Few things sound more wonderful than a new IBM product or system when it is described by one of the IBMers responsible for development of a part of the system – and some very enthusiastic IBMers, and ones with a romantic turn of phrase, are doing the rounds giving selected customers and analysts previews of the wonders of the Silverlake successor to System 36 and System 38. They of course come from the Frozen North, up in Rochester, Minnesota, and the Silverlake really exists, not far from the town. And, appropriately enough, given the area from which they hail, they are going around telling the chosen few that the machine will be announced before the first snow, and will be available by the second half of next year. Machine, however, is a misnomer: the French weekly 01 Informatique hears that Silverlake will be an entire range, starting with a bottom end processor that will cost no more than $15,000. Everyone talks in terms of the top-end model, rated at 1.75 times the power of the System 38 Model 700 – that will take the price up into the millions – $1.75m according to 01. The basic concept is well enough known – a rack-mounted series of machines that will be built in System 38 architecture, but will run applications written for the System Support Program of System 36. That now has a name – Extended Control Program Facility or XCPF. Dreamed of by the big boys The Operator Control Language of SSP will be executed by a microcode interpreter, and 36 applications will need to be recompiled for Silverlake. The new machines will accept most if not all the existing System 36 and 38 peripherals, and are tipped to come with 35% improvement in memory price-performance, 20% better disk price-performance, and a 5% more economical tape back-up system. One thing that will not be needed is anything like the 24-bit to 31-bit conversion needed when moving up to XA on the top-end machines – System 38 already has a 48-bit address range, still far beyond anything dreamed of by the big boys: there’s really room to grow there, fellahs. But while Silverlake will implement the integral relational database of the System 38, and is expected to have an enhanced version of the 38 user interface, one or two key parts of the system will come from the 36 side of the house, notably the communications support and the office automation offerings of System 36, as well as the latter’s local area networking capability. A C compiler is expected to join Cobol and RPG III as the key programming environments, and SQL is expected to be implemented on Silverlake. And where does Silverlake go in the medium term? The sky seems to be the limit. There is confident IBMer talk of a future top-end machine offering 10 times the performance of the System 38 Model 700, support for 600 terminals or more, and 100,000 transactions an hour. On the last measure, the 700 does 25,000 an hour, 10% less than the 4381 Model 14. And why is IBM doing all this? There is an army of 200,000 System 38 and System 36 users out there who would rather sell their grandmothers than convert to the hoary old 360 architecture of the 9370. But once Silverlake is out, IBM will still be left with the problem of having to support two totally different architectures and operating environments, and no doubt people will still keep insisting on buying the machine that the IBM marketing rules insist is the wrong one for their type of business.