About KCP and automatic initialize

Richard A. O'Keefe ok at cs.otago.ac.nz
Wed Sep 17 07:33:46 UTC 2003


	One of the main argument against the KCP and automatic
	initialize seems to be related to the portability of
	Smalltalk code across dialects...
	
There is no argument against the KCP.
The argument against automatic #initialize is for the most part
not an argument against automatic initialisation.  If it was
automatic #initializeYourself, there'd have been a grumble-but-go-ahead.
(And then we would have missed out on Andreas Raab's wonderful suggestion.)

	Well, I really think this isn't such a big deal...

There are four Smalltalks available for my machines that I looked at
at one time or another:  VAST, VW, GNU, and Squeak.  But the one that
I care about most is Squeak.  I care about whether my students can
use, next year, a package I wrote this year.  I care about whether
they can use, next year, a tutorial that was written last year.  I
care about how much use the wonderful Ducasse collection of free
Smalltalk books is to them (because the Squeak books that I've had
our library buy don't really do the job).

	This issue isn't worse than the fact that Squeak,
	Dolphin, VAST and VW have different frameworks for the
	Web (VisualWave, Seaside, Web Parts, etc), for
	persistence, for UI (Morph, AbtParts, MVC, Pollock,
	etc) and the list goes on...

Yes it is.  We're not talking about "add on" features that you can
completely ignore.  We're talking about a method (Object>>new)
that is used by *most* classes.

	And making Squeak different doesn't mean it's a bad
	idea... 

No.  That's why there has been a lot of argument on quite other grounds
to show that it would be a bad idea even if no other Smalltalk existed or
ever had existed.

	Great ideas always resurface and get adopted... eventually.
	
So do bad ideas.  Andreas Raab's idea is a great idea.  Making a core
method that is d--ned hard to avoid automatically call a widely used
method is not a great idea.

	And if this community is clever enough to come up with
	tools like VMMaker, I bet we'll be able to extend the
	SIF tools to convert Smalltalk code to the "Squeak'
	way of initializing things"
	
Why?  We've never done that before.  We don't have any SIF tools
that convert anyone else's file opening code to the Squeak way of
opening files.



More information about the Squeak-dev mailing list