The Unix Wars are all but lost and won, and Unix System V.4 is the victor in the core operating system battle – witness the fact that OSF/1 1.1 will be System V.4-compatible, while the Open Software Foundation has won the vast majority of hearts and minds for its Unix extensions. IBM Corp’s AIX Systems Manager Bill Sandve lights up when he reels off the numbers relating to the Foundation’s first big success, the Motif graphical user interface – over 350,000 binaries shipped, running on 145 machine types, 800 applications available, 1,400 source code licences out, 85 suppliers, over 50,000 developer’s systems in place. And that was as of last September. IBM is treading these days as it tries to put over the message that there will be strong interoperability between its Unix systems and its proprietary environments, and Motif is a case in point: it is claimed that Motif 1.2 will have the look and feel of both IBM’s Common User Access end of Systems Application Architecture as well as of Microsoft Corp’s Windows 3 – and support for the X Window System 11.5 release. Further out, release 1.3 or 2.0 will add style consistency, data exchange capabilities, and the intention is that graphical user interface development should become much easier.
Microkernel
Next up from the Software Foundation was the operating system itself, OSF/1. In release 1.1, X/Open Co Ltd’s Portability Guide 4.0 will be complied with by mid-1992, and the 1.1 release will offer Distributed Computing Environment integration, better performance and higher levels of security. Then in 1993, we can look for OSF/1 1.2 with real-time extensions, Greater Trust, System V.4 compatibility, fault-tolerance and transaction processing support. The operating system will move to a microkernel environment with Mach 3 from Carnegie Mellon University: functions and capabilities will then leave the kernel and move up into user space servers. This will make it possible to overlay other operating systems such as OS/2 – if anyone is still using it by then, Pick, Microsoft New Technology as well as Unix on the kernel. The next major Software Foundation offering is the Distributed Computing Environment, designed to make it possible for a user at a workstation attached to a network of heterogeneous machine should be able to log onto any machine on the network without needing to know where it is or much about it; work should also be able to be parcelled out to the machine most suited to executing it. IBM is promising to offer the Computing Environment with AIX 3 in the second half of next year, and there will also be some support for it under MVS – that depends where the IBMer you talk to is coming from: some will tell that there will be lots of support for DCE under MVS. Next up will be the Distributed Management Environment where the technologies have been chosen but no-one has yet been picked to integrate them – IBM is integrating the Distributed Computing Environment for the Foundation.
Genuinely new
The Distributed Computing Environment is something genuinely new, but the Distributed Management Environment is quite the reverse. The biggest weakness of Unix in a commercial environment is the well-nigh total lack of systems administration tools and facilities that mainframe users have long taken for granted: many such users will not consider Unix until that infrastructure is in place, and that is the problem that the Management Environment is conceived to address. The features the Foundation asked for in its Request for Technology were a standard means of software installation and distribution; a means of managing software licences; tools for configuration management; and standardised print services. Although the Management Environment is vitally needed, it is clear that the Software Foundation, which doesn’t yet have the Computing Environment out the door, shows signs of running ahead of itself in its dash to get everything the MVS junkie could possibly wish for in place under Unix. Problem is that in tomorrow’s brave new world of object-oriented technology, it is quite cle
ar that all the facilities that the Management Environment will embody should be implemented as objects, but the standards are not in place yet, and so although, as Bill Sandve says, we want to move to objects, the Foundation is having to start with a layered applications programming interface approach. All of which suggests that the Foundation is getting carried away with its own enthusiasm: all the sponsor members of the Foundation depend substantially on their mainframe or proprietary minicomputer lines, and while it is highly likely that in the medium term that the mainframe market will migrate to Unix, all those sponsor companies actually need to continue to sell mainframes in very substantial numbers for several years yet, so that eliminating one of the most telling arguments against moving to Unix at this stage in the game is just plain stoopid.
Installer
Content from our partners
And the aim of integrating open and IBM proprietary products hits a snag with systems management, because IBM’s SystemView requires a central host while the Distributed Management Environment, as the name suggests, does not. Finally – so far the Foundation is committed to producing an Architecture-Neutral Distribution Format so that a single application written to the rules will be able to load and go on any machine running OSF/1. The big worry is that it will work, but will impose such an overhead that applications will not run fast enough. The technology is still at the evaluation stage, and the efficiency of the installer written for each piece of hardware on which it is to be supported is the key to acceptable performance. The other problem lies in testing: when an application the developer thinks is written to the rules fails to run, where is the problem – in the application, in the installer on the hardware, or somewhere in the Format itself, with a condition no-one had thought of turning up. Here, says Bill Sandve, application testing suites will be big business for those that come out with convincingly comprehensive and exhaustive ones. Of course IBM is right at the beginning of what promises to be a very long road and its declared objective is to converge AIX and OSF/1 at the same time persuading the world to accept that AIX is the industry-standard implementation of Unix. And the company will take a very significant step forward down that long road early next year, with the introduction of the 3.1 release of AIX, which will include enough OSF/1 code that IBM will have to start paying licence fees on it to the Open Software Foundation for the first time…
