IBM Corp’s mastery of hardware technology becomes more impressive by the month, and the news that having wiped the floor with all but the very best of the Japanese in memory chip technology, it has now moved on to leadership in the small disk drives with its new 320Mb 3.5 Winchester, soon to arrive on the OEM market (CI No 1,159), will send shivers down the spines of the companies that depend on small disk drives for their livelihood. Even in super-high-speed logic circuitry, IBM is not nearly as far behind the Japanese as the yawning performance gap between its water cooled 3090-180S uniprocessor and Amdahl Corp’s new 5990-350 suggests: IBM demands so many guarantees of reliability from its designers that its top-end mainframes are quite absurdly over designed – the company cheerfully admitted that the chips in the 308X machines were driven at so far below their potential peak performance that the things would be as good in 40 years’ time as they were when they were shipped. By designing for a more realistic 15 years or so mean time between chip failures, Amdahl and Fujitsu feel able to get much faster machines to market much sooner than IBM. Also-ran But while no-one would deny IBM’s mastery of most of the key hardware technologies, few would regard the company as more than an also-ran in what is these days becoming more and more important to the company – software development. IBM still seems to think that the answer to any software problem is to throw more programmers at it, never having taken the wise advice of the good doctor, Gene Amdahl on board: the best way to solve these things is to keep them in as few minds as possible – preferably one. So it is that IBM cannot introduce its – no doubt superb – high speed disk drives yet because the microcode – simply software in silicon – for the 3990-3 controller still doesn’t work as it should. IBM’s legendary propensity for screwing up on large software developments has also thrown the schedule for the repository for the DB2 relational database hopelessly awry, so that while something may come out later this year, it certainly won’t be what IBM had hinted at or third party software developers and users were looking for. But there are now signs that an even more vital IBM development effort, that for Systems Application Architecture, is beginning to go wrong. The people closest to the way things are going on the SAA front are still independent software developers rather than users – and a ring round of some of the most committed by Computer Systems News turned up some very furrowed brows. What makes the worries and reservations so serious is that everyone in the IBM software world believes implicitly in the concept and aims of Systems Application Architecture – a single set of interfaces and languages that guarantee that if the developer writes religiously to and in them, applications will be simply portable between 370 architecture machines, the AS/400 and the PS/2. Desperate effort Moreover, for the first time, IBM has laid down the law on a user interface, so that if a program is front-ended by the Common User Access of Presentation Manager, anyone familiar with SAA applications on an AS/400 will feel completely at home when confronted by a screen attached to an SAA application under MVS, or under OS/2 – at last, no retraining of clerical staff. But IBM’s desperate effort to cover all bases means that far from Common User Access being its preferred user interface for everything, it looks as if you are going to be able to take your pick of NextStep from Next Inc, Motif from the Open Software Foundation, Metaphor from Metaphor Computer Systems, and one or two more shadowy offerings – without ever moving outside the IBM empire. Because IBM’s efforts to crib, confine and cabin Unix while keeping its options open in case the thing does catch on simply isn’t working. And the IBMer quoted in Computer Systems News as saying that the RT is targeted at specialised scientific and engineering environments is talking through the back of his head – try telling that to the host of
RT resellers in Europe who are gaily – and successfully – marketing the thing as a general purpose business computer with the cachet of an IBM badge. But if third party software developers are unhappy that IBM appears to be two-timing them on the user interface front, they are no happier with the situation with regard to languages – pressure from System 36 and 38 users has forced IBM to include RPG II in the Systems Application Architecture portfolio – but there ain’t no standard IBM RPG on System 370 or the PS/2. But – promises, promises – there should be versions of RPG II for those architectures by the summer of 1990 – announcement or availability? And in fact, say the developers, at present, if you want an SAA language that is common across all the architectures, you can have any SAA language you want provided it’s Cobol. But the biggest complaint of all is that SAA is now two years old, and yet there still doesn’t seem to be as much of it showing above the surface as there is of an iceberg – and what is visible has to be viewed through a kaleidoscope – the damn thing won’t stand still. Two guesses It’s difficult to know when this thing is going to stop so we can mass release products, says one third party. Today you have to read three manuals, talk to four IBMers and make two guesses, says another. Moreover, IBM’s timescales for SAA are typical of the leisurely way it approaches all software developments – announce the thing for delivery in nine months, take scores of orders for it, and then look around to find somebody capable of writing the thing. Not true! protests IBM we’re moving fast to complete the SAA groundwork within the next few years, the company says. And analysts – the ones who told you Summit would be out in 1987, no doubt, say that IBM will supply true co-operative processing and cross-system application portability by 1992 or 1993. That sounds like 1995 at the earliest. The problem for IBM is that other manufacturers are beginning to implement what they can of SAA with more urgency that IBM seems to be doing, and all those manufacturers, unlike IBM, have hitched their future to the Unix bandwaggon. All the signs are that once SAA is clearly discernible, IBM will be the only major Unix systems player that won’t be supporting SAA under Unix – and it is unlikely that IBM has not spotted that danger. Every Unix vendor would love to be able to pick up all the applications in the IBM world and support them efficiently on their machines, and SAA compliance could be the key. It’s unlikely that that frightening prospect has not occurred to IBM, so don’t be too surprised if you start to see SAA deadlines slipping… and slipping… and slipping…
