A cursory glance at the status of the once trumpeted ANDF Architecture Neutral Distribution Format efforts could easily be taken as an indication of the impending demise of the initiative. The Open Software Foundation has de-emphasised the Distribution Format within its programme and Novell Inc has deferred work on an ANDF implementation for Unix. Proponents say the picture is more complicated and that in fact the Distribution Format, which in any case means different things to different people, is caught up in the long-running debate about just what open systems ought to be. They argue that unilateral action by a vendor like Novell probably wouldn’t be effective in establishing the Architecture-Neutral Distribution Format as it was originally conceived, and now see a quite different and more subtle – though no less effective – route leading to its more widespread adoption.
Role to play
If an open computer systems and software market is deemed to include the supply of quality products made available by multiple vendors that will be able to work together by virtue of their adherence to standard application programming interfaces, then the Architecture-Neutral Distribution Format very definitely has a role to play, believes Paul Tanner, European Operations manager at X/Open Co Ltd, the Reading, Berkshire-based standards organisation which leads the effort to establish common application programming interface specifications based upon de facto industry technologies. The Common Open Software Environment grouping’s Common Desktop Environment and the Spec 1170 application programming interface are evidence of the increasing agreement that application programming interfaces must be standardised, and the greatest impact of a standard is felt where the standard is widely adopted and forms the boundary between products from different vendors. Buyers are used to going to more than one vendor to create an integrated computer system, but if the open systems market vision is to be achieved, Tanner says they will require increasingly sophisticated applications; standards covering all the application programming interfaces that will delineate applications from systems (including application programming interfaces to middleware required to support database access, distributed computing and systems management); ways to procure systems that support these standards; an application to be supported for a wide range of systems, and all at a reasonable price. Moreover, he argues, buyers do not care precisely how these objectives are achieved. The most common position is that it is ‘up to the industry’ to sort these things out, and the industry has to resolve a series of potentially conflicting interests if it is to do that.
By William Fellows
Issues such as product availability and the ability to respond quickly to market demand are becoming paramount. However, Tanner observers, where major investments are needed to move closer to the open systems vision, vendors are understandably reluctant to respond to ‘requirements’ like these unless they translate into a willingness to buy something additional. The the Architecture-Neutral Distribution Format was proposed by the Open Software Foundation as a means of providing an application with access to a wide range of disparate systems at a reasonable cost, and as such could address a component of the open systems vision. The concept is to split the compilation process into generic and architecture-specific parts. The intermediate architecture-neutral distribution format would be distributed instead of the multiple binaries needed at present. The the Distribution Format producer (the compiler front-end) would be used just once and the installer (the back-end) would reside on the target machine. By reducing the number of distributable versions, it was envisaged that the cost of converting and supporting an application on additional systems would be dramatically reduced, increasing the buyer’s choice. There were several steps conceived as being necessary to realise the potential of this
approach: identification of a two-stage compiler technology with sufficient performance and maturity; agreement on the right set of standards for the Distribution Format, especially the application programming interfaces to be supported; implementation of installers on a sufficient range of systems; and adoption by the providers of a sufficient number of applications. The first requirement has been met with the UK Defence Research Agency’s TenDRA technology.
Hurdle
The Open Software Foundation’s Application Environment Specification was not sufficient as the application programming interface set, and the final two conditions are a hurdle that won’t likely be crossed for some time. However, Tanner believes they can be, and given sufficient support from buyers they will be. X/Open, he says, is gradually answering the application programming interface question, although many third generation application programming interface issues are still unresolved, including system-to-system protocol interfaces, various language mechanisms and most importantly, the standardisation of object-oriented interfaces. Meanwhile, if buyers are convinced that adherence to a standard will give them a clear commercial benefit, they will make it a requirement, thereby putting a value on it, and demand like this will justify the necessary investment by suppliers to meet the remaining conditions, he argues.
