[Date Prev][Date Next][Thread Prev][Thread Next][Author Index][Date Index][Thread Index]

Re: This Week (nondescript subject)

>From xanadu!vlad!mark Thu Jun 28 01:40:39 1990
	From: xanadu!vlad!mark (Mark S. Miller)
	To: xanadu!xanadu!roger
	Cc: xanadu!xanadu!bobp, xanadu!xanadu!xtech
	Subject:  This Week
	   Date: Tue, 26 Jun 90 12:41:52 PDT
	   From: xanadu!roger (Roger Gregory)
	   On rereading your message I was shocked to see that you might not be doing this 
	   in smalltalk, that is a big mistake, you will be buying a lot of incompatibility
	   during development.  Not using xpp is a lot worse, the communications code
	   will be evolving in a way that will be very cheap to -track- using xpp and 
	   VERY hard to do by hand.  That was the whole point of all our discussion about
	   initialized static globals and garbage collection, which will not be needed
	   if you don't use xpp, but which will not be a problem if you do.
	I see no significant problem with BobP not using Smalltalk (especially
	no incompatability problems).  However, I agree with you regarding the
	need to base things on the xpp foundations (tofu, bombs, garbage
	collection(?), & comm support).
Although I haven't been a part of the beginning of this thread, I
will throw in my $0.02.  I think mark is dead-on right.  Ceran and
Scheran work from the C++ generated from Smalltalk.  Once the tools
are polished (note that formic is *very* close to being there, although
the scripts can be somewhat tricky), tracking changes will be pretty
easy (as easy as things get in this business, anyway).  The key, however,
is buying into xpp; not doing so will mean constant vigilance over
whatever "conversion" code you create.  In Scheran, I think of it as
grafting:  I'm stub/proxy paralleling febe and most of xlatexpp,
tying into the "bottom" of xlatexpp and xpp.  The only "implicit"
conversion I've implemented is Fixnum/Bignum <-> IntegerVar, and
that'll probably change once the XOCers implement unbounded IntegerVar's.