From: Marek Lugowski
Subject: C++ vs. any Lisp on a 386/Windows/on top of OS2 -- recommendations?
Date:
Message-ID: <2mn3tp$2bs@news.acns.nwu.edu>
Hello World. Has anyone a word of advice, from specs/from experience, on
the following:
We are building a client/server system where the server is an RS/6000/320
and the client a remote (14.4k bps or better SLIP line) PC, a 386 running
Windows, possibly Windows over OS2, if network needs require it (printer
deamon, etc.). The 386 has 4 meg of RAM. :(
We are wondering what the trade-offs are in a) development time, b)
compiled image size, c) runtime of compiled image -- of any known Lisp in
the universe :) vs. C++. Is the penalty hit with Lisp such that C++ is our
only way to go? Or are the modern Lisp compilers for Windows efficient.
How do the developer/debugging facilities under Windows compare? We wish
to standardize on the same Lisp, if it is Lisp, on both RS/6000 and the
386. Any suggestions? Description of task follows:
The client application requires putting up buttons, navigating through a
browsing interface, as well as prompting the server for, receiving and
locally assembling RTF-->Postscript documents for local printing under
a local queue control. A configuration file would accompany every document
sent from the server.
Please respond to me and I will summarize any general insights to the net.
Thank you.
-- Marek Lugowski
Senior Programmer Analyst
Institute for the Learning Sciences
Northwestern University
1890 Maple Avenue
Evanston, IL 60640
(708) 467-1872 ·····@ils.nwu.edu <--
pref. addr, not reply
From: Henry G. Baker
Subject: Re: C++ vs. any Lisp on a 386/Windows/on top of OS2 -- recommendations?
Date:
Message-ID: <hbakerCn3HzC.Hs6@netcom.com>
In article <··········@news.acns.nwu.edu> ·····@casbah.acns.nwu.edu (Marek Lugowski) writes:
>We are building a client/server system where the server is an RS/6000/320
>and the client a remote (14.4k bps or better SLIP line) PC, a 386 running
>Windows, possibly Windows over OS2, if network needs require it (printer
>deamon, etc.). The 386 has 4 meg of RAM. :(
>We are wondering what the trade-offs are in a) development time, b)
>compiled image size, c) runtime of compiled image -- of any known Lisp in
>the universe :) vs. C++. Is the penalty hit with Lisp such that C++ is our
>only way to go? Or are the modern Lisp compilers for Windows efficient.
>How do the developer/debugging facilities under Windows compare? We wish
>to standardize on the same Lisp, if it is Lisp, on both RS/6000 and the
>386. Any suggestions? Description of task follows:
>The client application requires putting up buttons, navigating through a
>browsing interface, as well as prompting the server for, receiving and
>locally assembling RTF-->Postscript documents for local printing under
>a local queue control. A configuration file would accompany every document
>sent from the server.
Your first order of business should be to junk the 386 and buy a
486 with at least 8Megs of RAM. Windows is slow as molasses in
January on a 386. It has nothing to do with C++ v. Lisp. You can
now get an absolutely screaming complete 486 system for well under $3K.
Now that you have a decent machine and a decent amount of memory,
Lisp should run well, and you should have no need for C++.
From: Cyber Surfer
Subject: Re: C++ vs. any Lisp on a 386/Windows/on top of OS2 -- recommendations?
Date:
Message-ID: <Cn5oJ4.42t@cix.compulink.co.uk>
In article <················@netcom.com>,
······@netcom.com (Henry G. Baker) writes:
> Your first order of business should be to junk the 386 and buy a
> 486 with at least 8Megs of RAM. Windows is slow as molasses in
> January on a 386. It has nothing to do with C++ v. Lisp. You can
> now get an absolutely screaming complete 486 system for well under $3K.
I think you should generalise that advise. Just get hardware
that's as fast as possible - while not spending more money than
you have. The point at which the cost for the added performance
is too much varies from person to person.
You _can_ do reasonable work on a 386 - I do, for example. Ok,
a 20 Mhz 386 with 8 MB of RAM isn't the best you cam get, but
there's a a practical limit. The first computer I used could
run Lisp, despite being a mere Z80 (not even a fast one) with
16K of RAM.
> Now that you have a decent machine and a decent amount of memory,
> Lisp should run well, and you should have no need for C++.
Here's a quote from my favourite road movie (Two-Lane Blacktop):
"You can never go fast enough."
Demand will always increase to meet the capacity. Here in the UK
we have an expensive example of that principle, called the M25.
Computers are no different from cars IMHO, and computer languages
are just like motorways, or something like that.
We've had the C vs Lisp argument before. Just buy faster hardware
isn't an argument. It's just a bigger budget. My budget will only
grow as fast as my income will allow - it'll never be big enough
for the stuff I want to do, whatever language I use.
I agree about $3K being enough for a "decent system". I'm now
looking for someone who'll give me the money, but when they do,
they'll probably think I'm going to spend it on hardware to
develop C++ software with.
My argument in favour of Lisp/Smalltalk/Prolog etc is always that
incremental interactive compilers are more productive, and I prove
it by pointing to all the Basic programmers I know. "Now imagine
a language that can develop larger apps, and make it easy to
maintain them afterward", I might say. I still see jobs for C++
programmers, and lots of them.
Martin Rodgers
--- Cyber Surfing on CIX ---
In article <··········@cix.compulink.co.uk> ············@cix.compulink.co.uk ("Cyber Surfer") writes:
| I think you should generalise that advise. Just get hardware
| that's as fast as possible - while not spending more money than
| you have. The point at which the cost for the added performance
| is too much varies from person to person.
Don't forget, though, that the original article was discussing using
C++ as an alternative to Lisp because of the 386 client machine(s). At
least, that's how I read it.
As soon as labour costs are brought into the equation, hardware costs
pale. Take a software designer who is familiar with Lisp and and force
that person to use C++ because of performance problems with Lisp on
the specific target hardware, and you can bet that any perceived
savings on hardware upgrade costs are long gone by the time you get
the finished system. (Deadline slips are an average of, say, $1000 per
week of _direct_ costs per designer, never mind downstream effects of
the slip.)
Unless of course they have dozens or hundreds of these client target
machines at the 386 level, in which case they have my deepest
sympathy.
--
_____________________________________
_/ _/ _/ _/_/_/_/ _/_/_/_/ _/ |Disclaimer: |
_/_/_/_/ _/ _/ _/ _/ | |
_/ _/ _/ _/ _/ _/_/_/_/ _/ |Should you not agree with the above, |
_/ _/ _/ _/ _/ _/ |my employer has authorized me to |
_/ _/ _/ _/ _/_/_/_/ _/_/_/_/ |take the heat as they are not willing|
|to take responsibility for it anyway.|
···@Software.Mitel.COM | |
|_____________________________________|
From: Cyber Surfer
Subject: Re: C++ vs. any Lisp on a 386/Windows/on top of OS2 -- recommendations?
Date:
Message-ID: <CnGq39.30F@cix.compulink.co.uk>
In article <·················@picard.Software.Mitel.COM>,
···@Software.Mitel.COM (Kim Letkeman) writes:
> Don't forget, though, that the original article was discussing using
> C++ as an alternative to Lisp because of the 386 client machine(s). At
> least, that's how I read it.
Oh, I agree. I just see so many people suggesting that any 386
should be junked. It's very subjective, unless the machine simply
can't run the software at _any_speed_. I think of it like this,
if I can run it on my present hardware, then I probably don't need
anything faster. If it won't pay for itself by running faster, then
I _definitely_ don't need it. Any other way of looking at it could
just be an excuse to spend money, which makes it easier for hardware
and software vendors to make something new to sell us, whether we
need it or not.
> Unless of course they have dozens or hundreds of these client target
> machines at the 386 level, in which case they have my deepest
> sympathy.
That may depend on the app. I've seen a justification for not
supporting any version of MS Windows later than 2.11 by using
an app for a large bank as an example. That means nothing at all,
as nobody else will likely use an app written for one costomer,
however large they may be. That fact that it was bank tells you
very little about the hardware demands of the app.
I'm using at least one app that still supports machines older
than 10 years. As it's a comms app, that might not matter, but
I sometimes wish that apps like that (for DOS) could at least use
the hardware features of a 286, like memory protection and extra
memory addressing. Note I know and care little about the 286,
so I should restrict my comments to the 386. However, there's
even less software about for DOS/Windows that makes use of the
32bit features of the 386. That's changing, but not for DOS,
and it's too late anyway.
Lisp should be above all this, but I guess it's not, judging by
the number of times hardware creeps into discussions about a
tools for writing software. Either Lisp is too high level, or
not high level enough, or something else. I don't know which.
Martin Rodgers
--- Cyber Surfing on CIX ---
From: Henry G. Baker
Subject: Re: C++ vs. any Lisp on a 386/Windows/on top of OS2 -- recommendations?
Date:
Message-ID: <hbakerCnHMBp.HrH@netcom.com>
In article <··········@cix.compulink.co.uk> ············@cix.compulink.co.uk ("Cyber Surfer") writes:
>Oh, I agree. I just see so many people suggesting that any 386
>should be junked. It's very subjective, unless the machine simply
>can't run the software at _any_speed_.
The reason that the 386 should be junked isn't because of Lisp, but because
of Windows. Lisp works fine on a 386 with enough memory (4 Mb may be
a bit shy, but it works), but Windows doesn't. However, once you get a
machine big and fast enough to run Windows, then Lisp runs well, too.
I run Lisp on my 386 (w/ 16 Mbytes of memory) all the time, but under
SCO Unix, not Windows.
From: Cyber Surfer
Subject: Re: C++ vs. any Lisp on a 386/Windows/on top of OS2 -- recommendations?
Date:
Message-ID: <CnKF4D.9nn@cix.compulink.co.uk>
In article <················@netcom.com>,
······@netcom.com (Henry G. Baker) writes:
> The reason that the 386 should be junked isn't because of Lisp, but because
> of Windows. Lisp works fine on a 386 with enough memory (4 Mb may be
> a bit shy, but it works), but Windows doesn't. However, once you get a
> machine big and fast enough to run Windows, then Lisp runs well, too.
I use 8 MB, and the performace is noticably better than when I
only had 4 MB, but that based on the perfornce of SC++, not any
Lisp system. The choice of Liis is a big factor, just as it is
with C++. For example, the time I'm told that VC++ takes to build
a simple MFC program on a 50 Mhz 486 with 16 of RAM is about the
same as I find SC++ takes to build a simple MFC program on my
20 Mhz 386 with 8 MB of RAM.
I'll keep my machine and just look for a Lisp that will run well
on it. The editor of Windows Tech Journal reviewed Allegro CL/PC
last month, and altho he didn't test it on a machine such as mine,
he assured me that it would on it. Perhaps I should ask for a demo
before buying it, but I don't know if that's possible, and I won't
be buying a system at that price just yet, as I'm still paying for
my last set of hardware/software upgrades.
> I run Lisp on my 386 (w/ 16 Mbytes of memory) all the time, but under
> SCO Unix, not Windows.
When I can afford another 8 MB of RAM (I'll need a Zenith memory
card to install it, first) and SCO Unix, I'll also be able to use
NetBSD/Linux and CLISP for X-Windows. Guess which I'd be using.
Martin Rodgers
--- Cyber Surfing on CIX ---
From: Jeff Dalton
Subject: Re: C++ vs. any Lisp on a 386/Windows/on top of OS2 -- recommendations?
Date:
Message-ID: <CnqyKn.DLJ@cogsci.ed.ac.uk>
In article <··········@cix.compulink.co.uk> ············@cix.compulink.co.uk ("Cyber Surfer") writes:
>I'm using at least one app that still supports machines older
>than 10 years. As it's a comms app, that might not matter, but
>I sometimes wish that apps like that (for DOS) could at least use
>the hardware features of a 286, like memory protection and extra
>memory addressing. Note I know and care little about the 286,
>so I should restrict my comments to the 386. However, there's
>even less software about for DOS/Windows that makes use of the
>32bit features of the 386. That's changing, but not for DOS,
>and it's too late anyway.
>
>Lisp should be above all this, but I guess it's not, judging by
>the number of times hardware creeps into discussions about a
>tools for writing software. Either Lisp is too high level, or
>not high level enough, or something else. I don't know which.
I'm not sure what the problem is. I use the same Lisp on a 386
and a SCARC. But then, I'm not using MS-DOS.
-- jeff
From: Cyber Surfer
Subject: Re: C++ vs. any Lisp on a 386/Windows/on top of OS2 -- recommendations?
Date:
Message-ID: <CnxDLo.6C4@cix.compulink.co.uk>
In article <··········@cogsci.ed.ac.uk>,
····@aiai.ed.ac.uk (Jeff Dalton) writes:
> I'm not sure what the problem is. I use the same Lisp on a 386
> and a SCARC. But then, I'm not using MS-DOS.
The problem is that there isn't yet a Windoes version of CLISP.
I'd have to use Allegro CL/PC, which is way to much money for me
just now. So, I can choose between full graphical Windows support,
or plain old DOS.
Unix isn't an option for me yet.
Martin Rodgers
Cyber Surfing on CIX
--------------------
"First you see Pringle sweaters, then you wear Pringle sweaters,
then you eat Pringle sweaters, then you _be_ Pringle sweaters."
-- Pat Cardigan
From: David Brabant
Subject: Re: C++ vs. any Lisp on a 386/Windows/on top of OS2 -- recommendations?
Date:
Message-ID: <David.Brabant.682.2D9248FF@csl.sni.be>
>We are building a client/server system where the server is an RS/6000/320
>and the client a remote (14.4k bps or better SLIP line) PC, a 386 running
>Windows, possibly Windows over OS2, if network needs require it (printer
>deamon, etc.). The 386 has 4 meg of RAM. :(
386/4MB + Windows 3.1 + Lisp ? Arrghh ! Are you kidding ?
Any MINIMAL decent Lisp implementation (with a good programming environment,
CLOS, Emacs|DWIM ...) requires at least 64MB of memory :-) :-).
Mmmh. Seriously, think twice before doing your work in Lisp. I'm an
absolute Lisp lover but I've not yet encountered a decent PC/Windows
implementation and I don't hope to see one soon.
Your application seems to need some low level stuffs and you'll have
to develop DLLs to do most of your work. Lisp, in this case, will
be a nuisance. You'll just have a few functions encapsulating DLL
calls.
>We are wondering what the trade-offs are in a) development time, b)
>compiled image size, c) runtime of compiled image -- of any known Lisp in
>the universe :) vs. C++. Is the penalty hit with Lisp such that C++ is our
>only way to go? Or are the modern Lisp compilers for Windows efficient.
>How do the developer/debugging facilities under Windows compare? We wish
>to standardize on the same Lisp, if it is Lisp, on both RS/6000 and the
>386. Any suggestions? Description of task follows:
Here is the references for a Lisp/Clos system. I know nothing about it.
These data are extracted from an advertisement in MSJ (march 1994, pp
69). It seems that this Lisp runs on both Unix/Dos machines and that
the code written is totaly portable (you can move your application
without recompiling).
Lisp and Clos for Windows/DOS ($795)
Venue,
1549 Industrial Road, San Carlos, CA 94612,
800-228-5325. FAX 415-508-9770
>The client application requires putting up buttons, navigating through a
>browsing interface, as well as prompting the server for, receiving and
>locally assembling RTF-->Postscript documents for local printing under
>a local queue control. A configuration file would accompany every document
>sent from the server.
> (708) 467-1872 ·····@ils.nwu.edu <--
HTH
David
From: Cyber Surfer
Subject: Re: C++ vs. any Lisp on a 386/Windows/on top of OS2 -- recommendations?
Date:
Message-ID: <CnD0AI.1JK@cix.compulink.co.uk>
In article <··························@csl.sni.be>,
·············@csl.sni.be (David Brabant) writes:
> 386/4MB + Windows 3.1 + Lisp ? Arrghh ! Are you kidding ?
> Any MINIMAL decent Lisp implementation (with a good programming environment,
> CLOS, Emacs|DWIM ...) requires at least 64MB of memory :-) :-).
The reviewer of Allegro CL/PC in a recent issue of Windows Tech
Journal thinks that it should run fine on a 386 with 8 MB of RAM.
I'd love to test that, as that's the machine I use every day.
CLISP runs fine (under DOS) on this 386, tho I've yet to try it
under Windows, as I don't have the extra bit of software that allows
the DOSX to use DPMI, and there's no Windows hosted binary yet.
> Mmmh. Seriously, think twice before doing your work in Lisp. I'm an
> absolute Lisp lover but I've not yet encountered a decent PC/Windows
> implementation and I don't hope to see one soon.
That's subjective. I say the same thing about C++ enviroments and
compilers, and yet most C++ programmers love them. It could be coz
I compare C++ with Smalltalk and Lisp systems. Even mediocre Lisp
systems for Windows beat the C++ systems I've seen into a bloody
pulp - IMHO.
Martin Rodgers
--- Cyber Surfing on CIX ---
From: Zoltan Janosy
Subject: Re: C++ vs. any Lisp on a 386/Windows/on top of OS2 -- recommendations?
Date:
Message-ID: <janosy.83.000E5821@tsys.hit.bme.hu>
In article <··········@cix.compulink.co.uk> ············@cix.compulink.co.uk ("Cyber Surfer") writes:
>> 386/4MB + Windows 3.1 + Lisp ? Arrghh ! Are you kidding ?
>> Any MINIMAL decent Lisp implementation (with a good programming environment,
>> CLOS, Emacs|DWIM ...) requires at least 64MB of memory :-) :-).
>The reviewer of Allegro CL/PC in a recent issue of Windows Tech
>Journal thinks that it should run fine on a 386 with 8 MB of RAM.
>I'd love to test that, as that's the machine I use every day.
Yes, it does. I've been using it on a 386/40 128k cache machine with 8 MB
RAM, and about 10 MB virtual RAM, and it worked fine. However, with 4 MB
it is so slow, that I couldn't use it at all.
My experience with the runtime generator is that the "smiley" demo (a
smiling face bouncing in a window) when compiled produced some 700k of
code, which is not really acceptable. But maybe a very complicated program
would make up just the same amount, I don't know.
----------- Zoltan Janosy <······@tsys.hit.bme.hu> ------ #8^) -----------
Technical University of Budapest, Hungary, Department of Telecommunications
Sztoczek u. 2., H-1111 Budapest, +36-1-166 4011 / 2093, fax: +36-1-166 5824
From: Cyber Surfer
Subject: Re: C++ vs. any Lisp on a 386/Windows/on top of OS2 -- recommendations?
Date:
Message-ID: <Co1575.2G3@cix.compulink.co.uk>
In article <··················@tsys.hit.bme.hu>,
······@tsys.hit.bme.hu (Zoltan Janosy) writes:
> Yes, it does. I've been using it on a 386/40 128k cache machine with 8 MB
> RAM, and about 10 MB virtual RAM, and it worked fine. However, with 4 MB
> it is so slow, that I couldn't use it at all.
That'swhat I'd expect.
> My experience with the runtime generator is that the "smiley" demo (a
> smiling face bouncing in a window) when compiled produced some 700k of
> code, which is not really acceptable. But maybe a very complicated program
> would make up just the same amount, I don't know.
ISTR Actor adding a few hundred K in the .EXE, and a few more in
the image file. I wonder how Smalltalk/V 2.0 compares with Actor
4 and Allegro CL/PC? Probably pretty well. Smalltalk/V 1.0 run well
enough on my 386 when I only had 4 MB of RAM, and the stand-alone
image for a simple program is _much_ smaller than 700K, not counting
the shared libs for the Smalltalk/V 1.0 runtime.
Martin Rodgers
Cyber Surfing on CIX
--------------------
"First you see Pringle sweaters, then you wear Pringle sweaters,
then you eat Pringle sweaters, then you _be_ Pringle sweaters."
-- Pat Cardigan
From: Fenton Ho
Subject: Re: C++ vs. any Lisp on a 386/Windows/on top of OS2 -- recommendations?
Date:
Message-ID: <CnEAt9.60u@watserv2.uwaterloo.ca>
In article <··········@tove.cs.umd.edu>, John R. Bane <····@cs.umd.edu> wrote:
>I've actually used two Common Lisps on IBM-compatible platforms for major
>project development:
>
(stuff on Allegro deleted)
>
>* Venue Medley Common Lisp
>
>Good news: Solid CLtL2. Mature development environment; lots of tools and
>library modules. Portable across platforms without recompilation;
>look-and-feel across platforms is identical.
>
>Bad news: Can be slow; images are byte-coded ala Smalltalk. No CLIM.
>Runtime fees for application delivery.
>
>Quirks: Development environment is residential: preferred method of code
>editing is a structure editor (albeit a very nice one).
>
>These impressions are about a year old; contract the vendors for their
>latest-and-greatest stuff.
>
>Disclaimer: I used to work for Venue.
I don't work for Venue but I've been using Medley for the past year and
have been extremely pleased with the product. The development environment
is excellent and to me makes up for the "bad news" and the "quirks".
Just my two cents.
Fenton
---
Fenton Ho "Nothing so cold as closing the
PAMI Group heart when all we need is to
Department of Systems Design Engineering free the soul. But we wouldn't
University of Waterloo be that brave, I know"
···@watisit.uwaterloo.ca Toad the Wet Sprocket