Object Specialization

Object Specialization

[Bill Wolfe]

  An interesting paper entitled "Object Specialization" appears in
  ACM Transactions on Information Systems, April 1989, pages 103-122;
  it combines the efficiency of class-based systems with the flexibility
  of prototype-based systems, and provides an elegant solution to the 
  problem of "intersection classes" which eliminates the need for (and
  the desirability of) multiple inheritance.  Additionally, it provides
  an elegant way to model different perspectives on (or implementations
  of) a single real-world entity -- for example, a person's address might
  differ depending on whether that person is viewed as a consumer (home 
  address) or as a worker (business address).  

  This paper has led me to modify my previous position that multiple
  inheritance should be incorporated into Ada 9X; single inheritance
  via specialization at the object level appears to be a fundamentally 
  better solution, capable even of representing semantic constraints
  which cannot be represented by traditional class-based approaches.

  The power, flexibility, and efficiency of this new approach should 
  be very seriously evaluated during the design of Ada 9X's improved 
  support for the process of object-oriented software construction.


  Bill Wolfe, wtwolfe@hubcap.clemson.edu

Re: Object Specialization

[Markku Sakkinen]

In article <8388@hubcap.clemson.edu> wtwolfe@hubcap.clemson.edu (Bill Wolfe) writes:
>
>  An interesting paper entitled "Object Specialization" appears in
>  ACM Transactions on Information Systems, April 1989, pages 103-122;
> ...
>  This paper has led me to modify my previous position that multiple
>  inheritance should be incorporated into Ada 9X; single inheritance
>  via specialization at the object level appears to be a fundamentally 
>  better solution, [...]

I agree that the paper (by Edward Sciore) is interesting, although
not so easy to understand (I'll read it again when I have time).

I am not quite as enthusiastic about Sciore's model as Wolfe is.
For one thing, it is rather complicated, with _separate_ hierarchies
for objects, classes, and templates. It also looks definitely less
disciplined than conventional inheritance: new objects can be
connected to a "superclass part" of an existing object without
the original owner's knowledge or consent.

One of Sciore's goals is the integration of classes and prototypes.
Other, perhaps more promising approaches are e.g. in the recent work
of Lynn Stein and others at Brown University. We are also developing
some ideas at this corner of the world.

It was useful to point out the reference in this newsgroup.
I suppose the Transactions on _Information Systems_ is not the
highest priority reading that most OOP people pick up from the weekly
pile of journals. Read the paper and make up your own mind.

Markku Sakkinen
Department of Computer Science
University of Jyvaskyla (a's with umlauts)
Seminaarinkatu 15
SF-40100 Jyvaskyla (umlauts again)
Finland
          SAKKINEN@FINJYU.bitnet (alternative network address)