Re: CLIM on CMU CL

[Scott_Fahlman@SEF-PMAX.SLISP.CS.CMU.EDU]

Mark,

This all sounds pretty good.  I'd like to carefuly go over any copyright
notice you propose to put on the thing, just to make sure it doesn't tangle
us up in ways that will become a problem later on.  We'll have to discuss
what happens when someone else takes our code and builds their Common Lisp
on top of it, as several groups have done.  I guess the right idea is that
if they plan to sell the result, they have no right to include the CLIM
code unless they negotiate a separate deal with you, but if they put the
result back into the public domain (for example, some people at University
of Utah are doing a public-domain port of CMU CL to the H-P snake), then
the result would still be CMU CL and the CLIM code could still be part of
it.
    
        Well, if all these companies feel they must charge extra for CLIM, instead
        of bundling it, then this whole public-domain idea is pretty risky...

    I'm not sure I agree with this.  I don't think CLIM will make things any
    worse than they already are.  I presume that Lucid et al aren't worried
    about CMU CL because they think their product is still on balance a
    better value - smaller, better documentation, better support, whatever.
    It's quite likely that the same will apply to their CLIM - after all, I
    imagine that the best we'll be able to do here with you folks is to
    produce, about six months from now, a single back-end (probably CLX)
    port of the reference implementation on your (then brand new) native
    CLOS.  In that same timeframe, Lucid and Franz will probably have Open
    Windows and Motif adapting look-and-feel back-ends in highly optimized
    implementations.  (And keep in mind that one vendor - Symbolics -
    already bundles CLIM with their Lisps, Genera and CLOE.)

Lucid and Allegro don't view us as a threat because they know that their
big customers don't mind paying for a guaranteed level of support and
hand-holding.  I think that our support has usually been better than
theirs, at least for semi-sophisticated university users, but we don't
promise anything at all.  Then again, I wouldn't be surprised if some
company were to start selling CMU CL support as a separate service -- that
makes the game a lot more interesting.

In my view, if these companies do their jobs, CMU CL should actually be
good for them.  By making Common Lisp more available to universities, and
by solving some of Lisp's traditional problems, we help to enlarge the
total community and hold back the C.  The companies should have no trouble
shipping a Lisp that is better than ours: if they can't build something
better on their own, they can take our code and devote all of their
programming resources to adding value.  All we've done is to raise the
floor a bit -- the vendors can't stand pat with their mediocre old compiler
technology while something better is out there for free.  But competition
from C was putting pressure on them anyway.  (It looks like Lucid has
decided to switch most of their resources over to the task of making C more
usable.)

The problem with charging extra for add-ons is that it really ticks off the
more price-sensitive customers.  For example, back when CMU CL only ran on
the IBM RT, CMU bit the bullet and bought a site license for Allegro to
cover all the other machines.  But their package of nifty environment tools
was so costly that CMU couldn't justify buying it -- it's not essential,
after all.  So all our undergrads run the stripped down Allegro and wonder
why people keep raving about the wonderful programming environment provided
by Lisp.
    
    We've bounced the idea of simply putting the proto-version or even alpha
    of CLIM 2.0 fully into the public domain with most of our vendor
    licensees, and none of them has yet objected very strongly.  This move
    with you folks seems to be all around a better approach, though, in that
    we avoid commercial free riders (Ibuki, Gold Hill, Delphi, NEC, Sony,
    ...) - which is what the existing CLIM vendors really care about - and
    we can be fairly confident that CLIM will still look fairly good (= run
    well) on your Lisp, which can't be said for all the others out there.
    
Yes.  You'd get much of the same benefit by just putting out a
Copyrighted version with blanket permission for ALL free systems but no
commercial ones, but then there is the danger of people seeing bad ports
that don't work well.

    I'm skeptical that any undergraduate can keep a secret, but I'll trust
    your judgment.  I suppose the worst that can happen is that a rumor
    starts floating around the big Lisp world that "CLIM will soon be put in
    the public domain" with nothing more said than that, leaving the market
    all confused, and the commercial vendors upset with us (ILA).  I'd like
    to get a plan in place with you folks, then go around and sell it to the
    vendors, then explain it to the users, in that order.  
     
I'm not too worried about the discretion of our two undergrads.  What is
more likely to happen is that some of the preliminary technical discussions
about how to do this will spill out onto a mailing list with outsiders on
it.  If a port were to begin in earnest, people would certainly get wind of
it.  (Our "internal" mailing list includes ex project members who have gone
to various companies and universities.)  So you would have to make peace
with your vendors and users before we really crank up a big technical
effort.

One possible strategy would be to pre-empt a "public-domain" rumor by
telling people that you are exploring with us what technical problems would
be faced in porting CLIM to CMU CL.  You don't have to say that the result
would be free.  If anyone asks, just say that you are exploring the
technical issues, and that you haven't yet decided how this would be
marketed.  Just a thought...

-- Scott