Re: <boolean> <true> & <false>

[alms@cambridge.apple.com (Andrew LM Shalit)]

At 26 Nov 16:50 GMT, eliot@dcs.qmw.ac.uk wrote:
> >
> >>   ...If you have separate classes for true & false then its possible
> >>   to define the basic control structures using polymorphism on #t &
> >>   #f and lambda (method)....
> >
> >While this is a simplification in theory, I don't think it would simplify
> >the practice of programming in Dylan.  It might be different in Self and
> >Smalltalk.
> >
> >If you allow the user to define methods on IF, then it also makes things
> >much harder for the compiler.
> >
> >   -Andrew
> >
> I'm not sugesting one replace the IF special form by methods.  The
> compiler can in-line the Smalltalk-style methods/lambda version of IF, but then
> debugging the code reveals that the compiler has treated the IF message
> specially.  So if one is concerned about efficiency then use hard-wired control
> structures.  
> 
> My feeling is that if you introduce a class <boolean> which allows
> you to specialise methods on <boolean> and non-<boolean> you should introduce
> <true> & <false>, since its likely you will want to specialise methods on
> <true> & <false>.

Do you want <true> and <false> to be subclassable?  If not, why not just use
(singleton #t) and (singleton #f) instead?  If so, do instances of subclasses
of <false> also appear as false to IF?