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

Re: [not about] Moving harddisk



    Date: Mon, 26 Aug 91 13:09 PDT
    From: JFK@BOLD-EAGLE.varian.dialnet.symbolics.com (Joe F. Karnicky)

	Date: Mon, 26 Aug 91 14:14:46 -0400
	From: hall@aplcen.apl.jhu.edu (Marty Hall)

	In message entitled "Re: [not about] Moving harddisk" on Aug 23, 
	Gooch writes:
	[...]
	> I realize you were just making a joke, but it made me think about how we
	> perceive our computers these days.  Ten years old and obsolete, even
	> though it still runs like (or better than) brand new and the software is
	> the best ever.  I see no real reason to buy an XL<whatever> (sorry about
	> that, Symbolics), until such time as I get work involving frame throwers
	> or other such stuff that doesn't work in my 36xx machines. [...]

	Our attitudes are interesting, I agree. A while back we got an XL1200
	with a FrameThrower to provide faster processing and color graphics for
	our current program. When I use it, do I think "Gee, this is fast"? Nope....

						- Marty

    There's another aspect to this issue that I run into on my lisp machine
    (a 3670) as well as other computing platforms I use:  As machines get
    faster, the software produced for these machines tends to be more
    demanding....

    Regards,
    Joe Karnicky

One theory has it that slower development machines make for better
(leaner, cleaner, more efficient) code.  I'm not too sure about this, but
I believe that there is a threshhold for each individual user - or more
precisely, a "wavefront" of performance threshholds across different
computation activities - beyond which investment the hardware tends to
reach rapidly diminishing returns.

Perhaps I shouldn't admit it, but when I am developing lisp code, I
usually don't need a machine much faster than a 3670 (especially if I
also have a spare machine nearby) in order to work close to my capacity.
Compiling, loading and testing may make me wait a bit, but they aren't
needed too frequently and can be done in parallel.  (Hm, I just realized
that this is the same sort of argument I've heard used many times in the
defense of switching from Symbolics to Unix on RISC machines.)

Just to be explicit about how I think a vendor should respond to this, 
I would recommend low-cost, compact products including the full-blown
software development environment, but with only moderate performance.
By "moderate" I mean "slow enough to keep the price down."  I would be
very interested in an xl400 level of performance in a desktop box for
under $10K, for example - wasn't this a planned product a few years ago?
A PC coprocessor UX board, or something like that?  Given the success of
the MacIvory, I would think the much larger PC market might be quite
lucrative.

I know the "Lispm for under $10K" idea is so old as to be almost cliche;
I just wonder why it never quite happened.