Slashdot Mirror
IBM to offer Linux support under AIX
cswiii writes "This c|net article talks about IBM, following the footsteps of Sun, to offer Linux support upon IBM's UNIX systems. " The most recent announcement from Sun, about LXRUN is quite similar. Essentially, both parties have modified their Unix flavors to enable AIX and Solaris to run Linux binaries.
looks like everybody wants to be able to play with everybody else (except M$ wants to be the cool kid on the block, the one who is condescending and snotty, but kids still want to be their friend). This is very good, with all this interoperability between Unixes' maybe linux will heal the UNIX rift that formed long ago. Just think of it, Linux binaries as the bandages for a weakness of your flavor of unix.
Now how about IBM helps us get AIX binaries running under linuxPPC?
This is my signature. There are many signatures like it but this one is mine..
As went OS/2, so shall AIX.
Yes, the proliferation of lxrun will reduce the number of Linux installations in the short run. In the long run, the commercial ISVs (especially smaller ones) will cut costs by cutting AIX, Solaris, and SCO Unix versions in favor of a single Linux version that runs on all of them.
IBM, Sun, and SCO will be caught in the trap of having to make significant investments in lxrun to keep compatibility with the newer versions of Linux, while people who want the fewest compatibility hassles will go with the Real Thing.
Only in areas where the non-Linux alternatives are clearly more capable will AIX, Solaris, and SCO Unix survive -- and slowly wither as Linux becomes more capable.
lxrun? Is that Finnish for world domination?
Because, the tighter one's grip, the more star systems will slip through one's fingers.
Sun Microsystems wanted control, specifically so that someone like Microsoft could not "embrace and extend". Linux, with the GPL, gives up that control. You can embrace and extend all you want...so long as you ship your extensions with the source code. In ESR's annotations of the Halloween Documents, he notes that the openness of a development environment paradoxically prevents forking, as forking is the last effort at moving a development process in one's own direction. In an open situation like Linux, where anyone can (and does) enter a change, few people have a reason to fork Linux.
I see two Microsoft strategies to dealing with the LX86 runtime environment. One is to submit code itself (likely under an astroturf front) which sabotages the kernel. This would have to be extremely subtle, if it is doable at all, since bad additions to the kernel will generally get thrown out (binary Darwinism).
The second is to embrace and extend, but outside the kernel level. They cannot E&E in deep kernel (because their extensions would necessarily be copylefted), but nothing is stopping them from creating a payware shared library with The Killer Feature. By doing that, they can get people to write software to their library, thus requiring all users of said software to buy said library. It would be the OS monopoly all over again, but sitting on top of LX86 rather than just the Intel hardware itself.
That is, Microsoft can sacrifice kernel control and still collect their tax. To do so, they would need a Killer Feature that Linux can't provide. The most obvious such feature is Win32 itself. If they ported Win32 to LX86, they would be able to convince vendors to stop porting to Linux, since every LX86 can become a (relatively) proper Windows box.
Do we have countermoves to this strategy? What are the odds of this happening, and the factors that change these odds? IMHO, MS will have to swallow some pride to do this, but they might at that. Then again, it would be at least a partial victory, as Those Of Us Who Know would be able to run around beneath the Gates layer.
--The basis of all love is respect
Yes, the article is deceiving. lxrun does NOT require any modifications to the operating system to run Linux apps.
.001% of Solaris users, but the Voodoo drivers do not currently work with lxrun. The reason is that Solaris's /dev/mem is only the size of your physical RAM, while Linux's is apparently 4GB no matter what. Since the Voodoo driver mmaps the card at whatever high RAM address the PCI/AGP bus gives it, this fails on Solaris. So no accelerated Quake 3 for Solaris (yet) :(
/dev/mem structure in Solaris be changed. I doubt that would happen.
:-)
;-)
However, there are a couple of modifications that do need to be made to make lxrun perfect (at least on Solaris). First of all, there is a way on Unixware (lxrun's original platform) to produce the necessary ld-linux.so.1 that actually is lxrun. It's a little confusing, but that would essentially allow you to type "linuxquake3" instead of "lxrun linuxquake3" to run Q3test.
Unfortunately, the current Solaris linker doesn't have an essential option (I think it controls the relocation) to allow this to happen. I have heard that people at Sun have done this modification and gotten lxrun to work in this transparent mode. Therefore, Solaris 8 may have this option (no guarantees though).
Second, this probably affects
Obviously, getting that to work would require that the
Also, kernel modules are out of the question... I think. Never tried modprobe or insmod. That could have some "interesting" effects though, given their close dependence on the kernel. Hopefully it just ends up with signal 11. I guess I get to test them today
Oh, BTW, don't use Stampede with lxrun... for some reason pathnames don't work properly. They were fixed when I put in SuSE. Dunno if it's the glibc2.1 or the Pentium optimizations.
In other words, it's a lot less buggy than I thought
Glückwünsche, haben Sie Slashdot ermordet, indem Sie zum korporativen Druck beugten und Subskriptionen einlei
Possibly.
:)
Keep in mind what these "rumblings" from some of the "big guns" COULD mean.
1. We have Corel already working on a distribution, one that, IMHO, will be a bit finer tuned to run whatever apps they've decided to port (ie: all of 'em).
2. Linux intreroperability between Unices could be (hell, IS) a step towards a possible reunification in the *nix world. Binary cross-platform compatibility would go a LONG way towards killing the current FUD with respect to "splintering" Linux.
3. The definition of the LSB needs to get set _REALLY_ quick so that if Big Blue, and anyone else, DOES decide to do a distro, we'll have a base in place from which to build. Packages built on one distribution would be even more easily pluggable into others using whatever package manager is deemed worthy to be deigned as the LSB package manger.
I'm rambling. YMMV
With Sun and IBM's OSes able to run Linux binaries out of the box, in addition to BSD, MacOS X, and (I think) BeOS able to run them with a recompile, it seems to me that Linux has suddenly become the cross-platform standard. Developers can truly "write once, run anywhere," and as more developers realize this, Linux will gain more applications. Once that happens, commercial OSes will have to be better than Linux to sell - after all, why buy what you can get for free? This will keep commercial vendors innovating, and Microsoft out of the picture unless it gets compatible real soon.
-Rafi Remove the Spanish to email me.
{sigh} I wish IBM hadn't killed AIX PS/2. It ran really well on a 386-genre hardware and the X11R4 implementation was good in terms of performance. The Token Ring driver is also better than the one in the Linux kernel {grin}.
Seriously, I have to wonder just what ``Linux'' binaries AIX is going to run, as GNU/Linux doesn't run in the same CPU mode as AIX does on the high-end PowerPCs (e.g. G4 and allegedly G5).
I need to get back to work on GNU/2 and OS/GNU (OS/2, GNU, and Linux integration software). It'll be cool once it works.
Cheers,
Joshua.
--jon. Postel is dead. May we all mourn his, and our, loss.
The last sentence of the first paragraph says:
"IBM also will package its other software with a version of Linux."
Does this mean what I think it means?
Ed
I find it interesting that Sun and IBM are basically indicating a willingness to take advantage of Linux-oriented applications while maintaining a proprietary operating system on their hardware.
This was, of course, the mode for most of what is now deemed 'open source' software prior to the development of the Linux kernel, only then it required the software be recompiled on the platform in question. It also renders the need to actually run the Linux kernel as moot -- Linux binaries will run quite nicely on any old proprietary operating system that puts in the appropriate hooks.
Goodbye Linux.
IBM is making the same mistake that Sun just made, and Digital made before them.
Emulation doesn't sell machines!
Anyone ever see NT running on a high end Intel box? Compared to a mediocre Alpha box? When the services you need are available as native Alpha code, Alpha always turned out to be superior.
Well the applications never came. Digital started marketing heavily for their x86 emulation onboard but it didn't work. Sure, Alpha's are still supported by NT, but out of the thousands of NT machines I've designed and installed, my clientelle only accepted five Alpha-based NT boxes.
Alpha is still the superior technology, and not all that much more expensive than Intel these days. But it doesn't matter. Without the native apps, the servers just aren't moving (at least in the NT industry).
If IBM wants to hitch its wagon to Linux, we need to see Linux running natively on RS/6000's. We need to see IBM shipping binaries of popular GPL applications, prepackaged for IBM systems (of course the source is available as well).
Emulation is just a kluge, and a poor marketing gimmick. IBM's customers are more clueful than that and won't fall for it. Show me a 43P running "Blue Hat Linux" with KDE and a full suite of GPL'd apps, with IBM improvements, running natively, and I'll show you a hot seller.
I though AIX only ran on the PowerPC architecture. Does this mean they will run PPC Linux binaries, or x86 binaries under cpu emulation?
Certainly, the GPL is filled with strict legalese... There are very strong restrictions on what you can and can't do with GPL'd code. Really, if you want to create a surviving standard, you have to have one of three things
1) Enforced Open source (Linux)
2) Installed Base (M$ Window$)
3) Strongly worded licenses (Java)
These are all strategies designed to stop people from killing your standard. M$ had some very strong business reasons to try to kill Java. As a result, Sun had some very good reasons to try to protect it - they just didn't do it the right way. Open Source ensures that your competitors can't steal the standard from under you, and twist it into something different, or kill it for reasons other than technical reasons.
apparently linux is turning into the dos of old: there was one ms-dos (the gold standard, i suppose) and several OS (dos) clones that could run dos executables. until, at least, MS decided they didn't like it, that is. where linux is different is that there's no MS to decide that this negatively affected their bottom line.
Emulation sold many machines for Apple Computer. Good emulation was necessary for allowing the "Power Macintosh" series to take off. In fact, the emulation was a little bit *too* successful: some vendors were satisfied enough with its performance that they continued to release 68040 binaries for a year or two after the Power Macs' release, electing to add new features rather than port what was already a sufficiently fast product.
What's the difference? Apple pushed the Power Macs as the next step in their line; they were the clear successors to the Quadras, the new cutting edge.
There is no such unambiguous message with Pentium vs. Alpha. Intel markets their chips beyond belief, regardless of the Alpha's superior performance, and people buy it. It's as though some other company had released the PowerPC Macs while Apple shouted the glories of their upcoming 68060 machines with colourful dancing bunnymen.
-Mars
I get the feeling that Linux x86 becoming another multi-platform runtime enviroment? How come Java needs such strict legalleze to "protect" the dieing standard from "corruption" and LX86 standard can survive and grow under GPL? The fact that the two major players in Java (Sun and IBM) are now supporting a LX86 RTE should shake up Sun Microsystems Java Legal dept to the fact that a bad license is not marketable no matter how "hot-word compliant" your "technology" is.
It is cool that IBM recognizes the desire from it's customers to run certian linux x86 binaries on AIX. However, I think that IBM still has missed out on one of the largest freebie improvements that they could take from the Linux movement. AIX's "installp" is ugly and has been overdue for a complette rewrite for a LONG time. The fact that so much of AIX v3's installp has remained with AIX v4 is proof that IBM does not know how to correct the monstrosity they created. One the other hand, RPM is probably less than 5,000 lines of code away from doing everything that installp does and do it cleaner. If only IBM could see the advantages there would be by moving to an Open Source package management system.
There are many questions with this.
The first is what kind of Linux binaries are they talking about? Linux x86 binaries? If so you have the problems of CPU emulation in addition to syscall (etc.) emulation.
Ok, now assuming they are only talking about PPC Linux binaries you run into another set of problems....
All of the "current" AIX machines use 64 Bit Power(PC) processors... CURRENTLY Linux PPC only runs on the 32-bit version of the PowerPC.
Is emulation required for that?
Or are they going to support 64-bit PPC Linux Binaries when they are done? (I believe the ABI and such are all the same for 32-bit and 64-bit binaries.....)
There are a lot of questions to how IBM is going to do this, and what exactly they are really doing.
Do Linux gurus blend into IBM's "Big Blue" mentality?
Yes. I work at one of IBM's major locations and Linux is probably almost as pervasive on people's screens as AIX (either that or folks are running GPL'd window managers on AIX). In any case, Linux is quite popular here and the management has been quite tolerant. The corporate lawyers are pulling their hair out over GPL issues though.
I have heard from others, too, that IBM has relaxed its strict personnel rules over the past few years. But can they integrate Linux developers into their fold?
We can and we have. I don't have to tell you of course that the Apache credits are thick with IBM contributors. But Apache!=Linux. The main talent behind PHP just started working for IBM as well. Of course, this doesn't mean IBM is looked favorably upon by Linux developers, but it does mean that "Open Source" (God, I hate that term) developers can find a happy home at IBM.
It was interesting to note that at Linux Expo, a great many of the attendees were incognito IBM'ers. IBM definitely had a stronger presence than even Red Hat (though the pretty young lady at the Linux Magazine booth, who shared space with Red Hat, did get me to come visit...)
So existing Linux developers can and are being brought into the IBM family. Existing IBM'ers are also learning to embrace Linux, despite the fact that our corporate lawyers are making it almost impossible to do business in the Linux industry through official channels (for example, we are not allowed to look at GPL'd source code, but at the same time we need to create binary modules for kernel code that contains intellectual property).
The final problem is that the bean counters don't know what to make of it. They are used to seeing nice clean sales figures and since Linus is so... well... so FREE (in the free speech sense, as well as the free beer sense), nobody really knows what the true numbers are or even if the estimates are close. This drives them crazy. Without hard numbers, it is too hard to procure funding to start producing products for Linux. The only reliable metric we have to go by is IBM customers calling their IBM marketing reps and demanding Linux versions of IBM products.
Getting back to Linux inside of IBM; there are several "early adopter" groups within the company that have fully embraced Linux and would be a great place to work. For example, the Netfinity group has embraced Linux whole hog and would be a great place to send resumes. There are other very big announcements that have yet to be made, and will make this AIX announcement forgettable.