In <··········@fido.asd.sgi.com>, ·······@engr.sgi.com (Mike McDonald) wrote:
> I believe that the reason Lisp has failed to be widely accepted is that for
> most programmers, their first and only exposure to Lisp is in some rinky dink
> "Introduction to AI" class. At the end of which, if they're lucky, they might
> have written programs like Eliza and a checkers player. I can understand why
> they don't think Lisp is good for anything real.
Maybe that's true, but I believe the reason is much simpler: for many years
Lisp was used in the AI community, which, according to some sources, was
swimming in money. This enabled Lisp developers to concentrate on making
the language useful, rather than feel obliged to make Lisp systems run on
limited hardware, so Lisp was out of reach for most programmers. By the
time the typically available hardware caught up with Lisp's demand,
thinking patterns were set. Right now Lisp faces about the same problem C
had 20 to 25 years ago in converting all those FORTRAN and COBOL
programmers.
My first exposure to Lisp was in the late 60s when I found the Lisp 1.5
Programmers Manual in a bookstore. I liked what I saw, but alas, there was
no usable Lisp system available anywhere I worked. My first practical
experience was about 10 years later when I worked in a Unix shop with a
Vax780 and several PDP11s. The BSD we had on the Vax came with Franz Lisp,
and, remembering my earlier fascination with the book, I played around with
it. Again, I liked what I saw, the ease of developing code, the
interactive environment. The problem was, everything we wrote had to run
bot on the Vax and the PDP11s, franz wouldn't even start up on the PDP11.
Later I found some versions of xlisp (PC) to keep my interest alive. It
wasn't until about 5 years ago that Lisp and Scheme became available for
most of us in a useable form. Even now, CMUCL needs at least 32MB, the
ducumentation for MITScheme states they haven't tried it on machunes with
less than 24(?)MB, which means that most machines must be upgraded before
Lisp can be used.
> As for Symbolics part in the downfall of Lisp, sure, they contributed to the
> problem in some respects. They also built some of the nicest machines to do
Symbolics has been mentioned quite a lot in this thread. I have a question:
what was the typical price range for their machines, how many users could
one handle, what was the performance compared to other widely used machines
at that time (PDP11? Vax? DG?). The (snipped) comparison between Sun's
and Symbolic's policies leads me to believe that they price/performance
ratio wasn't all that good
> ...
> Anyway, that's history and it's too late to do much about it now. The real
> question now is can anything be done to increase the acceptance of Lisp as a
> viable programming solution? I'm rather pessimistic about it. Heck, GUI's have
I'm not so pessimistic. Over the years I have seen it over and over again
that new languages (and in many environments Lisp is new) had problems
getting their initial entries in the market. The languages that merited it
eventually did make it (and the GUIs followed). Maybe not Lisp in its pure
form, but a language like ML or Dylan?
> ...
>
> Mike McDonald
> ·······@engr.sgi.com
>
Hartmann Schaffer
In article <············@nic.wat.hookup.net>,
··················@wat.hookup.net wrote:
> most of us in a useable form. Even now, CMUCL needs at least 32MB, the
> ducumentation for MITScheme states they haven't tried it on machunes with
> less than 24(?)MB, which means that most machines must be upgraded before
> Lisp can be used.
It should be possible to use the MCL 4.0 development environment
(the RISC version) in 16 MB. I'm using more, since I'm loading
a large system (CL-HTTP) with source information (etc.) and
some other libraries. And then I want to run Netscape Navigator
on the same machine. ;-)
A "typical" Symbolics has 4 MW (40 Bits -> 20 MB) more memory as VM. The
complete OS (incl. development environment)
runs quite nicely in 8 MW with. A Lisp image then is around 50 MB,
incl. a lot of baggage (all the Flavors-based stuff, ...).
These are not frightening numbers nowadays. Boot time on my
MacIvory is around 3 minutes. These are machines have a processor
with 5-8 MIPS. Imagine how Lisp would run on them today, with
manufacturing processes resulting in 100-300 MIPS
processors. DEC took a 20 Mhz ARM processor and now it runs
at >160 Mhz with their process (the processor
is still using battery power). The relatively simple PowerPC design
from Exponential runs at 533 Mhz. A few years ago the PowerPC
started with 66 Mhz.
The virtual Lisp machine runs a lot faster on DEC Alpha,
than every Ivory ever built. Still, you could get
a boost from "just in time compilation", not to talk about a
native Alpha compiler.
> Symbolics has been mentioned quite a lot in this thread. I have a question:
> what was the typical price range for their machines,
Depends what machine. A board for the Mac for universities once (1990)
was sold for around $10000, a high-end XL1200 with HDTV video and
3d graphics packages was good for much more than $100000.
> how many users could
> one handle,
Effectively one.
> what was the performance compared to other widely used machines
> at that time (PDP11? Vax? DG?).
For Lisp they were great.
> The (snipped) comparison between Sun's
> and Symbolic's policies leads me to believe that they price/performance
> ratio wasn't all that good
Typical comparisons lead to this conclusion. If you
wanted to have a comparable Lisp development system on a SUN, then
I think the Lisp machine was similar priced and offered more value.
Your SUN needed >24MB, a larger disk then typical, a Lisp environment, ...
But then not everyone wanted to use Lisp, so a SUN was
*effectively* cheaper.
> I'm not so pessimistic. Over the years I have seen it over and over again
> that new languages (and in many environments Lisp is new) had problems
> getting their initial entries in the market. The languages that merited it
> eventually did make it (and the GUIs followed). Maybe not Lisp in its pure
> form,
I think for a wide range of applications Lisp (Scheme, CL, EuLisp, ...)
can be a good choice.
> but a language like ML or Dylan?
Software in ML could make a difference. Still you need
good commercial quality development systems and libraries.
Basically we need to get rid of crappy operating systems.
Create a move back from things like WIN32 or Unix.
Let these guys try to integrate not the other way round.
I know, there is not much future in this view. Will
Bill *ever* offer a radical different OS? Until now
we have 95, NT and CE.
--
http://www.lavielle.com/~joswig/
methinks should warn: Very bad grammar warning! <grin>
On Sat, 22 Feb 1997, Rainer Joswig wrote:
> In article <············@nic.wat.hookup.net>,
> ··················@wat.hookup.net wrote:
>
> > most of us in a useable form. Even now, CMUCL needs at least 32MB, the
> > ducumentation for MITScheme states they haven't tried it on machunes with
> > less than 24(?)MB, which means that most machines must be upgraded before
> > Lisp can be used.
>
> It should be possible to use the MCL 4.0 development environment
> (the RISC version) in 16 MB. I'm using more, since I'm loading
> a large system (CL-HTTP) with source information (etc.) and
> some other libraries. And then I want to run Netscape Navigator
> on the same machine. ;-)
CMUCL <something - 3.x I think> runs fine on my system (12M ram, 32M swap
[it doesn't get used], 486-100, Linux :)
I just don't know how to convince it to spit out a useable application.
(they have, like no ACTUAL advice on how to start). I read cltl2
inside-out, but it barely covers compiling and doesn't even mention an
EQUIVALENT to the 'C' main() function.... Very strange. I read Lisp
fine now, just don't know where to start writing it. Or writing a
compiler for it.
Lisp very nice, just very poorly documented if ya wanna write apps.
> > I'm not so pessimistic. Over the years I have seen it over and over again
> > that new languages (and in many environments Lisp is new) had problems
> > getting their initial entries in the market. The languages that merited it
> > eventually did make it (and the GUIs followed). Maybe not Lisp in its pure
> > form,
>
> I think for a wide range of applications Lisp (Scheme, CL, EuLisp, ...)
> can be a good choice.
I think I agree. Friendliness through tying code-generators into GUI
interface :drag effect to button, get code doing that, or something like
that. Easy to implement in Lisp. Not feasible with any other language.
So why code GUI-stuff in a language OTHER than Lisp? [makes no sense to
me - and I wrote mine in C/Assembly. Go figure]
> > but a language like ML or Dylan?
>
> Software in ML could make a difference. Still you need
> good commercial quality development systems and libraries.
>
> Basically we need to get rid of crappy operating systems.
> Create a move back from things like WIN32 or Unix.
> Let these guys try to integrate not the other way round.
> I know, there is not much future in this view. Will
> Bill *ever* offer a radical different OS? Until now
> we have 95, NT and CE.
[there's also Linux :]
I'm giving a shot to this - but I'm hosting under the Linux kernel (helps
me 'cause everything I use is Linux). Not touching any of the libraries
or the like though. Gives me networking, memory management, and hardware
management for most things. GUI is my own [any suggestions?] [it's
VR-based and quite fast :]. Haven't decided on any other details yet,
other than I plan on putting a number of virtual-machines in it for
various purposes (eg. running Java code fast enough to be useful).
So I'm not following the mighty Gates' footsteps <grin>.
... Only prob I can see with Linux at moment (other than having to code
all my own graphics drivers - ah well) is that this beast might not run
on all computers in market. Is that important?
Tchau!
- Teunis