Multiple polymorphism / multi-methods

Dave Ungar ungar at Eng.Sun.COM
Thu Mar 28 01:07:06 UTC 1991


	From mntgfx!dad.MENTOR.COM!plogan at uunet.UU.NET Wed Mar 27 16:43:48 1991
	Return-Path: <mntgfx!dad.MENTOR.COM!plogan at uunet.UU.NET>
	Message-Id: <9103272218.AA01106 at dad.MENTOR.COM>
	To: self-interest at self.stanford.edu
	In-Reply-To: Dave Ungar's message of Tue, 26 Mar 91 18:56:45 PST <9103270256.AA12884 at self.Eng.Sun.COM>
	Subject:  Multiple polymorphism / multi-methods
	Status: R

	   >From uiucuxc!eng.sun.com!ungar  Wed Mar 27 14:04:42 1991
	   I love the idea of putting lookup within the language.
	   Perhaps the VM shouldnt implement any kind of inheritance at all?

	Can you say Scheme? This is not an entirely facetious suggestion. It
	sounds like this discussion is leading to a small, regular easily
	optimized language that can be used to build higher-level language
	features like inheritance and polymorphic procedures.

	Excuse me for leading this perhaps too far astray. I don't want to
	start discussing Scheme (or compiling for continuations, etc.) but it
	is a reference point when discussing core languages vs. extensions
	that can be built with a core.


Nice point. Scheme is ALMOST right, but wrong enough to be very.
THe problem is: Scheme puts variable binding at the bottom, separate from 
function invokation. We want something as elegant, but with message-passing
(method invokation) at the bottom. This may be too cryptic or 
inflamatory--sorry if it's the former.

Dave



More information about the Self-interest mailing list