Class 3 on September 29, 94

Paper Presented: 

Gilbert Cockton:
Spaces and Distances: Software Architecture and Abstraction and their
Relation to Adaptation

presented by: Bradley Fischer


Four discussion questions from Wednesday, 28-SEP-94 for CS8113D
       Walter Patterson / John Canning


[1] What is an atom?
      Discussion for this question focused on how small or how detailed
      can a menu system, etc, be decomposed before reaching the "atomic"
      level.  Does an atomic level exist?  Is there a theoretical limit?
      The class basically decided that:

       (a) You can't make the tiniest "atomic" components of an interface
           completely adaptable, because this will lead to confusion and
           lack of consistency.

       (b) a theoretical limit probably does not exist.
           Menus often contain sub-menus, references or
           links to other portions of the interface, etc.
           These links can be recursive or sometimes circular, and
           it is hard to define what the "bottom" level is anyway.

       (c) The question is moot, because the practical limit (what
           the user is willing and able to accept) is more relevant.

       (d) The practical limit depends heavily on user preferences
           and therefore can vary from user to user.


[2] Are the 5 variables proposed by Cockton ( agent, timing, mechanism, 
    components, composition of components ) valid for classification as
    mutually orthogonal axes?

      No, because they are NOT mutually orthogonal at all.  For example,
      "Components" and "Composition of Components" could be combined in
      some way, which reduces the space by one variable.  There is a great
      deal of interplay between the variables that prevents them from
      serving as mutually orthogonal axes.  
      There was also a general feeling that this classification is
      not very useful as a taxonomy in general, that in constructing an 
      adaptive system it is important to consider these variables together
      and their interplay, but that constructing a space in which to frame
      these variables doesn't buy you much.


[3] What is the difference between recomposition and simple editing?

      Recomposition is the addition or redefining of links between
      nodes, resulting in synthesis, rather than editing, which is
      simply changing the contents of a particular node.  Obviously,
      this question is one mainly of semantics.


[4] Do more complicated CRL libraries work as well as traditional GUIs?
      
      One person claimed that yes, you could do anything with a CRL library
      that you could with a GUI.  For example, you could build a large enough
      library of commands that all needed functionality is present.  Not 
      discussed, however, was rather or not this "works as well" in the
      sense that GUIs may be more intuitive.

      The most important thing, according to Dr. Schneider-Hufschmidt, is 
      that if you want a system to be self adaptive to a large number of 
      users, it won't work, you have to make smaller things in the system 
      adaptive.   A large number of users will have widely varying needs
      and the system cannot possibly include enough flexibility to adapt 
      properly to everyone.