[no subject]

[MOON@MIT-MC]

CC: (BUG LISPM) at MIT-MC
    CWH@MIT-MC 08/05/79 07:06:33
    To: MOON at MIT-MC
    CC: (BUG LISPM) at MIT-MC
    
        MOON@MIT-MC 08/04/79 18:09:51
            cwh@MIT-AI 08/04/79 15:17:49
            To: (BUG LISPM) at MIT-AI
            The instance variables of such widely used classes as WINDOW-CLASS and
            WINDOW-WITH-PC-PPR-CLASS should be interned on GLOBAL.
    
        Why?
    
    The reason for doing this is to have a consistent naming scheme for instance
    variables when defining methods.  If one is defining methods for a subclass
    of WINDOW-CLASS, then one must remember that NAME, SCREEN, LEFT, ... are on
    GLOBAL, and PROCESS, FRAME, and BIT-SAVE-ARRAY are on SI.  Alternately, one
    can simply say SI:NAME, etc. for all of the instance variables inherited
    from WINDOW-CLASS, but then one must distinguish between those variables
    inherited and those specific to the new class being defined.  I am assuming,
    of course, that the new class is being defined in a package other than SI.
    
    I think the solution RMS proposed of putting all instance variables on the
    keyword package is a reasonable one.  Besides, this would make our message
    passing code look more like LOGO!

My own preference would be to put all window methods in the SI package.
There's nothing wrong with users writing files of code that go in the SI
package, but aren't part of the initially-supplied system.  I don't feel
too strongly about this since it won't make message-passing code much
more unreadable than it is already.