Anybody using CLOE (was: LISP on PCs)

[gooch@TIJERAS.SW-SW.Dialnet.Symbolics.COM (William D. Gooch)]

    Date: Wed, 9 Jan 91 12:23+0100
    From: berni@iml.fhg.de (Stefan Bernemann)

	Date: Tue, 8 Jan 91 19:52 EST
	From: Qobi@ZERMATT.LCS.MIT.EDU (Jeffrey Mark Siskind)

	[...]
	Is
	anybody using CLOE? Are they happy with it?
		Jeff

    I'm also deeply interested in experiences (cheers? desperation?)
    from anybody using CLOE (on PCs. Anybody using it on UNIX boxes?).
    ....

Although I am not currently using CLOE, I had good success with it in
porting a neural net pattern recognition system from Genera to DOS for
MCC last year.  The development system was pretty solid at the time,
although CLIM was not yet ready for real use when I got it in the beta
version.  Instead, we used an MS-Windows lisp programmer interface
supplied by Symbolics, and were able to build our windows that way.  

By now, I would hope that CLOE CLIM is in a more fully usable condition.
The main problem with it was not just that the beta-test version had
bugs, but that the bugs were different between the CLOE development
environment in Genera and the CLOE delivery environment.  Debugging on
the PC is about as painful as you might expect, so the CLOE approach
works much better when code on the PC reliably behaves substantially the
same as it does in Genera (which of course is what it's supposed to do).

In case you like statistics: we ported around 500K bytes of source code in
about 6-7 man-weeks initially, and spent about one man-month adding a new
demo interface to it (we did not attempt to port existing interfaces,
because that would not have served any purpose).  The code initially ran
about 10x slower on a 20Mhz 386 than on a 3640, but after tuning to use a
387 coprocessor (with help from Symbolics), this improved by a factor of 4.