Just about the time most big customers thought they finally understood the different roles mainframes and smaller servers would play in their information processing strategies, IBM has decided to change – or possibly to throw away – all the rules. IBM is trying to persuade all its S/390 customers to migrate to the same operating system: OS/390 Open Edition. At the same time, IBM is trying to reassure its mainframe customers that their software will provide all the features and facilities offered by alternative operating environments (meaning Unix and Windows NT) while also retaining the qualities and strengths that have made MVS the world standard for large scale commercial information processing. The key points in IBM’s strategic initiative are: OS/390 will support all standard Unix function calls and libraries (called application programming interfaces, or APIs, these days). OS/390 will provide programming interfaces for many Windows NT applications as well as support for interaction among those applications. OS/390 will be a host for World Wide Web services by offering both native Web server technology and a high level of external Web server support. The core offering is called Net.Commerce. The advantages to IBM and OS/390 users are obvious, if they can be obtained. For IBM, the goal is standardization that would make its software development efforts, end user support services and facilities management services more efficient. The danger is that IBM would face more direct comparisons between its System/390s and servers offered by other companies; these comparisons would include performance benchmarks, critical assessments by consultants of software quality and the establishment of migration paths that could as easily enable applications to leave the mainframe as move to it. For customers, now divided into three software camps, a single expanded strategic operating environment would ultimately provide the best access to commercial applications, support and a vast pool of experienced personnel. The disadvantages would be the cost of migration for VM and VSE users and the risk that software suitable for machines with large capacity might work poorly on modestly-configured models and thus demand big hardware upgrades. And for IBM as well as its customers, there is a real possibility that will not meet the ambitious goals it has set for itself. Even now there is evidence that IBM’s marketing machine may have overstepped the boundaries of credibility. Unfortunately, that is not a problem confined to IBM, but rather one of the most repulsive traits of the software business in general. IBM intends to have OS/390 certified as Unix by the fourth quarter of this year. All the Spec 1170 programming interfaces that define true Unix will be available to OS/390 applications.
By Hesh Wiener
It remains to be seen how efficient these programming interfaces will be in their initial release, but clearly IBM is determined not only to make them work but to make them work well. Using a tool kit developed by Bristol Software Technology Inc of Ridgefield, Connecticut – it is based on an established product called Wind/U – application programs written for Windows NT can be run under OS/390. The tool kit translates NT programming interfaces into Unix programming interfaces. Bristol has an NT source license that gives it access to all Microsoft Corp’s code, but as a practical matter it has not provided translations of every programming interface. Missing, for instance, are the service programming interfaces that are used by Web server programs. Presumably, users of OS/390 could obtain access to Unix versions of Netscape Communications Corp Web servers anyway, if the two vendors saw a significant opportunity and decided to co-operate. The one company mentioned in press releases about the Bristol deal was Britain’s National Westminster Bank Plc, which is moving toward a two-tier system composed of IBM mainframes running OS/390 and NT servers. Some applications may have to be made available on both systems, and Bristol’s tools could make this possible. This is a big opportunity for Bristol, which just set up an office in Amersfoort, the Netherlands, presumably on the assumption that NatWest’s strategy could be exemplary across Europe. Finally, IBM will launch its Net.Commerce Web Server in September. This is the software that is being used to book tickets for the Olympics, so it has already had a public shakedown. While IBM says it will deliver Net.Commerce for all its platforms, at first it will provide only AIX and NT versions. Net.Commerce for OS/390 and OS/400 are due by mid-1997. The product is a secure Web commerce server with related back-end components such as DB2/WWW, page creation tools and lots of security features. The commercial showcase site will be L L Bean, which intends to put its Christmas catalog on the Web using IBM’s software and hardware. The Bean configuration is expected to use RS/6000s to run Net.Commerce and provide firewalls in front of mainframes that will continue to do their usual jobs in support of order processing and fulfillment. But if OS/390 is to be Unix, why will it take IBM six months or more to make Net.Commerce available under OS/390? How long does it take to recompile a suite of programs? Something is obviously missing. We suspect that AIX has programming interfaces that extend it beyond the so-called standard Unix and that Net.Commerce uses these programming interfaces to achieve its power and efficiency. Even so, IBM, knowing Net.Commerce was of vital importance, could have added those programming interfaces to the other 1170 it is putting into OS/390. For the moment, we think it is fair to conclude that IBM is exaggerating the openness of OS/390 and the consistency of its C/C++ compilers fr om machine to machine. This is more than a bit of vaporware. Rather, it is an example of the failure of Lou Gerstner’s effort to rationalize IBM’s use of components – a failure tacitly acknowledged when the man granted dictatorial powers to oversee it was assignd to other duties. Just as the PowerPC chips were not good enough in their standard form to suit the AS/400 or to enable IBM to open another front in the battle for desktops, so, too, it seems that IBM’s effort to co-ordinate its many internal software development groups is far from successful. Gerstner may yet get his way, but customers who base their plans on IBM’s intentions rather than its proven accomplishments are naive. By every indication, IBM is still Babel and users mus t still evaluate its products on a product line-by-product line basis, even if the products share names like DB2 and Net.Commerce.
