From Paul Newman, Technical DirectorPacific Associates LtdFrimley Business Park GU16 5SG
I have to take issue with several inaccuracies contained in Timothy Prickett’s article, The hurdles facing AS/400 users as they migrate to RISC models – when they are ready, (CI No 2,643). It is simply not true that AS/400 V3R1 customers have lost disk sectors and as a result, data. There is no fault that causes data or disk sector loss on the AS/400. The Storage Management components residing in the I/O processors were not written in C++. The V3R1 code, in common with all previous releases, has experienced the usual flurry of PTFs since its shipment and, as usual, is settling down well. Prickett may well be thinking of an unrelated SLIC problem on the Advanced S/36. The transition to RISC processors was not going to be easy… Well who for? Customers moving to the RISC boxes who will have their entire mach ines replaced (not all will) will need only save and then restore to the new machine. That’s all. That’s all there ever has been except for the very tiny number of customers who deleted their program templates to save space – and they were always well aware that they would need the source code to these programs if a processor, or other large scale technology change occurred. I have seen application programs written in Release 2 of CPF on System/38 that have been running on all processors since 1981 up until V2R3 of OS/400 without any re-compilation. There are several similar examples running in the labs. That it wasn’t easy for the developers in the labs isn’t in dispute, it’s not even news. The developers have been very open about the difficulties that they (or anyone else who has made the transition to object-oriented design and programming) have had. The rewrite of the microcode was the largest OO project in the world. And yes, most had to learn C++ and Smalltalk – this was partly a sign of the times: quite simply, nobody wanted to write PL/MI (not PL/1) or procedurally any more, and it was proving too difficult to recruit programmers. Who wants to modify and maintain an enormous system stretching back to 1974? …absurdly high numbers of lines of code and classes attesting to the growing complexity of AS/400. Well really! Do you think its possible to write all the function of an AS/400 into a few Cobol programs? And anyway, this rather proves the reverse: it is possible to build really large syst ems out of OO technology and further, it is desirable so to do. It was simply proving too costly to maintain 17m lines of PL/MI code. All the developers I have spoken to in Rochester have expressed at least satisfact ion at the move to OO and, for most, enthusiasm. Again, it simply is not true that templates account for 75% of the disk space of RPG programs. A more accurate figure lies between 10% and 15% for all programs, not just RPG, with some quite exceptional one line prog rams generating a relatively large template contended in this article. This is due to the automatic inclusions of compiler-generated functions such as exception handlers, I/O routines and so on, that are often a fixed size – the larger the program, the smaller the relative size of the tem plate. Very, very few customers ever deleted their program templates. Some software vendors do, in order to save space consumed by their products but it’s a practice few indulge in. And if they haven’t the source to their products, it’s hard to see how they could maintain them. As for AS/400s being late for delivery, I wasn’t aware there was an announ ced delivery date. You don’t announce delivery dates until you’re pretty sure you can stick to them. IBM has asked Pacific, as it usually does its major business partners, for its opnion on the announcement date of the new RISC machine should be, given the likely availability of the hardware. In common with virtually all European partners, we were in favour of June 21, even if a few months elapsed before general shipments. As far as I’m aware this is likely to be the date. The US partners largely favoured an ann
oun cement closer to the delivery date, which may explan Prickett’s acerbic comments, as the article originated in the US publication The Four Hundred. It’s a pity that an article with so many factual inaccuracies like this has appeared because unlike all the hyperbole generated by DEC, Intel, Microsoft Corp would have us believe, you can and will move to 64-bit RISC technology painlessly provided you’re an AS/400 customer – it’s just that the competition doesn’t want to hear that.
