Bitter as the battle for supremacy in the graphical user interface has been, the similar acrimonious split over a standard for distributed computing could be much more damaging for the industry. While users have proved able to put up with a choice between two user interfaces, two different standards for running distributed operating system environments would not only present serious compatibility problems for both manufacturers and users, but would be divisive for a community that is already reeling from the news that for the forseeable future there is to be no reconciliation between Unix International and the Open Software Foundation. Given the prospect of this scenario, temperatures are now rising across the industry in anticipation of the Open Software Foundation’s announcement to be made at its members meeting next week, where the choice of technology – or technologies – for its Distributed Computing Environment system will be revealed.
Sticky
Although 28 projects were originally submitted to the Request for Technology, the strongest contenders – and those making the most noise – are based around technology that supports Apollo’s Network Computing System, and Sun Microsystems’ Open Network Computing, and the former is though to have won (CI No 1,416). While Network Computing System has the backing of Foundation sponsors Hewlett-Packard, DEC and IBM in their submission known as DEcorum, Sun Microsystems and Netwise Inc, with systems based around Open Network Computing, are also believed to have been on the shortlist of contenders. The Open Software Foundation is in a sticky situation. For starters its sponsors have fared extremely well in its other technology choices so far – such as in the graphical user interface Request For Technology that led to Motif. Coming out in favour of its sponsors again could seriously damage the Foundation’s claim to have an open, and vendor-independent selection processes. To compound matters, there is reckoned to be some, not inconsiderable disagreement, between the Foundation’s business and technical fiefdoms about which path to follow. On the one hand Network Computing System as even Sun admits – has some technical advantages and extra features. On the other, the Apollo-developed technology still has some way to go before it is fully developed and stable, and Sun’s Open Network Computing-based technology is both widely used and has strong industry support from the likes of Novell Inc. Indeed the key Remote Procedure Call – RPC – and eXternal Data Representation – XDR – parts of Sun’s Open Network Computing were jointly developed in conjunction with AT&T and are included in Unix System V.4. If the Foundation opts for the Remote Procedure Call element of Network Computing System, it would put users of its distributed computing environment technology in danger of being incompatible with Unix System V.4 and its portfolio of distributed applications.
By William Fellows
Indeed, Eric Schmidt, vicepresident of Sun’s general systems group said, it would be a travesty if the de facto [Open Network Computing] standard is not approved and the Foundation goes and creates a competitor instead. One way that the Foundation may attempt to get around this, according to its European director Paul Wahl, is that more than one technology will be chosen. However the likelihood is that this could lead to more headaches than it cures. Merging the two technologies is likely to take at least a year – a year during which Sun’s Open Network Computing will become further established as a de facto standard – and even then, observers say, such an effort may mean having a less capable technology in the long run. Netwise’s Larry Lytle’s view is that it would depend on what they [the Open Software Foundation] have in mind, he added, we’d have to walk away from it if we had to merge with Network Computing System. Although the Network Computing System- and Open Network Computing-based submissions look the strongest contenders, newer technologies could be a practical – if less proven – means of incorporating
both. One such submission is a joint effort by Transarc Corp, IBM, Hewlett-Packard and Locus Computing, based around Transarc’s Andrew File System that was originally developed by Carnegie-Mellon University in Pittsburgh. It runs on top of the Network Computing System file system, but also includes support for Open Network Computing/Network File System clients. The Foundation is presently engaged in business negotiations with the various firms – yet another hot potato over marketing terms and rights, a choice is expected in Boston between May 15 and 17.
Rift
Looking further ahead, the decision will also have serious implications for de jure industry standards such as X/Open’s XPG. It has already laboured unsuccessfully for more than a year with rival offerings from the Unix camps to establish guidelines for developing graphical user interfaces. A similar rift over the choice of a distributed computing standard can only seriously undermine the credibility of an industry that is pressing for the adoption of open systems across the computer world. The problem of integrating the Remote Procedure Call elements of Network Computing System and Open Network Computing into one protocol set is not the time and expense required, but the fact that they have evolved from entirely different directions. However a good assessment of their relative merits and constraints is contained in an independent study of the two by Joshua Levy of Atherton Technology, Sunnyvale, California. Tests he carried out showed that Sun’s and Apollo’s Remote Procedure Call products performed comparably for small packets of data, but Apollo’s system degraded for increasingly large packets. However for lightly loaded machines, the two are equally dependable, but as the load increases, Sun’s system becomes less reliable. One of Levy’s notable findings is that most distributed programs can be modified to use any RPC system very quickly, in as little as a week.
