FUNCTION-TYPE

[David A. Moon <Moon@STONY-BROOK.SCRC.Symbolics.COM>]

    Date: 06 Jun 87  1630 PDT
    From: Dick Gabriel <RPG@SAIL.STANFORD.EDU>

    I regard this issue as a put-up-or-shut-up affair. I prefer
    the formulation that states that APPLY and FUNCALL take functions,
    and you have to coerce symbols and lambda-expressions to be functions
    before you can APPLY or FUNCALL them to more lenient formulations.

    If we do not go with this formulation, then I would say an
    important design criterion of the cleanup is:

[updated version of the following quote from a later message substituted
for the original]

    ``Changes that introduce or retain a multiplicity of rules where a single
    rule will do are preferable to changes that cause old programs to stop
    working.''

Perhaps I should just keep my mouth shut, but I think that quite a
plausible case can be made that requiring an explicit coercion before
calling APPLY or FUNCALL would be introducing an extra rule and would be
more confusing to new users than permitting the coercion.  I don't think
there is one "right" position on this issue, I think it is a matter of
taste and style.  I don't think incompatibility with existing programs
is the primary point of resistance on this issue.

I'd really prefer that this matter of taste not hold up agreement on the
important FUNCTION-TYPE proposal.  Can we separate out the coercion of
symbols and lambda expressions by function calling into a separate issue?
Dick, I don't know what you meant by "put-up-or-shut-up"; would treating
this as two separate issues be somehow unacceptable to you?