ASSERTION: A WILLINGNESS TO INVEST IN CROSS-FAMILY ARCHITECTURES WILL DISTINGUISH SUCCESSFUL FROM UNSUCCESSFUL IT USER ORGANIZATIONS

Syllabus: Systems developed over the next twenty years will continue a trend toward increasing system infrastructure. The software necessary to run a typical 1980s application consisted of approximately one part application (business rules, data descriptions and algorithms) to one part infrastructure (capacity to move and record data, manage system time, deal with hardware and software exceptions,translate among data formats, maintain object and class relationships, etc.). Today the balance between application and infrastructure has shifted so that a typical computer will dedicate at most a tenth of its resources to application specific processing, all the rest being used up by infrastructural processing. This balance will continue to shift toward infrastructure, mimicking the evolution of the human mind as a system in which application specific processing represents a miniscule fraction of necessary infrastructure.

Much of the infrastructure of any IT system is supplied by the operating system.. But part is domain specific. Domain specific infrastructure components will soon begin to dwarf the purely application-specific logic of any given IT system.

Organizations that use IT effectively will be those that invest in domain specific environments to support rapid development of systems needed to serve the domains. Such environments serve the role of cross-family architecture for systems in their domains.

Cross-family architectures are capital intensive. The cost of such an architecture will exceed by many times the cost of any given application that runs under it. Those organizations brave and aggressive enough to invest in cross-family architectures will reap a distinct advantage over those which continue the 1990s tradition of burdening each new application project with construction of its own domain-specific componentry.

-- Anonymous, October 22, 1999

Answers

My take on the architecture issue is that the division is somewhat different: (1) in-house developed, (2) externally purchased very large scale components, or (3) modified components working with the purchased components. However, any object technology we see is highly localized usage by a specific large-scale COTS vendor (SAP, PeopleSoft, J.D. Edwards, Seibel, etc.)

The problems that are most critical right now are that most customers are in the second phase of managing their large-scale purchased software. In the first phase, they highly customized the purchased software (option 3 above). Unfortunately, this left them unable to go from Version 5.3 to Version 6.0 of PeopleSoft (or made it very, very expensive.) As a result, many customer has tried to go cold turkey on modifications and run their purchased software in a vanilla configuration. As it turns out, they can't actually run their business with an out-of-the-box solution. Now they are looking for a third strategy, which in some cases is to throw out their packages.

-- Anonymous, October 22, 1999