IBM Corp has acknowledged that the long-awaited Repository for the AS/400 is a difficult child. Ron Fess, Rochester, Minnesota-based Foundation Manager for Openness, says that when the initial system did not offer adequate performance, the Repository team went back to the drawing board. Apparently they are still there, debating whether the rejigged version is up to scratch. The possibility exists, according to Fess, that IBM may skip from the present definition of repository to a system class library. Some commentators say that system class libraries are considerably less sophisticated than the concept of a repository as propounded by IBM. But that may be missing the point since they are two entirely different animals. A class library stores objects and handles things like images and speech, both set to become increasingly important on the AS/400. The Repository is inextricably linked with AD/Cycle, so the AS/400 must have one. Nonetheless, it can be at the bottom of the heap and no more than a data dictionary. Fess admits that the delay could lead to a loss of market share, possibly to Unix-based systems. But he claims that Unix systems are less scalable than the AS/400 – at least, based on his experience of testing Unix repositories from both third parties and IBM internal developments. Does this mean that IBM is developing a repository for the RS/6000? Sources say that the repository development for AIX continues apace with the Toronto labs about to launch a Portable Common Tool Environment-compliant repository. The twist in the tale is that it will be licensed back to Hewlett-Packard Co (CI No 1,921). Fess went on to discuss his plans for changing the object-based AS/400 into an object-oriented system. He says that system objects will act as a base, and users will be able to add self-defined objects. Unlike other forms of object technology which have temporary objects, Smalltalk for example, Fess says that he will keep the AS/400’s existing persistent objects, although the AS/400 uses Smalltalk. IBM is not collaborating with the Object Management Group in its quest to make the AS/400 object-oriented, but the fairly recent member is working with the Group on the Object Request Broker. Other topics under discussion include interfaces for disk attachments like RAID devices. Fess has reservations however, saying that the issues are quality of components and system recoverability. Other sources say that the RAID device will be launched next year, supposedly at level seven which focuses on fault-resistance in the control unit and active checksum capability. And what of Posix? It appears to be of limited value, but given that IBM committed itself last September to making OS/400 Posix-compliant, it’s not unreasonable to ask when its going to happen. No dates as yet, and Fess insists that a definite timetable would not be an indication of further commitment since his development cycles won’t change. No doubt, but commentators say that IBM Rochester is deliberately understating its Posix plans. Fess says that he can’t support X/Open Co Ltd’s Portability Guide, mainly because the test suite has ANSI dependencies, and the AS/400 still lacks SQL capabilities. However, he will support Release 4 of the Portability Guide although that’s a moving target – and that ties in with the forthcoming Posix machine in 1995. The new AS/400 architecture will be Posix-compliant and provide system calls and tools and libraries. It should be a comfortable box for both AS/400 and Unix users and should support concepts like threads, also keeping C users happy. Nonetheless the AS/400 will have to be mapped onto that since IBM can’t run the risk of another fiasco like the System/36-to-AS/400 non-migration. Unix and RISC processors are on the also cards. For the central processor to be RISC-based, IBM needs 64-bit implementation. But until that arrives – and it’s worth noting that Rochester is said to be collaborating on that with the Unix team in Austin – IBM will put RISC processors in the input-output processors. In terms of opening up the machine f
or software development and database companies, he plans to redo the C compilers and support a different programming model, currently focused on RPG. Fess says he has spoken with several database companies and discussed front-ends, but he claims that they are not particularly interested – which must come as a surprise to folks at Oracle Corp. If all this comes to fruition, then what goes on under the covers will not be the AS/400 as we know it. The AS/500 will have a different instruction set and supporting hardware, but users don’t operate at that level, and the one thing that IBM won’t do is mess around with the program level. – Janice McGinn
