gentlemen, my killbot features Lotus Notes and a machine gun. It is the finest available."
Actually, I've put my daughter on Ubuntu.
She used Gentoo here at home, but Friday I install her in her freshman dorm room. Dad won't be around to maintain her system for her, and to be quite honest, Gentoo does require frequent low-level maintenance. I wanted her to have something she could *usually* handle entirely on her own, and in this case I figured the popularity of Ubuntu worked in her favor. She can push the update button, I've shown her how to get at synaptic, and a few other basics.
Now for the down side.
* Shortly after installation, she wanted Flash. Not in synaptic for amd64. I was able to find a Flash-on-Ubuntu-on-amd64 procedure with a bit of browsing, and got it done handily for her. But it wasn't a newbie type of thing.
* Sound only worked on one side. Furthermore, to her 18yo ears, the sound that was there had a high-pitched screech that annoyed her enough that she kept it turned off. Google was my friend, and now she's happy. But again, to a newbie it was smoke and magic - CLI even.
* Last night she wanted to play DVDs, and the installed default video player complained about missing plugins. Unfortunately it wouldn't tell me *what* was missing, just that *something* was. Google was my friend, but the first script failed, sending me back to google. Apparently I had to install a whole set of development tools, and there may have been one other hitch in there. At some point, it uninstalled some default media thing, too. The I *still* couldn't play the DVD. At this point, I just used synaptic to install xine-ui, and combined with the other stuff I'd done, she's watching movies. Using binaries was nice especially when it came to installing development tools, including gcc. But then again, Gentoo would have already had that stuff installed. But all in all, it was generally easier to get media working on Gentoo.
The living have better things to do than to continue hating the dead.
>So you're the kind of person that would hate http://www.gentoo.org/proj/en/gentoo-alt/bsd/fbsd/ Gentoo/FreeBSD then?
/usr/src. :)
Actually I was reading a little (nothing in great detail) about the alternate platforms for gentoo the other day, so you're not presenting something new to me.
BSD -Unix- is a system, developed and planned as a whole unit. Linux -and 'meta' operating systems such as Debian and Gentoo all suffer from a lack of co-ordination. Microsoft gets many things wrong, but their ad in '99 with the illustration of Linux as being an animal made out of the parts of different animals is actually pretty spot-on. The file utilities come from one developer, your archiving programs come from several different projects and your bootstrap utility is made by someone else entirely -and none of these people are on the same page.
In other words, most Linux distributions are comprised of random, seperately developed programs as opposed to whole systems where the person developing a new kernel api communicates back and forth with the people who are working on the corresponding userland code (whatever that may be in that instance, file utilities, disk utilities, whatever).
Then you have Gentoo -which, along with Linux from scratch- adds even more randomness to the mix with its' update mechanism. Update when the latest diskutilities fuck up X (or whatever) and your system is hosed.
The strength of Unix (BSD, Solaris) is that it is _designed_ -and by replacing the userland designed to work with the kernel with the usual disparate array of GNU/utilities you detract from the strenth of using a designed system, and are probably better off sticking to Linux.
Unless you're simply talking about splicing gentoo in the place of the usual third party delivery mechanism (gentoo instead of ports, gentoo instead of pkgsrc). That would actually be interesting, but I don't think that's what you're talking about.
>Suck it down my friend - Gentoo is the meta distro.
First, a correction is in order - Debian is "the" meta distro, and has been since 2000, if not before (they have had ports to the Freebsd, Netbsd and Hurd kernels in development since then -hell I've heard unconfirmed rumors of a cygwin port of Debian).
That said, this statement is why I think that you're not talking about replacing package delivery mechanisms but are instead talking about ripping apart stable, designed systems and replacing them with randomly bits and pieces of cobbled-together pieces from GNU.
Lastly, the claim that Gentoo is a development platform "for developers by developers" is rather amusing considering that no one touts it as a superior programming enviroment (at least, not outside of the gentoo forums, I wouldn't know about there -that's not someplace I have any reason to hang out at).
Instead you hear two claims from most Gentoobies:
1)"I feel like I understand my system so much better by compiling it myself"
2)"It takes a long time to compile a system, but it runs sooo much faster once you do".
You never hear people talk about how great the documented gentoo is (they say that, instead, about BSD), you never hear about what a great environment Gentoo is to develop in -whereas people frequently comment about how much cleaner, elegant and easier to understand the code from -say- Net or OpenBSD is.
Now, as far as things to suck down -you might consider swallowing the fact that pkgsrc has existed longer -and does better- as a cross-OS packaging system and that Debian pretty much invented the concept of the Meta OS.
While Joe User may end up a developer after using Gentoo (though I've seen no evidence of that happening) I think that's much more likely to happen if Joe User either uses and follows the Linux From Scratch book or downloads one of the BSDs and gets his hands dirty experimenting with and seeing what breaks in