wm (12/07/82)
There is a correspondence between object oriented systems with message passing, and the more conventional procedure oriented systems. Large systems can be constructed either way. A good reference on this is "On the Duality of Operating System Structures" by HC Lauer and RM Needham in the Proc. Second International Symposium on Operating Systems, IRIA, Oct. 1978, reprinted in Operating Systems Review, 13,2 April 1979, pp 3-19. It is interesting that this paper was written at Xerox Parc (yes, the Smalltalk people here are agreeing with the person who was flaming against Smalltalk!), you might be able to get a copy from them. On a slightly different tack, I have used object oriented software systems to develop different systems, and I have noticed that the effectiveness of this technique only goes so far. When you start building really large systems, building things so modularly starts getting in the way. I am a proponent of applicative styles, so I have to admit this, but you really start longing for a few global variables, or the ability to reach into the guts of a data structure without going through some module's interface. Have others noticed this? Do you feel that it is a fundamental problem or just an artifact of our current implementation technology? Wm Leler - UNC Chapel Hill
freund (12/12/82)
Certainly current programming technology affects the way that we think. In a very real sense, the meer fact that we have been exposed to the tools and techniques that are now prevalent shapes not only the way that we implement systems, but limits what types of problems we attempt to solve as well as our very understanding of the problems themselves. Object oriented architectures express a type of ideal. Either the *REAL* world is not so neatly organised or our collective exposure to non-object oriented structures has made us perceive the world in terms with which we are familiar. It may be, after all, that object oriented architectures are an attempt at enforcing what many consider to be good systems implementation style. In reality, all implementations are attempts to solve one sort of applications problem or annother. The best solution is one that actually solves the problem. Until we can "think" in terms of object oriented structures, architectures that force us into the mold will be clumsy to some degree. Bob Freund (..nsc!freund) National Semiconductor