UNIX - Let's get together and feel all right

UNIX - Let's get together and feel all right

Summary: People have often accused me of dreaming too much and not having my thoughts anchored in reality, particularly in this world of multi-billion dollar technology companies and big business, when huge amounts of money are at stake and every competitor is vying for the top spot.In the world of high-end enterprise systems, that top spot - the world of large symmetrical processing UNIX systems - is divided between the tier 1 systems vendors - IBM, Sun and HP.


Unix ‘Mon!People have often accused me of dreaming too much and not having my thoughts anchored in reality, particularly in this world of multi-billion dollar technology companies and big business, when huge amounts of money are at stake and every competitor is vying for the top spot.

In the world of high-end enterprise systems, that top spot - the world of large symmetrical processing UNIX systems - is divided between the tier 1 systems vendors - IBM, Sun and HP. All of these have their own proprietary architectures - IBM has POWER with AIX, Sun has SPARC and Niagara with Solaris, and HP has the PA-RISC and Itanium platforms with HP-UX. Only two of these vendors, IBM and HP, are able to run enterprise-supported versions of Linux in addition to their proprietary Unixes on their native platforms, although Sun has experimented in the recent past with Ubuntu on their Niagra chip, a relationship which seems to be waning.

Back in the olden days of yore, when the earth was still new, there was one version of UNIX - AT&T's. Much like the current days of Open Source, this code was shared freely among vendors, and the systems were largely source compatible with each other, until the vendors decided to all go their separate ways and rule over their own feudal computing ecosystems. Fast forward 40 years later, and you have the situation where we are today, where many incompatible versions of UNIX now exist and an independent, rogue cousin -- which shares no lineage with UNIX despite having shared ideology and ported applications -- is now nipping at their heels for the same market share.

It has always been a dream of mine to see UNIX re-united within my professional lifetime into a single, standardized "mother" distribution. It could be argued that Linux displaces the need for UNIX, but as I have said before, commodity and utility computing will never fully displace enterprise-class systems. But it makes sense that they should be source compatible and migrating applications and code should be as easy as flipping a compiler switch or spitting out build packages with something like OpenSUSE Build Service. In an ideal world, there should be no such thing as a "Solaris-ism" or "AIX-ism" or "Linux-ism" or any kind of "ism" that inhibits code migration and application portability. Like the Linux Standard Base, which aspires to standardize the base distribution by which all Linux distributions are to conform to, the same should hold true of UNIX. POSIX -tried- to accomplish this, but proprietary OS extensions and varying platform differences effectively neutralized its effectiveness as a UNIX standard.

However, it seems, 40 years after the first AT&T UNIX release, there is a glimpse of hope that "Unixfication" may again be realized. At the most recent Oreilly MySQL Conference and Expo, Sun CEO Jonathan Schwartz discussed in his keynote the willingness of his company to further cooperate with the Open Source community and move more of their proprietary CDDL licensed material into GPL, the most predominant of the OSI-approved licenses. According to a trusted source and a significant community contributor who attended the keynote, it is my understanding that after completing his speech, Mr. Schwartz engaged in "locker room" chit-chat with members of the audience and discussed the possibility of moving projects such as Solaris into GPL2. Naturally, I asked Jonathan to clarify his comments, and this is what he said:

"Glassfish is dual licensed - CDDL and GPL. And as you're aware, MySQL is GPL(2), as is the Java (runtime) platform itself. So three very big components of what'd be a complete OpenSolaris platform are available to the broader GPL community.

My point was that Rich is considering making other portions of our "CDDL only" portfolio available to the GPL community - including elements like ZFS, which are today in BSD, OpenSolaris, etc., but not Linux. (Executive Vice President, Software) Rich Green is leading the charge..."

Now, of course, I would like to see some real commitments from Sun to actually address the fate of the CDDL and whether or not -all- of their major Open Source assets are up for grabs here rather than using couched and squirrelly double-speak. It is no secret to anyone who has followed my writings that I have always disliked the CDDL , because of the parallel "Mirror Mirror" universe it creates which divides OpenSolaris and Linux developers. Eliminating it altogether and embracing GPL2 would be a first step in uniting the Unix and Linux platforms -- because without any licensing barriers, it would allow the two systems to organically cross-pollinate each other, share code -- and dare I say it -- eventually merge.

Once Solaris is GPLed, it will be only natural, in my opinion, for the other major UNIX vendors to follow suit, which would bring about the eventual platform unification through cross-pollination. However, another well known advocate of Unixfication, Eric S. Raymond, who is one of the authors of the Open Source definition and co-founder of the Open Source Initiative, doesn't necessarily agree with me:

"The problem with changing the licensing on a huge project like AIX or HPUX is that due diligence requires you to do an audit for 3rd-party IP claims first.

That's an extremely expensive and time-consuming proposition -- so much so that it may just be easier (for IBM, in particular) to move to Linux and let the proprietary-Unix code quietly expire. " The AIX kernel core is fairly old at this point; and given the breadth of Linux driver support, it seems unlikely that they need to salvage anything from the driver set. Heck, Linux even has support for stuff like DASD and magnetic tapes. And if the stuff they need to keep is in userspace only, why not just GPL those bits and port them to Linux?

I'm not saying there's anything wrong with the idea of GPLing the entirety of AIX, it's just not obvious that this would be their least-effort path. HP-UX has exactly parallel problems."

I certainly understand where Eric is coming from, but I have to think that once Solaris and Linux start mixing it up, the other two big guys will want to be able to play in the standardization sandbox as well. HP or IBM ceding the world to SoLinux? I don't see them doing that, regardless if all they want as part of the "mother" distro is userspace bits, as Eric puts it. And don't get me started on Apple and Darwin and all the other BSDs.

Will the GPLing of Solaris lead to Unixfication? Talk back and let me know.

Disclaimer: The postings and opinions on this blog are my own and don’t necessarily represent IBM’s positions, strategies or opinions.

Topics: Linux, Open Source, Oracle


Jason Perlow, Sr. Technology Editor at ZDNet, is a technologist with over two decades of experience integrating large heterogeneous multi-vendor computing environments in Fortune 500 companies. Jason is currently a Partner Technology Strategist with Microsoft Corp. His expressed views do not necessarily represent those of his employer.

Kick off your day with ZDNet's daily email newsletter. It's the freshest tech news and opinion, served hot. Get it.


Log in or register to join the discussion
  • Eric may be right, it ttmight just be too much legal work, but, Sun could

    GPL important parts that could be mixed with Linux. Of course that would mean they would have to move to Linux at some point and give up the advantage of having the best Unix.
  • Way to go!

    You just lost all cred!:

    "...but as I have said before, commodity and utility computing will never fully displace enterprise-class systems."

    As though Linux is not enterprise-class? What? Where have you been the last 5 years?
    • You misread

      I am talking about a class of service -above- utility computing and mid-range. Linux does not scale well to large, multiprocessor systems (16-way, 32-way, etch) with sophisticated, high performance I/O. While Linux will run on pSeries platforms and the IBM zSeries and HP Integrity, it is largely performing in a virtualized role, using LPARs and nPar, not running large monolithic systems on its own. It does not perform as efficiently as UNIX when configured in this fashion.

      I don't count clusters as enterprise computing for large transactional systems, those are CPU intensive and not I/O bound and require major architectural changes to applications in order to make them scale to the level of a large UNIX machine such as a Sun E25K or a IBM pSeries 595.

      Solaris and AIX are still superior to Linux in these respects. For a large number of applications you don't need systems like this, but it would be good for Linux to inherit UNIX DNA for this purpose.
    • No Cred Loss

      No techboob_z, he didn't lose cred. You can't compare an x86 machine to a RISC Enterprise server...and that's pretty much what he's saying. You are comparing motorcycles to dragsters and it just doesn't work. The most powerful motorcycle can't compete with the most powerful dragster and the same is true of CISC vs. RISC computing.
  • RE: dead wrong about Sun

    1) Niagara is a SPARC chip.

    2) Sun's x86 gear runs Linux for x86 - including the "enterprise supported" variants.

    3) Most Linux code runs unchanged on Solaris for SPARC - essentially its a re-compile with minor path changes and you can even avoid those if you want to use a Linux run time container.

    4) Hobbyists often load native Linux on SPARC - Debian, Fedora, and Ubantu all compile easily. There's little business support for this because it doesn't make sense - since Solaris will run the applications at least as well, why bother?

    5) I stopped reading after the AT&T bit - that's so far wrong I'd need more time than it's worth to correct you.
    • Niagra runs SPARC but it is not SPARC.

      The architecture is so wildly different from traditional SPARC architecture that it deserves to be in its own category. The SPARC version of Solaris 10 might run on it, but the kernel does very special things on that chip to make it "beahave" like a traditional SPARC system.
  • Been Using Unix, AIX and Ubuntu , love it!

    I've been using Unix since college, AIX and zLinux , and Ubuntu for fun, cant go wrong here. If you havent ran Ubuntu check out my articles on setting it up.


    • Re: BeenUsing Unix, AIX and Ubuntu, Love it

      Nice website!

  • RE: UNIX - Let's get together and feel all right

    Yes, AT&T did design the original UN*X back in the late 1970's. Here's what I was told happened.
    Two bright coders (Kernigan and Ritchie) decided to write a system. At the same time, an AT&T manager/exec needed a new operating system. He saw their first release and decided that their play toy would then become the next OS. And so, UN*X was born. Along with that you have the C compiler, which was needed in order for UN*X to function. In the early days its my understanding that people wrote their own COMMANDS in C if a system command was not available. The OS I am referencing here became (eventually) AT&T SYSTEM V UNIX REL 4.0 which is what many of the current releases of UN*X today have at their very core. Sometime later, a group of people at Berkley university decided to write the Berkley toolset. Eventally you then had the divergent path of SYS V (AT&T) and BSD. The first releases of SUNOS were BSD based and stayed that way until SUNOS 5.0 AKA SOLARIS 1.0. So regarding compatability, SYS V based versions of UN*X were compatible at some levels and not at others. You also had releases of BSD that sprung up such as FreeBSD, netbsd, etc. In addition there was the venerable old INTERACTIVE Unix by Sun which was a System V release of UNIX for i386. Of course who could forget Xenix (or was it zenix?), DG-Ux, HP-Ux (still around in SYS V flavour). IBM went their own way with AIX early on. In fact porting from Solaris to IBM or HP to IBM early on was not so simple.
    So while I believe your article is very informational, UN*X is a very wide and varied beast and while our favorite shell may look the same on different platforms now what's underneath is quite a bit different!
    Regarding linux and the enterprise. Wow that's a tricky one there. I like linux for small app servers (telephony, web servers, etc) clustered or load balanced. However, there are a plethora of driver issues with many versions of linux and tweaking to get the best performance takes some doing, whereas for the extra moeny you can buy hareware certified to run a vendors build of Un*X (Sun, HP, IBM) or go Winx64 and get very good reliability. nothing is simple, obviously any platform you choose requires a great deal of effort to setup and run properly but UN*X is STILL KING for very large boxes (16CPU's and 128GB of RAM). Or you can go mainframe for stuff like Oracle/DB2 but then you're really paying the big bucks.
    So, linux will get there eventually. The major vendors have embraced it for some time now, and the 64bit builds ARE getting better.
  • IBM sells AIX

    Why should the company sacrifice revenue?

    With the influence IBM has over Linux, it's unlikely that Linux will soon be able to match AIX, so the money may be expected to continue for some time.

    When speculating about the future actions of for-profit organizations, especially IBM, shouldn't attention be given first to profits?
    Anton Philidor
  • RE: UNIX - Let's get together and feel all right

    i think if this does happen, the computing world would do nothing but benefit. maybe someday we can see systems that have always been "incompatible" become one system.
    the "Solinux" would be great. internet applications would become much easier to standardize, companies would have a much easier time writing software. and not to mention the benefits in security under a united Unix system.

    (but as always. someone wont like the idea, and try to break away from everyone else by creating something completely new and different.)
  • RE: UNIX - Let's get together and feel all right

    Well... I'm unconcerned about Unix coming together or whatever. I think what is more important than processors and OS's is that data formats "come together." Do that, and people will imho be happier than a hamster in its wheel!

    Disclaimer: That doesn't mean with OXML!
  • Then we can sing Kum Ba Yah

    Unify Unix? Who cares about Unix or any OS for that matter? Unify the document formats.

    End users care less about the OS and more about getting what they want. That covers playing Everquest, calculating quarterly profit/loss statements, or typing a paper for school. They just want to do what they want to do - period.

    From a strategic standpoint, older code needs to be replaced by better code for a number of reasons. Perhaps the conjecture is right that the various 'nix vendors need to get behind Linux and make their best features available for incorporation.

    It's not that they will be losing clients as the clients are forced to upgrade equipment and are therefore forced to update OS from time to time. (damn 1st law of thermodynamics)

    Doing this would allow the newer code - Linux included - to work seamlessly with whatever existing older parts still exist within any organization.

    At some point, Unix will change or die, just like people. The question is do they want a graceful transition or a horriffic crash?

    Customers want the less painful route.
    • The article is not about "desktop" computing...

      you missed the point or don't understand what's being talked
  • There never was "one" unix. There never will be.

    "Back in the olden days of yore, when the earth was still new,
    there was one version of UNIX - AT&T???s"

    That was true for only as long as UNIX was internal. As soon as
    it went out into the world, changes began to be made, and in
    no time people were licensing from AT&T and then picking up
    enhancements from Berkeley. The divergence took off from

    There have been many attempts to rejoin, at many different
    levels. AT&T's SVID acceptance tests. X Window in an attempt
    to have a standard window system. Motif in an attempt to
    standardize the GUI toolkit. Each time there was always a
    company, typically the market leader (e.g. Sun) who fought the
    standard. Why? Because it eroded their market dominance. As
    soon as their was a real standard, installed base no longer
    became a factor, and newcomers could undercut the larger
    company with its overhead and R&D costs.

    Furthermore, even those companies signed up for the standard
    would subvert the system indirectly. Everyone needed a value
    add that made their system standout. So they would add
    particular features that targeted their market. And some of
    these would become important for customers, who would then
    force other vendors to implement these non-standard features
    in order to bid. And in no time at all, the standard was

    There will be no standard Unix. The market won't allow it.
  • RE: UNIX - Let's get together and feel all right

    As long as there are open, patent free, standards, for creating a homogeneous network of disparate machines and data, something that UNIX/Linux systems are good at, there is no real problem.

    As more high end applications migrate to the Open Source development model, porting them to differing hardware platforms running a variant of UNIX/Linux becomes a no brainer.

    When Web2 becomes ubiquitous, desktop systems will disappear, replaced by human interface devices directly networked to the cloud, either through a physical pipe or a wireless link.

    All we will need is access to a computing network which serves applications and relevant media, and has access where required to the big iron UNIX centres for processing and storing massive data sets.

    This is robust togetherness in a system that caters for the disparate needs of hardware vendors who modify their OS for optimal performance on their hardware
  • RE: UNIX - Let's get together and feel all right

    • Why Not

      If SGI has the source and there are materials there that would be of benefit to the Open Source community to release in GPL2, then I don't see why not
  • Incompatible?

    Sure, there is binary incompatibility... you cannot simply plop a binary from one variant into another and expect it to work. But saying POSIX failed is very misleading.

    Linux was designed around POSIX and most programs written for UNIX work on most variants, including Linux, with little more than a recompile. You all but ignore the very important source compatibility that Unix-like systems are famous for.
  • RE: UNIX - Let's get together and feel all right

    Sun bought MySQL, They bought Innotek and Solaris is available as a .iso imgae for X86ers. Java is being worked on and now does hardware acceleration.

    I've always been tempted to get a later model sparcstation on Ebay, but never have.

    But, this shows me that Sun is putting more effort into what they have than M$ ever dreamed of.

    But Apple is Unix and always was, Koppix (Ubuntu), Mandriva, Debian, Red Hat, PCLinuxOS, Slackware, etc, etc.

    It would be a good idea to GPL Aix or parts of it. Believe or not, there's alot of old hardware that's laying around that still works. I have a Irwin Tape drive on my 466-133 even today and a 2mb Tseng Pci local bus card. Boxes and boxes of old software......then there's Minix for the ST and the Amiga of course which could serve as a lesson here because of OS 1.2-3.1 and it's compatibilty. CBM didn't bail till they were broke. Atari never used a 68k Unix as a desktop replacement. But it's available.

    If Sun could manage to stun everyone here with a OS that can handle all the Unix-Linux packages, i'd use it today.

    If this post is off base, I apologize right now.