Okay ... I've been lurking here for a while and have been swayed by the
mountain of (mostly anecdotal) evidence that I should revisit my old LISP
haunts in search of development productivity. I've excavated and dusted off
my "Common Lispcraft" by Wallensky, and am making my way through it trying to
remember all the LISP I once thought I knew.
Now, though, I'm getting the feeling that I'm not writing LISP, but bad C++
and bad Perl, just with LISP syntax. The bibliography at www.lisp.org didn't
have anything that I recognized as this, but are there any books that discuss
the philosophy of LISP and/or thinking in LISP more than the syntax or
function library?
Thanks for your time,
+Mitchell
--
Mitchell Morris
X-Windows: It was hard to write; it should be hard to use.
-- Jamie Zawinski
Mitchell Morris wrote:
>
> Okay ... I've been lurking here for a while and have been swayed by the
> mountain of (mostly anecdotal) evidence that I should revisit my old LISP
> haunts in search of development productivity. I've excavated and dusted off
> my "Common Lispcraft" by Wallensky, and am making my way through it trying to
> remember all the LISP I once thought I knew.
>
> Now, though, I'm getting the feeling that I'm not writing LISP, but bad C++
> and bad Perl, just with LISP syntax. The bibliography at www.lisp.org didn't
> have anything that I recognized as this, but are there any books that discuss
> the philosophy of LISP and/or thinking in LISP more than the syntax or
> function library?
>
"On Lisp: Advanced Techniques for Common Lisp" -- Paul Graham, Prentice
Hall.
Mitchell Morris:
> have anything that I recognized as this, but are there any books that discuss
> the philosophy of LISP and/or thinking in LISP more than the syntax or
> function library?
Peter Norvig: Paradigms_of_Artificial_Intelligence_Programming shows
the philosophy of Lisp in action. The book presents both polished
code and tentative/partial solutions that lead to the final glory.
Cf. http://www.norvig.com/.
rthappe
···@unpkhswm04.bscc.bls.com (Mitchell Morris) writes:
...
> Now, though, I'm getting the feeling that I'm not writing LISP, but bad C++
> and bad Perl,
You mean that it is possible to write "good" Perl? :) :) :) <- that's
3 three 3 smileys. :)
Cheers
--
Marco Antoniotti ===========================================
PARADES, Via San Pantaleo 66, I-00186 Rome, ITALY
tel. +39 - 06 68 10 03 17, fax. +39 - 06 68 80 79 26
http://www.parades.rm.cnr.it/~marcoxa
From: Christopher R. Barry
Subject: Lisp and Perl. (was Re: Book suggestions)
Date:
Message-ID: <87zoymf4ar.fsf_-_@2xtreme.net>
Marco Antoniotti <·······@copernico.parades.rm.cnr.it> writes:
> ···@unpkhswm04.bscc.bls.com (Mitchell Morris) writes:
>
> ...
>
> > Now, though, I'm getting the feeling that I'm not writing LISP, but bad C++
> > and bad Perl,
>
> You mean that it is possible to write "good" Perl? :) :) :) <- that's
> 3 three 3 smileys. :)
If you keep your Perl scripts under 100 lines or so, and try not to
use the most disgusting syntax you possibly can (thus breaking from
what is traditionally practiced among "Perl hackers"), then yes. In
other words, if you use Perl as a _scripting_language_ for writing
_scripts_, and understand when a script is called for and when a
program is, you can use Perl productively. It provides far more
connectivity than all the add-on features of all commercial Lisp
systems combined, and there is a lot of easy stuff that is trivial to
do in Perl that would take great effort in Lisp. (Like a lot of basic
text processing tasks.)
I think Perl makes any task that could be construed as "easy, just
tedious" into "a simple cakewalk", while Lisp makes many "very
difficult" tasks "doable", but still leaves much to be desired for
many of the easy but tedious tasks....
Christopher
In article <·················@2xtreme.net>, ······@2xtreme.net (Christopher R. Barry) wrote:
> Marco Antoniotti <·······@copernico.parades.rm.cnr.it> writes:
>
> > ···@unpkhswm04.bscc.bls.com (Mitchell Morris) writes:
> >
> > ...
> >
> > > Now, though, I'm getting the feeling that I'm not writing LISP, but bad C++
> > > and bad Perl,
> >
> > You mean that it is possible to write "good" Perl? :) :) :) <- that's
> > 3 three 3 smileys. :)
>
> If you keep your Perl scripts under 100 lines or so, and try not to
> use the most disgusting syntax you possibly can (thus breaking from
> what is traditionally practiced among "Perl hackers"), then yes. In
> other words, if you use Perl as a _scripting_language_ for writing
> _scripts_, and understand when a script is called for and when a
> program is, you can use Perl productively. It provides far more
> connectivity than all the add-on features of all commercial Lisp
> systems combined,
Bullshit.
······@lavielle.com (Rainer Joswig) writes:
> > If you keep your Perl scripts under 100 lines or so, and try not to
> > use the most disgusting syntax you possibly can (thus breaking from
> > what is traditionally practiced among "Perl hackers"), then yes. In
> > other words, if you use Perl as a _scripting_language_ for writing
> > _scripts_, and understand when a script is called for and when a
> > program is, you can use Perl productively. It provides far more
> > connectivity than all the add-on features of all commercial Lisp
> > systems combined,
>
> Bullshit.
What is bullshit? I also find 'far more connectivity' somewhat
exaggregated, but otherwise I agree, although I prefer to keep my
perl scripts under 1 line or so ;-)
Speaking of connectivity: I'm looking for a gdbm, or generally,
*dbm-interface for CL (ACL5.0 if it's not portable).
--
(espen)
From: Fernando Mato Mira
Subject: Re: Lisp and Perl. (was Re: Book suggestions)
Date:
Message-ID: <37E25360.C6D58D2F@iname.com>
Espen Vestre wrote:
> Speaking of connectivity: I'm looking for a gdbm, or generally,
> *dbm-interface for CL (ACL5.0 if it's not portable).
Grab FFIGEN and adapt the `backend' to ACL.
Then set the backend free, so I can give a turnkey answer next time ;->
In article <··············@wallace.nextel.no>,
Espen Vestre <·····@*do-not-spam-me*.vestre.net> wrote:
>······@lavielle.com (Rainer Joswig) writes:
>
>> > If you keep your Perl scripts under 100 lines or so, and try not to
>> > use the most disgusting syntax you possibly can (thus breaking from
>> > what is traditionally practiced among "Perl hackers"), then yes. In
>> > other words, if you use Perl as a _scripting_language_ for writing
>> > _scripts_, and understand when a script is called for and when a
>> > program is, you can use Perl productively. It provides far more
>> > connectivity than all the add-on features of all commercial Lisp
>> > systems combined,
>>
>> Bullshit.
>
>What is bullshit? I also find 'far more connectivity' somewhat
>exaggregated, but otherwise I agree, although I prefer to keep my
>perl scripts under 1 line or so ;-)
I've done quite a bit of programming in both Lisp and Perl, and have been
able to be quite productive in both, in their proper contexts. Perl is a
very good match for the Unix environment; Unix used to be a "C Machine" (by
analogy to Lisp Machines), but nowadays I find it feels more like a Perl
Machine.
I've written some Perl scripts up to about 1,000 lines. One was a rewrite
of a C program at least twice its size (yet I was able to add several
features and make it easier to extend, and still achieve the same
performance), the other a replacement for several Informix ACE reports (the
Perl version is about half the size, due to modularity improvements, and
implements additional features).
Lisp and Perl both have their places. Lisp is best when dealing with
complex relationships among data. Perl is great when doing lots of string
processing. These Perl scripts are organized very similarly to how they
would be done if they were in Lisp, but I didn't have to hunt down a good
pattern matching library -- it's a built-in feature, since that's what Perl
was designed to support.
--
Barry Margolin, ······@bbnplanet.com
GTE Internetworking, Powered by BBN, Burlington, MA
*** DON'T SEND TECHNICAL QUESTIONS DIRECTLY TO ME, post them to newsgroups.
Please DON'T copy followups to me -- I'll assume it wasn't posted to the group.
In article <·······················@pbg3.lavielle.com>,
······@lavielle.com says...
> Bullshit.
Could you expand on that, please? It's a bit vague.
In my experience, Perl helps solve small nasty problems with quick and
dirty solutions. Lisp is great at creating elegant and maintainable
solutions to complex problems. Two different tools for different jobs.
Your experience may differ, of course. For years, I never encountered
problems that required quick and dirty solutions. Today I may find
myself fixing nasty problems created other people. This is where Perl
really shines, but it took me years to appreciate it.
If this is bullshit, then it's because there's a lot of it about.
OTOH, as you were rather vague, you may have meant something else.
Please consider this an opportunity to avoid misunderstandings.
--
Remove insect from address | You can never browse enough
will write code that writes code that writes code for food
From: Christopher R. Barry
Subject: Re: Lisp and Perl. (was Re: Book suggestions)
Date:
Message-ID: <87u2otfbh1.fsf@2xtreme.net>
······@lavielle.com (Rainer Joswig) writes:
> > If you keep your Perl scripts under 100 lines or so, and try not to
> > use the most disgusting syntax you possibly can (thus breaking from
> > what is traditionally practiced among "Perl hackers"), then yes. In
> > other words, if you use Perl as a _scripting_language_ for writing
> > _scripts_, and understand when a script is called for and when a
> > program is, you can use Perl productively. It provides far more
> > connectivity than all the add-on features of all commercial Lisp
> > systems combined,
>
> Bullshit.
Name one connectivity feature any Lisp system provides -- even the
Lisp Machine -- that is not provided by Perl. Oh, and don't try to be
a lawyer and say something like "the Lisp Machine provides Statice
connectivity", since for example SQL and DBM connectivity and
Oracle/Informix/etc is all that really matters. (And commercial Lisp
systems do provide good SQL connectivity as well as many other
important kinds -- somtimes for a lot of extra money -- but I'm
interested in you naming a connectivity feature that a Lisp system
provides that is not provided by Perl, since you are calling me a
liar.)
Christopher
From: Kent M Pitman
Subject: Re: Lisp and Perl. (was Re: Book suggestions)
Date:
Message-ID: <sfw7llpp3vp.fsf@world.std.com>
······@2xtreme.net (Christopher R. Barry) writes:
> Name one connectivity feature any Lisp system provides -- even the
> Lisp Machine -- that is not provided by Perl.
Transparent access to any file anywhere on the internet using a uniform
filename syntax.
Programming at the service level so that you can just ask that a piece
of mail be sent or a terminal window be established without caring what
protocol is used or what network medium that protocol goes over and the
system takes care of the details.
In article <···············@world.std.com>,
Kent M Pitman <······@world.std.com> wrote:
>······@2xtreme.net (Christopher R. Barry) writes:
>
>> Name one connectivity feature any Lisp system provides -- even the
>> Lisp Machine -- that is not provided by Perl.
>
>Transparent access to any file anywhere on the internet using a uniform
>filename syntax.
Is that really a Lisp system feature, or an Lisp Machine OS feature? It's
true that the Common Lisp pathname mechanism includes a hostname field, I
suspect it's not really useful except on systems that have a corresponding
OS feature. And if the OS has it, I imagine Perl would be able to take
advantage of it just as well.
--
Barry Margolin, ······@bbnplanet.com
GTE Internetworking, Powered by BBN, Burlington, MA
*** DON'T SEND TECHNICAL QUESTIONS DIRECTLY TO ME, post them to newsgroups.
Please DON'T copy followups to me -- I'll assume it wasn't posted to the group.
From: Kent M Pitman
Subject: Re: Lisp and Perl. (was Re: Book suggestions)
Date:
Message-ID: <sfwpuzhcc8f.fsf@world.std.com>
Barry Margolin <······@bbnplanet.com> writes:
> In article <···············@world.std.com>,
> Kent M Pitman <······@world.std.com> wrote:
> >······@2xtreme.net (Christopher R. Barry) writes:
> >
> >> Name one connectivity feature any Lisp system provides -- even the
> >> Lisp Machine -- that is not provided by Perl.
> >
> >Transparent access to any file anywhere on the internet using a uniform
> >filename syntax.
>
> Is that really a Lisp system feature, or an Lisp Machine OS feature? It's
> true that the Common Lisp pathname mechanism includes a hostname field, I
> suspect it's not really useful except on systems that have a corresponding
> OS feature. And if the OS has it, I imagine Perl would be able to take
> advantage of it just as well.
I don't agree. Common Lisp programs mostly transparently work on
arbitrary file systems even when they parse and manipulate filename
components on the lisp machine if they're written in the way that's
documented. I haven't don't much serious programming in Perl, and
even that not recently, so I'm not an expert in it, but I'll be
surprised if you tell me Perl uses other than string manipulation for
managing pathnames and so is going to lose when doing any kind of
component defaulting, pathname merging, search paths, etc. that work
across multiple file systems.
The critical thing is that the Lisp Machine is mapping from host names to
information about the host name's file system to mechanisms for parsing
pathnames properly for those hostnames.
In article <···············@world.std.com>,
Kent M Pitman <······@world.std.com> wrote:
>Barry Margolin <······@bbnplanet.com> writes:
>
>> In article <···············@world.std.com>,
>> Kent M Pitman <······@world.std.com> wrote:
>> >······@2xtreme.net (Christopher R. Barry) writes:
>> >
>> >> Name one connectivity feature any Lisp system provides -- even the
>> >> Lisp Machine -- that is not provided by Perl.
>> >
>> >Transparent access to any file anywhere on the internet using a uniform
>> >filename syntax.
>>
>> Is that really a Lisp system feature, or an Lisp Machine OS feature? It's
>> true that the Common Lisp pathname mechanism includes a hostname field, I
>> suspect it's not really useful except on systems that have a corresponding
>> OS feature. And if the OS has it, I imagine Perl would be able to take
>> advantage of it just as well.
>
>I don't agree. Common Lisp programs mostly transparently work on
>arbitrary file systems even when they parse and manipulate filename
>components on the lisp machine if they're written in the way that's
>documented.
Generic filename processing is not the same thing as transparent access
across the Internet. I agree that generic filename processing is pretty
much unique to Common Lisp.
>The critical thing is that the Lisp Machine is mapping from host names to
>information about the host name's file system to mechanisms for parsing
>pathnames properly for those hostnames.
True. But as far as I know, no other Lisp implementation does this. The
CL specification allows for it, but except for logical pathnames, I believe
they all just pass pathnames down to the OS's file system API. And I think
this is the right thing -- I don't think it's really appropriate for
network file system to be part of a language's runtime library, this is the
responsibility of the OS.
--
Barry Margolin, ······@bbnplanet.com
GTE Internetworking, Powered by BBN, Burlington, MA
*** DON'T SEND TECHNICAL QUESTIONS DIRECTLY TO ME, post them to newsgroups.
Please DON'T copy followups to me -- I'll assume it wasn't posted to the group.
From: Kent M Pitman
Subject: Re: Lisp and Perl. (was Re: Book suggestions)
Date:
Message-ID: <sfwogf025tj.fsf@world.std.com>
Barry Margolin <······@bbnplanet.com> writes:
> True. But as far as I know, no other Lisp implementation does this.
He said it was ok to have it just be the lispm or I wouldn't have mentioned it.
In article <···············@world.std.com>,
Kent M Pitman <······@world.std.com> wrote:
>Barry Margolin <······@bbnplanet.com> writes:
>
>> True. But as far as I know, no other Lisp implementation does this.
>
>He said it was ok to have it just be the lispm or I wouldn't have mentioned it.
I know that. But even on a Lisp I think it makes sense to distinguish
between features provided by the Genera OS and those that are part of the
"Lisp system". I realize the dividing line is fuzzy. The only way I can
think of finding the division is to think in terms of which things would
still remain even when programming in something other than Lisp on the Lisp
Machine. For instance, when using C or Fortran on a Lispm you still get
translarent access to remote filesystems, because their runtime libraries
also interface with the generic filesystem component of the OS.
So, on a Lispm, the things that are part of the Lisp system are the things
you can only do in Lisp. For instance, with respect to filenames, you
don't have access to pathname components in C/Fortran, like you do in Lisp;
in those other languages you have to treat them as strings, like you would
on another operating system.
--
Barry Margolin, ······@bbnplanet.com
GTE Internetworking, Powered by BBN, Burlington, MA
*** DON'T SEND TECHNICAL QUESTIONS DIRECTLY TO ME, post them to newsgroups.
Please DON'T copy followups to me -- I'll assume it wasn't posted to the group.
From: Kent M Pitman
Subject: Re: Lisp and Perl. (was Re: Book suggestions)
Date:
Message-ID: <sfwg109bksn.fsf@world.std.com>
Barry Margolin <······@bbnplanet.com> writes:
> In article <···············@world.std.com>,
> Kent M Pitman <······@world.std.com> wrote:
> >Barry Margolin <······@bbnplanet.com> writes:
> >
> >> True. But as far as I know, no other Lisp implementation does this.
> >
> >He said it was ok to have it just be the lispm or I wouldn't have
> >mentioned it.
>
> I know that. But even on a Lisp I think it makes sense to distinguish
> between features provided by the Genera OS and those that are part of the
> "Lisp system".
Yeah, but we draw the line in different places. Having dealt with the file
system a lot, I perceive the "operating system" level support as being each
of the individual file access paths. That is, the support for LMFS or FEPFS.
On Cadr-27, I think it was, there were multiple "file hosts", both LMFS (BSG's)
and the MIT file system (Stallman's?). I consider the language level support
to be what allowed programs to access each of those transparently by use of
appropriate file naming that selected one or the other of these.
> So, on a Lispm, the things that are part of the Lisp system are the things
> you can only do in Lisp.
Which includes the pathname system and the "host colon" convention, IMO.
The "file host" support I regard as OS-level, and I never programmed it
directly.
I would have said the dispatch mechanism for getting you the various hosts
was part of the OS, but for the fact that thwere were so many high-level
programming constructs for accessing that.
And, btw, very little of what you could do in Lisp couldn't be done in C or
Fortran on the Lisp Machine, too, with the right stuff loaded. That wasn't
wholly an aspect of operating system issues, but also with the integrated
environment.
One might say that this is entirely an issue of what the definition of an
operating system is--exactly the issues Gates is up against now.
My sense is that you would find an ally in him, in that he'd want to say
that some operating systems grow to include more things. For myself, I
prefer to think of the operating system as a relatively fixed set of basic
operations upon which many layers of applications are written. The reason
I prefer this model is that I think the very term "operating system" is
off-putting, and I think users are more comfortable fooling around in
"low-level applications" than "high-level operating systems", though I
admit there is little conceptual distinction. I think if you ever have a
level that users are intended to program below, it is not the operating
system.
From: Christopher R. Barry
Subject: Re: Lisp and Perl. (was Re: Book suggestions)
Date:
Message-ID: <877llo7mjb.fsf@2xtreme.net>
Kent M Pitman <······@world.std.com> writes:
> ······@2xtreme.net (Christopher R. Barry) writes:
>
> > Name one connectivity feature any Lisp system provides -- even the
> > Lisp Machine -- that is not provided by Perl.
>
> Transparent access to any file anywhere on the internet using a uniform
> filename syntax.
>
> Programming at the service level so that you can just ask that a piece
> of mail be sent or a terminal window be established without caring what
> protocol is used or what network medium that protocol goes over and the
> system takes care of the details.
You still have to tell the system what protocol to use though _before_
it "takes care of the details". At least that's how I remember it. For
example, I remember my first exposure to the network transparent
filesystem features was when I wanted to install CL-HTTP on the Lisp
Machine. The instructions were to type this to the Command Processor:
Restore Distribution :Use Disk Yes Pathname for dummy tape file:
ftp.ai.mit.edu:/pub/users/jcma/cl-http/lispm.reel-1
I thought that was one of the coolest things I've seen. It didn't work though.
Not until I told the system that FTP.AI.MIT.EDU was a FILE host using FTP via
IP-TCP using the Edit Namespace Object CP command or something. (My memory is
foggy.) It makes things easier for future users I suppose though by giving
them a uniform interface once the administrator has told the system how to
take care of the details....
If I have to know that I need to use FTP to get a file as well as its full
pathname on the system, I could just as easily tell Perl (or Netscape or wget)
to grap ftp://ftp.ai.mit.edu/pub/users/jcma/cl-http/lispm.reel-1. I don't
think the Lisp Machine's transparent filesystem access is as important now as
it was back in its heyday now that everyone has standardized on URLs and a few
transports over TCP/IP and UDP instead of myriad different services and
transports over DECNET (?) and CHAOS (or whatever).
Anyways, this feature qualifies as an answer to my original question, but I
don't see how it really saves time in a _modern_ networking environment.
Example scenarios are welcome. (If Symbolics never went under, I'm sure Genera
would have continued to stay ahead of everything else out there
connectivity-wise. Oh well.)
Also, it's worth mentioning that this _one_ thing is a feature unique to a
single, expensive, slow, and barely supported Lisp system that only runs on
expensive Alpha workstations or on aging hard-to-locate hardware. I'll happily
write and debug Lisp on my Lisp Machine day and night (it's much better for
that than anything else), but I wouldn't want to make it a component of a
production networking environment.
Christopher
From: Kent M Pitman
Subject: Re: Lisp and Perl. (was Re: Book suggestions)
Date:
Message-ID: <sfw1zbw3pfy.fsf@world.std.com>
······@2xtreme.net (Christopher R. Barry) writes:
> You still have to tell the system what protocol to use though _before_
> it "takes care of the details".
No, you have to on a one-time basis declare what protocols it is
able to use. If it knows at least one protocol that the remote host
supports, it doesn't fuss with you.
> If I have to know that I need to use FTP to get a file as well as its full
> pathname on the system, I could just as easily tell Perl (or Netscape or wget)
> to grap ftp://ftp.ai.mit.edu/pub/users/jcma/cl-http/lispm.reel-1.
You're misunderstanding. The system is telling you that the host is not
properly declared and allowing you a chance to correct that, and you're
confusing that with the user interface.
> Also, it's worth mentioning that this _one_ thing is a feature unique to a
> single, expensive, slow, and barely supported Lisp system
Was this some kind of one-two punch where first you allow this as an answer
and then you criticize the answer? I'm sorry I responded at all.
Perhaps next time I won't.
From: Christopher R. Barry
Subject: Re: Lisp and Perl. (was Re: Book suggestions)
Date:
Message-ID: <87g10b5kfn.fsf@2xtreme.net>
Kent M Pitman <······@world.std.com> writes:
> ······@2xtreme.net (Christopher R. Barry) writes:
>
> > You still have to tell the system what protocol to use though _before_
> > it "takes care of the details".
>
> No, you have to on a one-time basis declare what protocols it is
> able to use. If it knows at least one protocol that the remote host
> supports, it doesn't fuss with you.
I'm not clear how this statement contradicts the one it is replying
to. (But it's possible that the lack of intellectual stimulation I've
been getting lately has dimmed my wits....)
> > If I have to know that I need to use FTP to get a file as well as its full
> > pathname on the system, I could just as easily tell Perl (or Netscape or wget)
> > to grap ftp://ftp.ai.mit.edu/pub/users/jcma/cl-http/lispm.reel-1.
>
> You're misunderstanding. The system is telling you that the host is not
> properly declared and allowing you a chance to correct that, and you're
> confusing that with the user interface.
I'm not sure how this "that" I am confusing is distinct from the user
interface. The Lisp Machine doesn't do URLs (unless you've installed
CL-HTTP, which doesn't integrate as super-smooth with the system like
the rest of the Genera components, but I digress...), so instead of
including the protocol as part of the "unique resource namestring", you
tell the system to associate the protocol with the host and then when
you request a resource from the host a valid protocol you've associated
with that host is used.
Maybe I'm not being as clear as I could be. Let's take this scenario:
Say I want to get the file
lispm.reel-1
in directory
/pub/users/jcma/cl-http/
of host
ftp.ai.mit.edu
using protocol
FTP
To do this from the Unix command-line I could say:
wget ftp://ftp.ai.mit.edu/pub/users/jcma/cl-http/lispm.reel-1
To do this from the Lisp Machine (as best as I know how), it would be
something like [from memory; after a few brown-outs here I turned off
the machine and decided not to power it on again for fear of damage to
it until I get a UPS to protect it and I haven't had time to use it
lately anyways because of work and other crap (though I sure want to
use it again and plan to soon)]:
Edit Namespace Object
Create Object (blah) HOST (blah) FTP.AI.MIT.EDU
[The previous command was possibly several steps...]
<some create/add/edit command> (service) FILE (medium) TCP (protocol) TCP-FTP
[There may be additional stuff like "Network INTERNET", I don't remember.]
then finally you do something like
Copy File (a file) ftp.ai.mit.edu:/pub/users/jcma/cl-http/lispm.reel-1
To (blah) >cbarry>
I hope you understand where I'm coming from. Now if instead of me having
to do all that Edit Namespace stuff, it would be cool if the system
tried to figure it out by itself. It could determine that:
FTP.AI.MIT.EDU
is not on the local network, and is an INTERNET address, thus it is
perhaps a reasonable first guess that it is on the
INTERNET
network, so we are probably going to want to try the
TCP
medium first. And we are requesting a FILE, since we are using the
Copy File command, so we want the
FILE
service. The next part (determining the protocol) is the trickiest,
but the best first guess would probably be FTP since
* we are requesting a FILE
* there is a good extra hint from the host beginning with
FTP.*..., just like WWW hosts often begin with WWW.*....
failing that, HTTP could be tried, then something else.... Perhaps the
best thing would be if the system just asked which protocol to use (like
the Terminal program AFAICR does) instead of making you add a namespace
database entry for every single host you ever want to snarf a single
file from.
I'm not disagreeing about any connectivity functionality you said the
machine has; I'm saying that the functionality can be tedious to take
advantage of, and trying to illustrate why. You definitely can't do with
Unix something like
$ cp ftp.ai.mit.edu:/.... ~/...
[Minor Digression: you can do stuff like this with the GNU HURD....]
because cp isn't object-oriented like Copy File is. (cp isn't object
oriented on the HURD either; it's the same GNU cp as the Linux one, but
the OS uses a translator mechanism to feed cp a byte-stream from across
a network [AFAICR].)
> > Also, it's worth mentioning that this _one_ thing is a feature unique to a
> > single, expensive, slow, and barely supported Lisp system
>
> Was this some kind of one-two punch where first you allow this as an answer
> and then you criticize the answer? I'm sorry I responded at all.
> Perhaps next time I won't.
I'm sorry for criticizing your answer, which was correct, well
written, and in fact the only actual answer I got.
Christopher
In article <··············@2xtreme.net>, ······@2xtreme.net (Christopher R. Barry) wrote:
> To do this from the Unix command-line I could say:
>
> wget ftp://ftp.ai.mit.edu/pub/users/jcma/cl-http/lispm.reel-1
I don't know what you are doing, I'm typing the pathname
to "restore distribution" on my Lispm and it installs that stuff
directly from FTP. Super easy.
From: Kent M Pitman
Subject: Re: Lisp and Perl. (was Re: Book suggestions)
Date:
Message-ID: <sfwd7vevrjc.fsf@world.std.com>
We're talking at crossed purposes.
The lisp machine system isn't for the purpose of putting web files on the
wrong protocol and still letting you address them. If it were, it could
be easily extended to have the protocol name in the filename.
The purpose of the lisp machine system is to let you do "finger" on a machine
and have it be able to finger the person on a remote machine whether the
remote machine was running ascii-name, lispm-finger, etc. Or to be able to
get a network terminal on a remote machine whether the remote terminal is
running 3600-login, supdup, or telnet. Or to be able deliver mail whethern
the remote system runs SMTP or other protocols. Or to be able to read and
write files on the host file system whether the remote system runs FTP
or NFS or lisp machine NFILE or some other such thing. And the lispm system
does this regardless of whether the medium connection in between is the
internet or chaosnet. Neither the protocol nor the medium is part of the
address, it is a mechanism that can be used to get to the address.
Increasingly, though, because machines have been so bad about about
being able to negotiate protocols, the real world solution is to
promote the protocol to be user-visible so that you are used to saying the
protocol because there is no meaning to you saying "find a protocol that
would suffice to perform the service of getting me the following remote file"
nor "find a service that would suffice to get me a login connection to
the following remote machine". Instead, you say "get me an ftp service
to this remote host" or "get me a telnet connection to this remote machine".
It's a sad, sad, sad thing that the world has devolved to a point where
it's actually more natural for you to think of asking for such a thing
and you think of the lisp machine as less functional for not wanting you to
have to deal with that.
Kent M Pitman <······@world.std.com> writes:
> Increasingly, though, because machines have been so bad about about
> being able to negotiate protocols, the real world solution is to
> promote the protocol to be user-visible so that you are used to saying the
> protocol because there is no meaning to you saying "find a protocol that
> would suffice to perform the service of getting me the following remote file"
> nor "find a service that would suffice to get me a login connection to
> the following remote machine". Instead, you say "get me an ftp service
> to this remote host" or "get me a telnet connection to this remote
> machine".
Different protocols have different usage patterns, requirements and
performance envelopes, and it is often very beneficial to be able to
choose which protocol you want. It's not simply a matter of failing
to automate protocol negotiation. I'm assuming that the LispM
provides the abstraction you talk about for negotiating protocols for
performing these tasks, as well as the ability to require a specific
protocol to be used. That would indeed be marvelous.
So, where can I buy one? I already got 220V pulled into my new house,
just in case that is needed.
--
Craig Brozefsky <·····@red-bean.com>
Free Scheme/Lisp Software http://www.red-bean.com/~craig
"riot shields. voodoo economics. its just business. cattle
prods and the IMF." - Radiohead, OK Computer, Electioneering
From: Christopher R. Barry
Subject: FAQ: How to get a Lisp Machine. (was Re: Lisp and Perl. (was Re: Book suggestions))
Date:
Message-ID: <873dwa63b8.fsf_-_@2xtreme.net>
Craig Brozefsky <·····@red-bean.com> writes:
> Different protocols have different usage patterns, requirements and
> performance envelopes, and it is often very beneficial to be able to
> choose which protocol you want. It's not simply a matter of failing
> to automate protocol negotiation. I'm assuming that the LispM
> provides the abstraction you talk about for negotiating protocols for
> performing these tasks, as well as the ability to require a specific
> protocol to be used. That would indeed be marvelous.
>
> So, where can I buy one? I already got 220V pulled into my new house,
> just in case that is needed.
Seems to be asked every now and then. If you want to get a Lisp
Machine, one of the most affordable options is to purchase one from
Peter Paine in the UK. Unfortunately, international shipping and tax
(if you are in the USA) can be exorbitant -- matching or exceeding the
cost of the machine. Get started here:
http://www.abstractscience.freeserve.co.uk/symbolics/welcome.html
The next way would be to send an email to the Symbolics Lisp User's
Group mailing list, ····@AI.SRI.COM, stating what you are looking for.
With luck, someone not too far of drive away will have one they are
willing to part with for the right offer. (I got a really nice deal on
a XL1201 this way.)
You will also probably be contacted by David Schmidt in Burke, VA if
you post to the SLUG-list. If you want to actually officially purchase
a supported Lisp Machine from Symbolics Technology, Inc. then he is
your man (really nice guy). This will probably be the most expensive
option. His contact information is:
(minor spam-proofing)
dkschmidt at compuserve.com
+1 703 455 0430
You can also just go to http://www.symbolics.com and send an email to
Kalman Reti who will likely forward it to David.
As for the machine; They are kinda slow. So if you plan to do serious
work with one then you will be very frustrated with performance if
don't at least have one of the 20MHz rev 4a Ivory models. So get at
least one of those unless you can afford an Alpha + Open Genera 2.0
which will be 4 [21164 processor] to 8 times faster [21264 processor]
plus 10-20 times the memory and faster disk. (If you work for the US
gov. then Open Genera is free. Just ask Kalman Reti to send you the
CD. (Not sure why that is, but I remember hearing it.)) AFAIK, the
20MHz machines are:
MacIvory III (LispM-on-a-card for old 68k Macs)
XL1200
XL1201 (Smaller, "pizza-box" version of XL1200 [What I have])
NXP 1000 (Netboot, X-Window only LispM. I would recommend the
others over this one.)
I would recommend the MacIvory III the most because it is just a card
that fits into a little Mac that will draw very little power. Even the
"tiny" XL1201 is far, far larger than your average workstation with
it's _huge_ clunky cube monitor (which I've got a hunch is kinda far
from meeting TCO emissions compliance ;-) and about 4 sq. ft. of desk
space it needs (not including mouse and keyboard). It also draws a lot
more power than your average workstation, though it will at least work
with standard wall socket power.
Anyone reading this that is a big-time Lisp user that doesn't already
have one of these things should really try to go out of their way to
get one. It will be worth it. I promise. (Several of the
comp.lang.lisp regulars that are very knowledgable of Lisp still don't
seem to own one or have used one like Erik Naggum, Pierre Mai,
etc....) They would seem very good candidates for satisfaction and
happiness with one.
Christopher
From: Rainer Joswig
Subject: Re: FAQ: How to get a Lisp Machine. (was Re: Lisp and Perl. (was Re: Book suggestions))
Date:
Message-ID: <joswig-1909992254250001@194.163.195.67>
In article <·················@2xtreme.net>, ······@2xtreme.net (Christopher R. Barry) wrote:
> MacIvory III (LispM-on-a-card for old 68k Macs)
> XL1200
> XL1201 (Smaller, "pizza-box" version of XL1200 [What I have])
> NXP 1000 (Netboot, X-Window only LispM. I would recommend the
> others over this one.)
UX1200 (VME-Bus board for older SUNs, you can plug
multiple of these into one SUN)
> "tiny" XL1201 is far, far larger than your average workstation with
> it's _huge_ clunky cube monitor
There used to be an adapter for other monitors (like Sony Trinitrons).
From: Tim Bradshaw
Subject: Re: FAQ: How to get a Lisp Machine. (was Re: Lisp and Perl. (was Re: Book suggestions))
Date:
Message-ID: <ey3yae2sc2l.fsf@lostwithiel.tfeb.org>
* Christopher R Barry wrote:
> It also draws a lot
> more power than your average workstation, though it will at least work
> with standard wall socket power.
As far as I know almost all the symbolics machines run off standard
power, or at least they do in the UK -- I guess the current
requirements are higher in the US. I certainly ran a completely
stuffed 3650 (floating point, colour graphics, frame grabber, video
support, no slots free in the cardcage) in my house. It was kind of
noisy but it worked OK. I guess a similar machine with a couple of
eagles might be pushing it a bit, but even they don't eat that much
once they're spun up.
In response to a bit I elided about performance a 36xx is adequate if
you mostly program, and maybe cheaper & easier to come by than an
ivory. (No, I don't have one any more, it died (disk & screen)).
Who wants an average workstation anyway?
--tim
From: ··@nanostructure.utdallas.edu
Subject: Re: FAQ: How to get a Lisp Machine. (was Re: Lisp and Perl. (was Re: Book suggestions))
Date:
Message-ID: <ur9jtfu3c.fsf@nanostructure.utdallas.edu>
······@2xtreme.net (Christopher R. Barry) writes:
> You will also probably be contacted by David Schmidt in Burke, VA if
> you post to the SLUG-list. If you want to actually officially purchase
> a supported Lisp Machine from Symbolics Technology, Inc. then he is
> your man (really nice guy). This will probably be the most expensive
> option. His contact information is:
>
> (minor spam-proofing)
> dkschmidt at compuserve.com
> +1 703 455 0430
>
> You can also just go to http://www.symbolics.com and send an email to
> Kalman Reti who will likely forward it to David.
I would imagine that David Schmidt can get you a machine cheaper than
having it shipped from Peter Paine. Do you have evidence to the contrary?
> As for the machine; They are kinda slow. So if you plan to do serious
> work with one then you will be very frustrated with performance if
> don't at least have one of the 20MHz rev 4a Ivory models.
I strongly disagree. A 3640 (basically the slowest and least capable
in the 36xx line, which predates the Ivory line) is plenty fast enough
for development. I'm much happier on my 3640 than I am on my AMD K6/2
450 running Harlequin under NT (at work); Harlequin is a fairly nice
environment but suffers from being a Windows program. Technically the
hardware is much faster, but so what? The LispM is so nice to you
that the hardware speed doesn't really even matter. Not that I
wouldn't mind getting an XL1201--So if you get sick of your wimpy,
slow XL1201, Chris, I will happily take it off your hands and let you
get back to your lovely perl coding. =)
Mysteriously, though, my 3640 certainly seems to -page- faster than NT
does, even using much slower disks, much slower memory, and a much
slower processor. NT's paging system is hideous. Be afraid.
So get at
> least one of those unless you can afford an Alpha + Open Genera 2.0
> which will be 4 [21164 processor] to 8 times faster [21264 processor]
> plus 10-20 times the memory and faster disk.
I would much rather develop my code on an XL1201 or even a 3640 than
under Open Genera. There is something about having a real Symbolics
console. X keyboard mappings to LispM keys are tiresome. Also,
there's just something about having lisp all the way down to the
chips. Also, theres something really sickening about having Genera
run on top of Unix, as it does on the Alpha. I would feel much
happier if Open Genera ran natively on the Alpha.
Now for *running* my production code, sure, I'd want to run it on Open
Genera, freeing up my precious Lisp Hardware for more important
development tasks.
(If you work for the US
> gov. then Open Genera is free. Just ask Kalman Reti to send you the
> CD. (Not sure why that is, but I remember hearing it.)) AFAIK, the
> 20MHz machines are:
>
> MacIvory III (LispM-on-a-card for old 68k Macs)
> XL1200
> XL1201 (Smaller, "pizza-box" version of XL1200 [What I have])
> NXP 1000 (Netboot, X-Window only LispM. I would recommend the
> others over this one.)
>
> I would recommend the MacIvory III the most because it is just a card
> that fits into a little Mac that will draw very little power. Even the
> "tiny" XL1201 is far, far larger than your average workstation with
> it's _huge_ clunky cube monitor (which I've got a hunch is kinda far
> from meeting TCO emissions compliance ;-) and about 4 sq. ft. of desk
> space it needs (not including mouse and keyboard). It also draws a lot
> more power than your average workstation, though it will at least work
> with standard wall socket power.
The MacIvory III is definately a win for the space-concious, but again
you run into the problem of not having a real Symbolics console
(though you can get an ADB adaptor to use a Symbolics keyboard on the
mac). Also, since it has to share its console and all its I/O with
the Mac, the MacIvory tends to be much slower at screen and disk I/O
than an XL is).
Still, using MacOS to communicate with your LispM is much more
tasteful than using X and Unix.
Also, You sound like you are using an old console. Get you one of the
new-style 19" consoles, boy! They are really sweet, and are very
shallow. Talk to David Schmidt, he'll sell you one. Or get a CCU and
run a 21" color monitor on it.
>
> Anyone reading this that is a big-time Lisp user that doesn't already
> have one of these things should really try to go out of their way to
> get one. It will be worth it. I promise. (Several of the
> comp.lang.lisp regulars that are very knowledgable of Lisp still don't
> seem to own one or have used one like Erik Naggum, Pierre Mai,
> etc....) They would seem very good candidates for satisfaction and
> happiness with one.
>
> Christopher
I totally agree with this statement. Any serious Lisp hacker needs a
LispM, of any kind, even emulating it under Unix if thats all thats
available.
P.S. I don't know why everyone has this superstition that older
Symbolics machines somehow have to take special power (some people
even think it is 3-phase!). I don't know of any Symbolics machine
that requires anything other than 120V, single-phase, 30A max
(typically they only draw 9-13A I think though). Twist-lock
connectors are preferred, but that's $15 item at your local hardware
store.
From: Rainer Joswig
Subject: Re: FAQ: How to get a Lisp Machine. (was Re: Lisp and Perl. (was Re: Book suggestions))
Date:
Message-ID: <joswig-2109991159140001@pbg3.lavielle.com>
In article <·············@nanostructure.utdallas.edu>, ··@nanostructure.utdallas.edu wrote:
> Mysteriously, though, my 3640 certainly seems to -page- faster than NT
> does, even using much slower disks, much slower memory, and a much
> slower processor. NT's paging system is hideous. Be afraid.
Sigh. That's my experience, too. NT is a dog.
> The MacIvory III is definately a win for the space-concious, but again
> you run into the problem of not having a real Symbolics console
> (though you can get an ADB adaptor to use a Symbolics keyboard on the
> mac).
That's what I'm using. The keyboard is way cool on the Mac.
> Also, since it has to share its console and all its I/O with
> the Mac, the MacIvory tends to be much slower at screen and disk I/O
> than an XL is).
Well, you need to get an accelerated graphics card for the Mac.
The MacIvory generally is nicely integrated into the MacOS.
You can attach newer monitors easily, etc. My MacIvory currently
runs in a Quadra 950 using the internal graphics system and
a second ethernet card. It has a big power supply and cooling
is okay, too. You need a newer revision of the MacIvory 3
board to have it running in a Quadra.
> P.S. I don't know why everyone has this superstition that older
> Symbolics machines somehow have to take special power (some people
> even think it is 3-phase!). I don't know of any Symbolics machine
> that requires anything other than 120V, single-phase, 30A max
> (typically they only draw 9-13A I think though). Twist-lock
> connectors are preferred, but that's $15 item at your local hardware
> store.
Well, the older machines are drawing quite a bit electricity.
You can turn off your heating system, though.
From: Pekka P. Pirinen
Subject: Re: FAQ: How to get a Lisp Machine. (was Re: Lisp and Perl. (was Re: Book suggestions))
Date:
Message-ID: <ixlna01afl.fsf@gaspode.cam.harlequin.co.uk>
··@nanostructure.utdallas.edu writes:
> Mysteriously, though, my 3640 certainly seems to -page- faster than NT
> does, even using much slower disks, much slower memory, and a much
> slower processor. NT's paging system is hideous. Be afraid.
Don't know if 3640 does anything faster, but certainly smarter. The
virtual memory system is integrated with the garbage collector, so it
does clever things, like scanning pages for inter-generational
pointers before they're swapped out and not writing unused pages to
swap. We'd like to build a system like that, too, but on Windows,
we're at the mercy of the OS.
--
Pekka P. Pirinen
Adpative Memory Management Group, Harlequin Limited
Only fools learn by their experience; smart people use the experience
of others. - Bismarck
From: Dirk Fieldhouse
Subject: Re: FAQ: How to get a Lisp Machine. (was Re: Lisp and Perl. (was Re: Book suggestions))
Date:
Message-ID: <7saok9$o15@romeo.logica.co.uk>
In article <··············@gaspode.cam.harlequin.co.uk>, Pekka P. Pirinen
(·····@harlequin.co.uk) says...
>
>··@nanostructure.utdallas.edu writes:
>> Mysteriously, though, my 3640 certainly seems to -page- faster than NT
>> does, even using much slower disks, much slower memory, and a much
>> slower processor. NT's paging system is hideous. Be afraid.
>
>Don't know if 3640 does anything faster, but certainly smarter. The
>virtual memory system is integrated with the garbage collector, so it
>does clever things, like scanning pages for inter-generational
>pointers before they're swapped out and not writing unused pages to
>swap. We'd like to build a system like that, too, but on Windows,
>we're at the mercy of the OS.
If you're prepared to stick to W9x, it should be possible to hook into the
paging system. I don't know about the equivalent in NT.
--
Dirk Fieldhouse Logica UK Limited
··········@logica.com 75 Hampstead Road
c=gb;a=attmail;p=logica; London NW1 2PL
o=LOGICA;ou1=UK;s=fieldhouse UK
+44 (20) 7637 9111
From: Christopher R. Barry
Subject: Re: FAQ: How to get a Lisp Machine. (was Re: Lisp and Perl. (was Re: Book suggestions))
Date:
Message-ID: <87zoyg2bre.fsf@2xtreme.net>
··@nanostructure.utdallas.edu writes:
> I would imagine that David Schmidt can get you a machine cheaper
> than having it shipped from Peter Paine. Do you have evidence to the
> contrary?
Some of the price quotes I got from him on a few parts seemed a tad
steep. (Others though (like an external SCSI disk loaded with layered
products) seemed pretty good.) Example:
Refurbished 17" basic console: $500
Refurbished 19" premium console: $800
Unused new-style keyboard: $300
Used new-style keyboard: $150
Kinda expensive. A long time ago I got a quote from Peter Paine on a
complete 3620 system which included one of the 17" consoles for $300
or so. On his web page he has a loaded color 3650 with a Frame grabber
and everything you need for $490. Even though international shipping +
tax would probably be a few hundred dollars, you still would probably
be spending less.
> > As for the machine; They are kinda slow. So if you plan to do serious
> > work with one then you will be very frustrated with performance if
> > don't at least have one of the 20MHz rev 4a Ivory models.
>
> I strongly disagree. A 3640 (basically the slowest and least capable
> in the 36xx line, which predates the Ivory line)
The 3620 is the smallest, most minimalist of the 36xx line.
> is plenty fast enough for development. I'm much happier on my 3640
> than I am on my AMD K6/2 450 running Harlequin under NT (at work);
And how many hours of development on the 3640 are you doing per week?
If I'm using my XL1201 many hours per day (as I was earlier in the
summer), the performance of Apropos and DocEx searches and other
commands you often do starts to get really tiring, as does the
constant swapping. (Mine only has 4MW though. I bet swapping would be
a lot better with 8MW.)
> Harlequin is a fairly nice environment but suffers from being a
> Windows program. Technically the hardware is much faster, but so
> what? The LispM is so nice to you that the hardware speed doesn't
> really even matter.
I pretty much agree. I prefer the LispM on the 20MHz Ivory to
Franz-ELI/Ilisp/Lispworks on this 200MHz MMX Pentium machine for
programming and debugging.
> So get at
> > least one of those unless you can afford an Alpha + Open Genera 2.0
> > which will be 4 [21164 processor] to 8 times faster [21264 processor]
> > plus 10-20 times the memory and faster disk.
>
> I would much rather develop my code on an XL1201 or even a 3640 than
> under Open Genera. There is something about having a real Symbolics
> console. X keyboard mappings to LispM keys are tiresome. Also,
> there's just something about having lisp all the way down to the
> chips. Also, theres something really sickening about having Genera
> run on top of Unix, as it does on the Alpha. I would feel much
> happier if Open Genera ran natively on the Alpha.
These are just your own psychological/perception issues. [Well I'd be
happier with a native Open Genera as well -- but for performance
reasons.] The Open Genera keyboard mapping in general lets you do
things with a similar number of keystrokes as using a real Symbolics
keyboard by using the function keys + others. It definitely isn't as
nice/cool as the real keyboard, but you can still be very comfortable
and productive with it, unless your own perception issues prevent
that. [It is certainly better than Emacs, the next best thing.]
> Now for *running* my production code, sure, I'd want to run it on Open
> Genera, freeing up my precious Lisp Hardware for more important
> development tasks.
I sure wouldn't want to use [Open] Genera for running any production
application. Unless your system can afford to run 10-100 times more
slowly than what a modern multiprocessor Pentium II/III, Sparc, or
Alpha system running native applications can do.
> Still, using MacOS to communicate with your LispM is much more
> tasteful than using X and Unix.
Again, these are your own personal perception/psychological/preference
issues. Despite user-interface and programmer-interface Unix
brain-damage, it is still about the most stable and
customizable/extensible enviroment there is. [And it is currently the
most suitable host for the {GNU|X} Emacs environment.]
> Also, You sound like you are using an old console. Get you one of the
> new-style 19" consoles, boy! They are really sweet, and are very
> shallow. Talk to David Schmidt, he'll sell you one. Or get a CCU and
> run a 21" color monitor on it.
I'm interested in this adapter that lets you run ordinary monitors
with a Symbolics. I'll have to investigate cost and availability
furthur. (I get the feeling my 17" console is approaching the end of
its years....)
Christopher
From: ··@nanostructure.utdallas.edu
Subject: Re: FAQ: How to get a Lisp Machine. (was Re: Lisp and Perl. (was Re: Book suggestions))
Date:
Message-ID: <uogevey8w.fsf@nanostructure.utdallas.edu>
······@2xtreme.net (Christopher R. Barry) writes:
> Kinda expensive. A long time ago I got a quote from Peter Paine on a
> complete 3620 system which included one of the 17" consoles for $300
> or so. On his web page he has a loaded color 3650 with a Frame grabber
> and everything you need for $490. Even though international shipping +
> tax would probably be a few hundred dollars, you still would probably
> be spending less.
Well, it depends on what you are buying, I guess. And also you have
to take into account that SMBX's prices are for refurbished equipment,
while Paine's prices are not. And I really think that shipping
charges would negate Paine's price advantage in many cases (certainly
in the case of a 3650, perhaps not for a 3620).
Really, though, both of them are much more expensive than the
alternative: find one locally. Sometimes you can find them for free
or trivial cost (I bought a 3640 for $25 once!). Or there are various
sources for machines in the ~$300 price range in the US.
I don't consider the prices SMBX has for keyboards and consoles to be
that out-of-line. Both are high-quality items with a lot more
sophistication than similar items found on stock hardware. (Doesn't
the console even run its own version of LIL?).
> > > As for the machine; They are kinda slow. So if you plan to do serious
> > > work with one then you will be very frustrated with performance if
> > > don't at least have one of the 20MHz rev 4a Ivory models.
> >
> > I strongly disagree. A 3640 (basically the slowest and least capable
> > in the 36xx line, which predates the Ivory line)
>
> The 3620 is the smallest, most minimalist of the 36xx line.
Actually, the 3610 is the smallest, most minimalist of the 36xx line.
But that's not what I was saying. Both the 3610 and 3620 are
G-machines, which makes them inherently faster and more capable than a
3640, which is an L-machine (even though a 3640 has more bus slots
than both combined).
> > is plenty fast enough for development. I'm much happier on my 3640
> > than I am on my AMD K6/2 450 running Harlequin under NT (at work);
>
> And how many hours of development on the 3640 are you doing per week?
> If I'm using my XL1201 many hours per day (as I was earlier in the
> summer), the performance of Apropos and DocEx searches and other
> commands you often do starts to get really tiring, as does the
> constant swapping. (Mine only has 4MW though. I bet swapping would be
> a lot better with 8MW.)
Well, unfortunately I have not been able to spend as much time with it
recently; but I used to spend "many hours per day" on it--like most of
my waking hours every day, for about two years. It is completely
adequate in performance (not that I wouldn't -mind- having a faster
machine, but I don't sit there grumbling while waiting for it to look
up docs or aproposing or whatever), and as I mentioned earlier, Genera
is so comfortable to use, and is so nice to you, that speed isn't an
issue.
If speed were what really mattered, we would all be hand-coding
everything in assembly or C, and we would eschew the benefits of
high-level languages (as many fools do). But I think people who use
high-level languages should realize that the entire point of a HLL is
to make life more comfortable for the programmer, and to make features
more readily implementable, and that performance is secondary. I
personally believe that performance is a hardware issue, and should
never take priority over functionality. Which is why I'm totally in
favor of Lisp hardware and large-scale computing systems in
general--then you are able to get functionality AND speed. The 3640
was a really fast system for its time. (BTW, anyone got a Connection
Machine they are wanting to get rid of?)
Genera has more than enough functionality to make up for the speed
difference of the hardware.
I bet David Schmidt would trade you your processor board for one with
8MW for a reasonable cost. Life really is better with more memory.
> > Harlequin is a fairly nice environment but suffers from being a
> > Windows program. Technically the hardware is much faster, but so
> > what? The LispM is so nice to you that the hardware speed doesn't
> > really even matter.
>
> I pretty much agree. I prefer the LispM on the 20MHz Ivory to
> Franz-ELI/Ilisp/Lispworks on this 200MHz MMX Pentium machine for
> programming and debugging.
Then why are you arguing with me? You seem like you are wanting to
have things both ways: "LispM's are slow, I use Perl and Unix. DocEx
and APROPOS are too slow." and "I like my LispM, I prefer it for
programming and debugging." I don't think you can really have it both
ways, Chris. You need to decide which way you want to go with this.
If you want to be a loser, that's fine, there are many LispM-hungry
Symbolians that would love to have your machine, and would treat it
with the respect it deserves, and would reap the rewards that you are
too ambivalent and confused to accept.
> > Now for *running* my production code, sure, I'd want to run it on Open
> > Genera, freeing up my precious Lisp Hardware for more important
> > development tasks.
>
> I sure wouldn't want to use [Open] Genera for running any production
> application. Unless your system can afford to run 10-100 times more
> slowly than what a modern multiprocessor Pentium II/III, Sparc, or
> Alpha system running native applications can do.
Again, functionality takes priority over performance. Genera has way
too many facilities that would be painful to reimplement on another
system for me to want to use anything else, given the choice (I use
Harlequin here because I don't have that choice at work). I spend a
lot of time missing things that I'd really like to use, when I try to
do things at work. I figure out something that will get me by, and
make my projects work, but I know they aren't as nice as they would be
if I were doing them in Genera.
> > Still, using MacOS to communicate with your LispM is much more
> > tasteful than using X and Unix.
>
> Again, these are your own personal perception/psychological/preference
> issues. Despite user-interface and programmer-interface Unix
> brain-damage, it is still about the most stable and
> customizable/extensible enviroment there is. [And it is currently the
> most suitable host for the {GNU|X} Emacs environment.]
Ugh. I really don't think you can defend that statement. Unix
doesn't hold a candle to the customizability and extensibility of
Genera. And who would want to use Emacs if they had Zmacs? Unix does
have a small amount of customizability/extensibility, but what it has
is so twisted up with so much other crap, and it is so hard to get to
and frob, that it is useless. The only people who get anything done
with Unix are a bunch of monomaniacal unix weenies that don't have
anything better to do than spend long years making it behave.
The customizability and extensibility of Genera, by comparison, is
practically quivering with anticipation as it exposes all of its
private parts to your heart's desires.
From: Tim Bradshaw
Subject: Re: FAQ: How to get a Lisp Machine. (was Re: Lisp and Perl. (was Re: Book suggestions))
Date:
Message-ID: <ey34sgnj25d.fsf@lostwithiel.tfeb.org>
* Christopher R Barry wrote:
>> I strongly disagree. A 3640 (basically the slowest and least capable
>> in the 36xx line, which predates the Ivory line)
> The 3620 is the smallest, most minimalist of the 36xx line.
But a 3620 is a faster machine than a 3640. It's a 2nd-generation,
gate-array machine (g-machine I think was what they were called).
3620, 30 and 50 are all pretty much the same board set, and are
noticeably faster than the 40 and 70 which were the earlier board set.
The 3600 itself may be basically the same as the 40 and 70, or perhaps
it is different again. There were also subvariants, the 3645 and 3675
which are souped-up somehow (better memory interface?). There are
also ivory variants.
A 3630 stuffed with memory is a pretty nice machine. *If* a 20 has
enough slots it would be pretty nice too.
--tim
From: Fernando Mato Mira
Subject: Re: FAQ: How to get a Lisp Machine. (was Re: Lisp and Perl. (was Re: Book suggestions))
Date:
Message-ID: <37E8FBF1.E57D6324@iname.com>
Tim Bradshaw wrote:
> But a 3620 is a faster machine than a 3640. It's a 2nd-generation,
> gate-array machine (g-machine I think was what they were called).
Could one make a Symbolics clone out of a single [FP]GA these days?
Writing a LispM in VHDL would be fun.
Even better, imagine a CL->VHDL translator.. *-)
I vote for this!
The VHDL+CL combinaison is really hard to beat.
Marc Battyani
Fernando Mato Mira <········@iname.com> wrote in message
······················@iname.com...
> Tim Bradshaw wrote:
>
> > But a 3620 is a faster machine than a 3640. It's a 2nd-generation,
> > gate-array machine (g-machine I think was what they were called).
>
> Could one make a Symbolics clone out of a single [FP]GA these days?
> Writing a LispM in VHDL would be fun.
> Even better, imagine a CL->VHDL translator.. *-)
>
>
>
>
From: Paolo Amoroso
Subject: Popularity of Symbolics Lisp machines [was: Re: FAQ: How to get a Lisp Machine. (was ...)]
Date:
Message-ID: <37ec1b22.473405@news.mclink.it>
On Sun, 19 Sep 1999 20:25:47 GMT, ······@2xtreme.net (Christopher R. Barry)
wrote:
> Anyone reading this that is a big-time Lisp user that doesn't already
> have one of these things should really try to go out of their way to
> get one. It will be worth it. I promise. (Several of the
> comp.lang.lisp regulars that are very knowledgable of Lisp still don't
> seem to own one or have used one like Erik Naggum, Pierre Mai,
> etc....) They would seem very good candidates for satisfaction and
> happiness with one.
As far as I know, there were other Lisp machine vendors besides Symbolics,
i.e. Lisp Machines Inc., Xerox, Texas Instruments and Integrated Inference
Machines. Yet people seem particularly fond of Symbolics systems. Why? Was
Symbolics the market leader? Were its technologies significantly more
advanced than other vendors'? Are there interesting features or
technologies of Lisp machines other than Symbolics worth remembering? I'm
just curious.
Paolo
--
EncyCMUCLopedia * Extensive collection of CMU Common Lisp documentation
http://cvs2.cons.org:8000/cmucl/doc/EncyCMUCLopedia/
From: Barry Margolin
Subject: Re: Popularity of Symbolics Lisp machines [was: Re: FAQ: How to get a Lisp Machine. (was ...)]
Date:
Message-ID: <CC9G3.58$854.1724@burlma1-snr2>
In article <···············@news.mclink.it>,
Paolo Amoroso <·······@mclink.it> wrote:
>As far as I know, there were other Lisp machine vendors besides Symbolics,
>i.e. Lisp Machines Inc., Xerox, Texas Instruments and Integrated Inference
>Machines. Yet people seem particularly fond of Symbolics systems. Why? Was
>Symbolics the market leader? Were its technologies significantly more
>advanced than other vendors'? Are there interesting features or
>technologies of Lisp machines other than Symbolics worth remembering? I'm
>just curious.
Yes, Symbolics was the market leader. I don't recall ever seeing a Xerox
Lisp Machine or hearing of anyone outside of Xerox who had one (on the
other hand, I've used the Alto -- we had about 4 of them at MIT in 1980);
we all know how bad Xerox was about turning their computer research into
products (if they weren't, Xerox would be selling what Apple does, and
Microsoft Windows probably wouldn't exist). I remember meeting people from
IIM at X3J13 meetings, but I don't know if they ever got far beyond the
prototype stage. Either LMI or TI was probably second place in the market
(I think TI was mostly repackaging LMI technology).
Symbolics was very good at developing technology on the Lispm. Their UIMS,
which inspired CLIM, was incredible, and has some features that are still
unmatched in modern GUIs. I hope I don't start a flamewar with this
comment, but LMI was doing quite a bit of copying. I'm not accusing them
of copyright violation, but merely of reimplementing interfaces. But this
meant that they were generally a release behind in these advanced
features. I'm sure they had some innovations of their own, but I never
used Lambdas enough to get familiar with them. I'm not putting LMI down;
Symbolics was significantly larger than them, so they could accomplish
more. One thing I remember about Lambdas was that they had customizable
microcode and a microcompiler, which was supposed to be good for optimizing
the machine for specialized tasks.
--
Barry Margolin, ······@bbnplanet.com
GTE Internetworking, Powered by BBN, Burlington, MA
*** DON'T SEND TECHNICAL QUESTIONS DIRECTLY TO ME, post them to newsgroups.
Please DON'T copy followups to me -- I'll assume it wasn't posted to the group.
From: Rainer Joswig
Subject: Re: Popularity of Symbolics Lisp machines [was: Re: FAQ: How to get a Lisp Machine. (was ...)]
Date:
Message-ID: <joswig-2209992234540001@194.163.195.67>
In article <·················@burlma1-snr2>, Barry Margolin <······@bbnplanet.com> wrote:
> Yes, Symbolics was the market leader. I don't recall ever seeing a Xerox
> Lisp Machine or hearing of anyone outside of Xerox who had one (on the
> other hand, I've used the Alto -- we had about 4 of them at MIT in 1980);
I think Xerox Lisp machines were commercially available
earlier than the Lispms from Symbolics. The Xerox InterLisp
systems were nicer (smaller, more Personal Workstation like,
had a "nicer" UI, ...) and cheaper. They were successful in
research labs and in applications. Here in Germany for example
Siemens was distributing them under the Siemens label. Xerox also
already was an established company with a brand name - Symbolics
was a newcomer. One lab here in Hamburg had several of these systems
and one Symbolics 3600.
People were switching to Symbolics, because their Lisp machines
had more power (in terms of speed, disk space, word size, memory
size...), IIRC. Also as time went by, some of the Symbolics technology
was more attractive (networking for example) and Common Lisp
was more "compatible" with ZetaLisp than with InterLisp. InterLisp
was kind of a dead end. I think Xerox was on the
right track for a lot of things and keeping some of the spirit
from the InterLisp days (lots of influence from Smalltalk,
Xerox's UI research, etc.) would be worthwile. Symbolics
on the other hand had too much influence from MIT, so
their systems had a more "techie" approach and were often
not really targetting the (end) user - but the mythical Lisp (or AI)
programmer. When in doubt, Symbolics was using a file and
mixing Lisp code with configuration information - Xerox
would present a GUI-based dialog and store the configuration
behind the scenes (really simplifiying, but this I think
describes roughly one philosophical difference).
From: Espen Vestre
Subject: Re: Popularity of Symbolics Lisp machines [was: Re: FAQ: How to get a Lisp Machine. (was ...)]
Date:
Message-ID: <w6k8piqaxl.fsf@wallace.nextel.no>
······@lavielle.com (Rainer Joswig) writes:
> size...), IIRC. Also as time went by, some of the Symbolics technology
> was more attractive (networking for example) and Common Lisp
> was more "compatible" with ZetaLisp than with InterLisp. InterLisp
> was kind of a dead end.
The Xerox machines ran Common Lisp starting with software version 'lyric'
(released in '88, I think) which was 'Xerox Common Lisp' and not 'Xerox
Interlisp' anymore, although all of Interlisp was still accessible from
Common Lisp (simply as a package, as far as I recall).
My main problem with the Xerox machines were that they behaved badly
in stand-alone or small-network environments (without a file server),
probably because the Xerox people never used them that way. And my
biggest frustration was trying to get TCP/IP (and not only PUP) work
on them (*when* it worked, it was really fun, you could integrate
ftp servers into your filesystem).
--
(espen)
From: Rainer Joswig
Subject: Re: Popularity of Symbolics Lisp machines [was: Re: FAQ: How to get a Lisp Machine. (was ...)]
Date:
Message-ID: <joswig-2309991115080001@pbg3.lavielle.com>
In article <··············@wallace.nextel.no>, Espen Vestre <·····@*do-not-spam-me*.vestre.net> wrote:
> My main problem with the Xerox machines were that they behaved badly
> in stand-alone or small-network environments (without a file server),
> probably because the Xerox people never used them that way. And my
> biggest frustration was trying to get TCP/IP (and not only PUP) work
> on them (*when* it worked, it was really fun, you could integrate
> ftp servers into your filesystem).
Btw., there is an emulator of the Xerox Lisp Machines available.
http://Top2bottom.net/venue.html
It's a bit expensive, though.
If you download the "Grammar Writer's Workbench for Lexical Functional Grammar"
(http://www.parc.xerox.com/istl/groups/nltt/medley/) you can
check the environment out yourself
(see the License ftp://ftp.parc.xerox.com/pub/lfg/license.txt).
Rainer Joswig
From: Espen Vestre
Subject: Re: Popularity of Symbolics Lisp machines [was: Re: FAQ: How to get a Lisp Machine. (was ...)]
Date:
Message-ID: <w6yadyotoe.fsf@wallace.nextel.no>
······@lavielle.com (Rainer Joswig) writes:
> Btw., there is an emulator of the Xerox Lisp Machines available.
>
> http://Top2bottom.net/venue.html
good grief! I didn't know that thing still existed!
(..first&last time I used it was on a Sun 4/110 at Xerox PARC 11 years ago)
Btw, the name 'medley' was the name for the second release of XCL ('koto'
was the last Interlisp-only release, then followed lyric and then medley.
The medley release was actually quite good also on the 'real thing'.
--
(espen)
From: Tim Bradshaw
Subject: Re: Popularity of Symbolics Lisp machines [was: Re: FAQ: How to get a Lisp Machine. (was ...)]
Date:
Message-ID: <ey3r9jqggdc.fsf@lostwithiel.tfeb.org>
* Espen Vestre wrote:
> My main problem with the Xerox machines were that they behaved badly
> in stand-alone or small-network environments (without a file server),
> probably because the Xerox people never used them that way. And my
> biggest frustration was trying to get TCP/IP (and not only PUP) work
> on them (*when* it worked, it was really fun, you could integrate
> ftp servers into your filesystem).
This reminds me of an amusing thing. If you asked the right people at
PARC, there was actually an NFS package for them. If you could get
that to work (I remember it being a bit of a pain, as you had to do YP
and stuff), then on an 1186 an NFS-mounted drive from a Sun was *much
faster* than the local disk (but still probably much slower than a
reasonable floppy drive)
--tim
From: William Deakin
Subject: Re: Popularity of Symbolics Lisp machines [was: Re: FAQ: How to get a Lisp Machine. (was ...)]
Date:
Message-ID: <37EB8269.DA7D2804@pindar.com>
Tim Bradshaw wrote:
> YP a pain. I will never hear it said! ;)
Will
From: Tim Bradshaw
Subject: Re: Popularity of Symbolics Lisp machines [was: Re: FAQ: How to get a Lisp Machine. (was ...)]
Date:
Message-ID: <ey3yadyhef6.fsf@lostwithiel.tfeb.org>
* Barry Margolin wrote:
> Yes, Symbolics was the market leader. I don't recall ever seeing a Xerox
> Lisp Machine or hearing of anyone outside of Xerox who had one (on the
> other hand, I've used the Alto -- we had about 4 of them at MIT in 1980);
> we all know how bad Xerox was about turning their computer research into
> products (if they weren't, Xerox would be selling what Apple does, and
> Microsoft Windows probably wouldn't exist).
There were a fair number of Xerox D machines (several 10s, less than
100) at least in Edinburgh, and I suspect in other places too. Of
course, being Xerox, you couldn't actually get the good machines
(Dorados) but had to make do with lesser ones (dandelions, dandytigers
and most likely doves (? 1186s anyway), which were slower and, in the
case of the 1186s of significantly awful physical quality.
Although the window system was much more mainstream-looking than the
Symbolics one, they really were nowhere near symbolics in terms of
functionality or performance.
--tim
From: Espen Vestre
Subject: Re: Popularity of Symbolics Lisp machines [was: Re: FAQ: How to get a Lisp Machine. (was ...)]
Date:
Message-ID: <w6n1ueqb7a.fsf@wallace.nextel.no>
Barry Margolin <······@bbnplanet.com> writes:
> Yes, Symbolics was the market leader. I don't recall ever seeing a Xerox
> Lisp Machine or hearing of anyone outside of Xerox who had one (on the
> other hand, I've used the Alto -- we had about 4 of them at MIT in 1980);
we actually had *two* at the math department at the university of Oslo
(one 1109 and one 1186, on which I programmed *prolog* (Xerox Quintus
Prolog!) for my masters thesis), and I know of several others at
scandinavian universities. I also spotted a 1109 with a Siemens
label at the university of Saarbr�cken (I think the also had several).
> products (if they weren't, Xerox would be selling what Apple does, and
> Microsoft Windows probably wouldn't exist).
there are several possible variants here: They might also have joined
forces with apple and made Interpress to what Postscript became...
--
(espen)
From: Paul Fuqua
Subject: Re: Popularity of Symbolics Lisp machines [was: Re: FAQ: How to get a Lisp Machine. (was ...)]
Date:
Message-ID: <83btatmp4u.fsf@elissa.hc.ti.com>
Date: Wed, 22 Sep 1999 18:55:30 GMT
From: Barry Margolin <······@bbnplanet.com>
Subject: Re: Popularity of Symbolics Lisp machines [was: Re: FAQ: How to get a Lisp Machine. (was ...)]
Either LMI or TI was probably second place in the market
(I think TI was mostly repackaging LMI technology).
I remember a celebration within TI when we had sold more machines
(overall) than Symbolics. I don't think anyone else noticed. Symbolics
stuck with the market much longer, though, so they won in the end.
TI's initial Explorer, though "repackaged LMI technology," was the first
lispm to be deskside-tower-size instead of refrigerator-size, and run on
normal room power. While both TI implementations were very CADR-like,
in the sense of being 32-bit microcoded machines in the CADR tradition,
the later Lisp chip did add several hardware enhancements; we just
didn't go as far as Symbolics' 36- and 40-bit architectures.
That's probably the significant difference between Symbolics and the
rest: they did the breakthrough things, and did them first, and people
heard about it. I really liked the presenting listener and the command
processor, and envied the modularity and flexibility of the network
code, and loved the idea of having a Lisp card in a Sun that you opened
an X window to.
TI tended to concentrate on cost and size and ease-of-use and OEMs (a
substantial fraction of Explorers were sold through Sperry; LMI's
Lambda-E was a rebadged and remicrocoded Explorer). We had some nice
interface features for inexperienced users, but one could argue that
that was neglecting the power of the system in the hands of an expert.
We also had the distractions of a government contract for an embedded
military lispm, a multiple-parallel-DSP board, Lisp/Unix, parallel Lisp
boards in one big chassis, and private contracts for AI stuff. (It was
also suggested to me in private email that TI's lispm business was a
side trip on the way to the Unix-box business with the 1500s.)
Symbolics had a single focus and enough resources to go after it. TI
had the resources but not the focus, LMI the focus but not the
resources. Besides, Symbolics had Zippy.
Paul Fuqua
Texas Instruments, Dallas, Texas ··@hc.ti.com
From: Rainer Joswig
Subject: Re: Popularity of Symbolics Lisp machines [was: Re: FAQ: How to get a Lisp Machine. (was ...)]
Date:
Message-ID: <joswig-2309992121140001@194.163.195.67>
In article <··············@elissa.hc.ti.com>, Paul Fuqua <··@elissa.hc.ti.com> wrote:
> TI tended to concentrate on cost and size and ease-of-use and OEMs (a
> substantial fraction of Explorers were sold through Sperry; LMI's
> Lambda-E was a rebadged and remicrocoded Explorer). We had some nice
> interface features for inexperienced users, but one could argue that
> that was neglecting the power of the system in the hands of an expert.
> We also had the distractions of a government contract for an embedded
> military lispm, a multiple-parallel-DSP board, Lisp/Unix, parallel Lisp
> boards in one big chassis, and private contracts for AI stuff. (It was
> also suggested to me in private email that TI's lispm business was a
> side trip on the way to the Unix-box business with the 1500s.)
I still remember visiting the giant German computer fair
"CeBIT" years ago and seeing an Explorer board for the Mac. Well,
I haven't had the spare money to buy one at that time (or even
an MacIvory from Symbolics).
But I always remember the TI sales guy with his (slight)
"arrogancy" saying that he couldn't remember all of the 30000
Lisp functions. ;-)
> Symbolics had a single focus and enough resources to go after it. TI
> had the resources but not the focus, LMI the focus but not the
> resources. Besides, Symbolics had Zippy.
Two questions:
- wouldn't it be fun to have more info about the TI Lispm
days on the web?
- wouldn't it be cool if TI would release some of the "old" Lisp
technology - they won't use it anyway?
Rainer Joswig
From: Harley Davis
Subject: Re: Popularity of Symbolics Lisp machines [was: Re: FAQ: How to get a Lisp Machine. (was ...)]
Date:
Message-ID: <37ea79a8$0$233@newsreader.alink.net>
Barry Margolin <······@bbnplanet.com> wrote in message
······················@burlma1-snr2...
> In article <···············@news.mclink.it>,
> Paolo Amoroso <·······@mclink.it> wrote:
> >As far as I know, there were other Lisp machine vendors besides
Symbolics,
> >i.e. Lisp Machines Inc., Xerox, Texas Instruments and Integrated
Inference
> >Machines. Yet people seem particularly fond of Symbolics systems. Why?
Was
> >Symbolics the market leader? Were its technologies significantly more
> >advanced than other vendors'? Are there interesting features or
> >technologies of Lisp machines other than Symbolics worth remembering? I'm
> >just curious.
>
> Yes, Symbolics was the market leader. I don't recall ever seeing a Xerox
> Lisp Machine or hearing of anyone outside of Xerox who had one (on the
> other hand, I've used the Alto -- we had about 4 of them at MIT in 1980);
> we all know how bad Xerox was about turning their computer research into
> products (if they weren't, Xerox would be selling what Apple does, and
> Microsoft Windows probably wouldn't exist).
We had two Dandytigers (upgraded Xerox D-machines) at Raytheon in the mid
80's, and there were also a fair number of them at MIT (believe it or not)
at Sloan School, where, if I recall correctly, Tom Moran and his students
did a fair amount of work on next-generation e-mail filtering and sorting.
I also had one in my apartment for about a year doing offsite work for
Raytheon and my thesis. It was a good replacement for a radiator during a
cold Cambridge winter.
I loved those machines. Computing environments have been on a downward
trend in terms of programming tools since then.
-- Harley
From: His Holiness the Reverend Doktor Xenophon Fenderson, the Carbon(d)ated
Subject: Re: Popularity of Symbolics Lisp machines [was: Re: FAQ: How to get a Lisp Machine. (was ...)]
Date:
Message-ID: <w4o7llaaia0.fsf@nemesis.irtnog.org>
>>>>> "BM" == Barry Margolin <······@bbnplanet.com> writes:
BM> Yes, Symbolics was the market leader. I don't recall ever
BM> seeing a Xerox Lisp Machine or hearing of anyone outside of
BM> Xerox who had one...
I own a Xerox 1186. Unfortunately, I'm missing the keyboard, without
which it will not boot. And even when I had the keyboard, the machine
would lock up about five minutes after powering it up for the first
time. Too bad I can't get a hold of anyone at Xerox (or on c.l.l) to
help me get this beast fixed. I'd *love* to get this monstrosity
running again!
--
Rev. Dr. Xenophon Fenderson, the Carbon(d)ated, KSC, DEATH, SubGenius, mhm21x16
Pope, Patron Saint of All Things Plastic fnord, and Salted Litter of r.g.s.b
``The lyf so short, the craft so long to lerne.'' - Chaucer
From: Tim Bradshaw
Subject: Re: Popularity of Symbolics Lisp machines [was: Re: FAQ: How to get a Lisp Machine. (was ...)]
Date:
Message-ID: <ey31zbiaaa8.fsf@lostwithiel.tfeb.org>
* d wrote:
> I own a Xerox 1186. Unfortunately, I'm missing the keyboard, without
> which it will not boot.
I can *maybe* find you one, mail me if interested.
> And even when I had the keyboard, the machine
> would lock up about five minutes after powering it up for the first
> time. Too bad I can't get a hold of anyone at Xerox (or on c.l.l) to
> help me get this beast fixed. I'd *love* to get this monstrosity
> running again!
My experience of 1186s that did this is that you're doomed. Mine
failed in exactly this way and I spent far too many hours trying to
fix it. With a complete other machine to board-swap you can probably
find the dead board, but I didn't have one by the time mine died (it
was the last survivor of the tribe of Edinburgh dmachines...)
Better to get a symbolics I think, unless you have antiquarian
interests in the xerox machines.
--tim
From: Chuck Fry
Subject: Re: Popularity of Symbolics Lisp machines [was: Re: FAQ: How to get a Lisp Machine. (was ...)]
Date:
Message-ID: <37e944a4$0$204@nntp1.ba.best.com>
In article <···············@news.mclink.it>,
Paolo Amoroso <·······@mclink.it> wrote:
>As far as I know, there were other Lisp machine vendors besides Symbolics,
>i.e. Lisp Machines Inc., Xerox, Texas Instruments and Integrated Inference
>Machines. Yet people seem particularly fond of Symbolics systems. Why? Was
>Symbolics the market leader? Were its technologies significantly more
>advanced than other vendors'? Are there interesting features or
>technologies of Lisp machines other than Symbolics worth remembering? I'm
>just curious.
Yes, in my somewhat biased opinion, Symbolics was the market leader.
LMI and TI were selling repackaged MIT CADRs; the microExplorer was a
CADR-on-a-chip for all intents and purposes. Symbolics's first product,
the LM-2, was also a repackaged CADR, but they designed two new
architectures, the 36-bit 3600 and the 40-bit Ivory, specifically
intended as Lisp machines. The 3600 series (L-machine, G-machine) was
CISC; the Ivory was a RISC Lisp machine.
I never used a Xerox machine, but I have met people who did. I've never
heard of anyone who used an IIM machine.
Symbolics software led the *workstation* industry in the '80s. The
integrated networking environment was mostly inherited from the MIT Lisp
Machine project, but was extended and was one of the first OSes to
integrate the Internet, particularly the domain name system. You could
access files transparently whether they were in the next room, or a
continent away via FTP. Symbolics was first to deploy a generational
garbage collector, and the virtual memory paging system was integrated
with the GC for optimum performance. They shipped a hypertext system
(Document Examiner) in '86, long before the World-Wide Web was even a
gleam in Tim Berners-Lee's eye. CLOS is basically the result of merging
Symbolics's New Flavors with Xerox's Common LOOPS. CLIM is the second
UI system of the folks who designed Symbolics's Dynamic Windows,
released again around '86. I think Statice was the first commercially
available object-oriented DBMS in the very late '80s.
Of course, they did a lot wrong too, mostly in the marketing and
financial arenas, or they'd be where Sun is now... think how different
the world would be had Symbolics won instead of Sun.
As I said, I'm biased; I worked at Symbolics during the glory days of
the mid-'80s. It was a wonderful experience, and I am glad I was there.
I've worked with some extremely talented people over the years, but I
don't think I've ever worked, or will ever work again, with a software
team as talented and capable as SCRC had in the mid-'80s.
-- Chuck, waxing nostalgic
--
Chuck Fry -- Jack of all trades, master of none
······@chucko.com (text only please) ········@home.com (MIME enabled)
Lisp bigot, mountain biker, car nut, sometime guitarist and photographer
The addresses above are real. All spammers will be reported to their ISPs.
From: Tim Bradshaw
Subject: Re: Popularity of Symbolics Lisp machines [was: Re: FAQ: How to get a Lisp Machine. (was ...)]
Date:
Message-ID: <ey3u2omghoe.fsf@lostwithiel.tfeb.org>
* Chuck Fry wrote:
> LMI and TI were selling repackaged MIT CADRs; the microExplorer was a
> CADR-on-a-chip for all intents and purposes. Symbolics's first product,
> the LM-2, was also a repackaged CADR, but they designed two new
> architectures, the 36-bit 3600 and the 40-bit Ivory, specifically
> intended as Lisp machines. The 3600 series (L-machine, G-machine) was
> CISC; the Ivory was a RISC Lisp machine.
LMI also worked on new architectures, including a final one (the k
machine?) which looked a whole lot more advanced than Ivory, really an
almost reasonable RISC one-instruction-per-clock load-store design I
think, (though I've seen better descriptions of it than I have ivory).
Some board sets were even made I believe.
--tim
From: Chuck Fry
Subject: Re: Popularity of Symbolics Lisp machines [was: Re: FAQ: How to get a Lisp Machine. (was ...)]
Date:
Message-ID: <37ea4f87$0$222@nntp1.ba.best.com>
In article <···············@lostwithiel.tfeb.org>,
Tim Bradshaw <···@tfeb.org> wrote:
>* Chuck Fry wrote:
>> LMI and TI were selling repackaged MIT CADRs; the microExplorer was a
>> CADR-on-a-chip for all intents and purposes. Symbolics's first product,
>> the LM-2, was also a repackaged CADR, but they designed two new
>> architectures, the 36-bit 3600 and the 40-bit Ivory, specifically
>> intended as Lisp machines. The 3600 series (L-machine, G-machine) was
>> CISC; the Ivory was a RISC Lisp machine.
>
>LMI also worked on new architectures, including a final one (the k
>machine?) which looked a whole lot more advanced than Ivory, really an
>almost reasonable RISC one-instruction-per-clock load-store design I
>think, (though I've seen better descriptions of it than I have ivory).
>Some board sets were even made I believe.
True enough, but no customer ever saw them to my knowledge. LMI was in
a bad place financially, with Symbolics on the high end and TI doing
their best to undercut every Lisp Machine vendor on price. I'm still
not quite clear on exactly how they met their demise, with a Canadian
firm (Gigamos) involved at the end, but LMI definitely lacked the
resources to compete with the rest.
I don't wish to belittle the contributions made by the other Lisp
Machine vendors -- TI did make an honest effort on the software end, and
of course at LMI, Richard Stallman put in a tremendous amount of work
reverse-engineering much of what Symbolics did -- but I don't believe
anyone else had made as much progress during the '80s as did Symbolics.
-- Chuck
--
Chuck Fry -- Jack of all trades, master of none
······@chucko.com (text only please) ········@home.com (MIME enabled)
Lisp bigot, mountain biker, car nut, sometime guitarist and photographer
The addresses above are real. All spammers will be reported to their ISPs.
From: Joe Marshall
Subject: Re: Popularity of Symbolics Lisp machines [was: Re: FAQ: How to get a Lisp Machine. (was ...)]
Date:
Message-ID: <P1LG3.5286$FW6.6540@dfw-read.news.verio.net>
Chuck Fry <······@best.com> wrote in message
···················@nntp1.ba.best.com...
> In article <···············@lostwithiel.tfeb.org>,
> Tim Bradshaw <···@tfeb.org> wrote:
> >
> >LMI also worked on new architectures, including a final one (the k
> >machine?) which looked a whole lot more advanced than Ivory, really an
> >almost reasonable RISC one-instruction-per-clock load-store design I
> >think, (though I've seen better descriptions of it than I have ivory).
> >Some board sets were even made I believe.
3 board sets were made by LMI, approx 15 were made by Gigamos.
None of them were sold to customers.
> True enough, but no customer ever saw them to my knowledge. LMI was in
> a bad place financially, with Symbolics on the high end and TI doing
> their best to undercut every Lisp Machine vendor on price.
The initial LMI Lambdas were plagued with hardware problems. While the
hardware
and software reliability was tremendously improved by 1985, it was too
little
too late.
> I'm still
> not quite clear on exactly how they met their demise, with a Canadian
> firm (Gigamos) involved at the end, but LMI definitely lacked the
> resources to compete with the rest.
Guy MontPetit, the chairman of Gigamos, was apparently involved in
bribing Canadian senator Michel Cogger. In addition, he was sued by a
Japanese investor (Takahashi Tsuru, I believe) who claimed that MontPetit
defrauded him of about $40M. (I got this from a Canadian newspaper
that broke the story circa 1987). The Canadian courts siezed the assets of
Gigamos and asked the US courts to do the same.
Gigamos (US) went into receivership and the assets were managed by
Coopers and Lybrand who were to liquidate them. The remaining physical
assets (boards, artwork, components, etc.) of LMI ended up at Eli Hefron
and Sons (an electronics salvage company).
~jrm
From: Howard R. Stearns
Subject: Great software team? (Re: Popularity of Symbolics Lisp machines [was: Re: FAQ: How to get a Lisp Machine. (was ...)])
Date:
Message-ID: <37EBBF08.5EC9099D@elwood.com>
Chuck Fry wrote:
> ...
> As I said, I'm biased; I worked at Symbolics during the glory days of
> the mid-'80s. It was a wonderful experience, and I am glad I was there.
> I've worked with some extremely talented people over the years, but I
> don't think I've ever worked, or will ever work again, with a software
> team as talented and capable as SCRC had in the mid-'80s.
If anyone can say with a straight face that they think they know of such
a team today, in any industry, doing any thing, then please contact me
privately.
Seriously.
From: Christopher J. Vogt
Subject: Re: Popularity of Symbolics Lisp machines [was: Re: FAQ: How to get a Lisp Machine. (was ...)]
Date:
Message-ID: <37EBAAD2.60CD8316@computer.org>
Chuck Fry wrote:
>
> [ ...]
>
> LMI and TI were selling repackaged MIT CADRs; the microExplorer was a
> CADR-on-a-chip for all intents and purposes. Symbolics's first product,
> the LM-2, was also a repackaged CADR, but they designed two new
> architectures, the 36-bit 3600 and the 40-bit Ivory, specifically
> intended as Lisp machines. The 3600 series (L-machine, G-machine) was
> CISC; the Ivory was a RISC Lisp machine.
>
Just a small historical footnote: Symbolics also had an S-machine which was
a load-store single-cycle/instruction architecture. We were 2 days away from
tape-out on the chips in May of 1987 when the s-machine devo group (otherwise
know as Symbolics Western Development Center SWDC) was RIF'd.
> [ ... ]
From: Paolo Amoroso
Subject: Re: Popularity of Symbolics Lisp machines [was: Re: FAQ: How to get a Lisp Machine. (was ...)]
Date:
Message-ID: <37ed0029.4039761@news.mclink.it>
Thanks to all those who provided information and comments in this thread.
In the golden age of Lisp machines I was a kid, technology-wise.
Paolo
--
EncyCMUCLopedia * Extensive collection of CMU Common Lisp documentation
http://cvs2.cons.org:8000/cmucl/doc/EncyCMUCLopedia/
From: Rainer Joswig
Subject: Re: Popularity of Symbolics Lisp machines [was: Re: FAQ: How to get a Lisp Machine. (was ...)]
Date:
Message-ID: <joswig-2309991439160001@pbg3.lavielle.com>
In article <················@news.mclink.it>, ·······@mclink.it (Paolo Amoroso) wrote:
> Thanks to all those who provided information and comments in this thread.
> In the golden age of Lisp machines I was a kid, technology-wise.
When I heard about these things (Lisp machines, Smalltalk 80, GUIs,
Prolog, ...) in, say 1985, I had this feeling that I was born
to late and that the relevant technologies now (then) existed.
Not much really interesting seemed possible to be invented after
that.
Now, if you look at the sorry state of the computer technology
(especially when it comes to software) today, "the revolution
has not happened, yet".
From: Martin Rodgers
Subject: Re: Popularity of Symbolics Lisp machines [was: Re: FAQ: How to get a Lisp Machine. (was ...)]
Date:
Message-ID: <MPG.1254961bd92675a798a030@news.demon.co.uk>
In article <·······················@pbg3.lavielle.com>,
······@lavielle.com says...
> When I heard about these things (Lisp machines, Smalltalk 80, GUIs,
> Prolog, ...) in, say 1985, I had this feeling that I was born
> to late and that the relevant technologies now (then) existed.
I had a similar experience. I read about Smalltalk 80 in the Aug 1981
Byte issue, then read the Forth and Lisp issues from earlier years. I
was in my mid teens. I wrote my first Forth in my late teens.
Then I took another look at Lisp.
> Not much really interesting seemed possible to be invented after
> that.
A few footnotes, perhaps. ;)
> Now, if you look at the sorry state of the computer technology
> (especially when it comes to software) today, "the revolution
> has not happened, yet".
Do people using other languages give you blank looks when you mention
Lisp? I wonder if Douglas Engelbart got that "Duh" expression from
people when he showed them the kind of computing that we now have, 30
years later. When he said, "The rate at which a person can mature is
directly proportional to the embarrassment he can tolerate. I have
tolerated a lot", he could have been talking to us.
--
Please note: my email address is munged; You can never browse enough
"There are no limits." -- tagline for Hellraiser
In article <··············@2xtreme.net>,
Christopher R. Barry <······@2xtreme.net> wrote:
>To do this from the Unix command-line I could say:
>
> wget ftp://ftp.ai.mit.edu/pub/users/jcma/cl-http/lispm.reel-1
>
>To do this from the Lisp Machine (as best as I know how), it would be
>something like [from memory; after a few brown-outs here I turned off
>the machine and decided not to power it on again for fear of damage to
>it until I get a UPS to protect it and I haven't had time to use it
>lately anyways because of work and other crap (though I sure want to
>use it again and plan to soon)]:
>
> Edit Namespace Object
> Create Object (blah) HOST (blah) FTP.AI.MIT.EDU
You don't need to do that if you're using DNS.
--
Barry Margolin, ······@bbnplanet.com
GTE Internetworking, Powered by BBN, Burlington, MA
*** DON'T SEND TECHNICAL QUESTIONS DIRECTLY TO ME, post them to newsgroups.
Please DON'T copy followups to me -- I'll assume it wasn't posted to the group.
* Christopher R Barry wrote:
> Name one connectivity feature any Lisp system provides -- even the
> Lisp Machine -- that is not provided by Perl.
A generic filesystem that hides multiple underlying filesystems in a
completely transparent way. (And is extensible of course).
--tim
In article <···············@lostwithiel.tfeb.org>, Tim Bradshaw <···@tfeb.org> wrote:
> * Christopher R Barry wrote:
> > Name one connectivity feature any Lisp system provides -- even the
> > Lisp Machine -- that is not provided by Perl.
>
> A generic filesystem that hides multiple underlying filesystems in a
> completely transparent way. (And is extensible of course).
It is cruel to compare the Lisp machine with a bunch
of PERL hacks. Especially when you know that the Lisp machine
pioneered things like generic networking (which nowbody has - even
nowadays), distributed objects,
has been used to control all kinds of devices (from capture
boards to massively parrallel computers like the connection machine),
had stuff like remote 3d graphics rendering a decade ago,
ann all this was actually implemented in Lisp in a sane
way (not glued together).
If people think they cannot use Lisp to connect things or even
that Lisp is bad at connecting things, then they are just
losers or liars - it's a kind of brutal revisionism
that always amazes me. It's the ignorance of people like
Christopher Barry who have not a single clue what is going
on. I mean Lisp systems are being used in commercial
environments where they are connected to mainframes
(like at Swissair) or they are handling the whole
luggage transport at airports or they controlling large
painting systems for cars (at Daimler Benz) or ...
If you give up early, you may better use PERL, be happy
and spare us with your negativism...
I'd prefer people who can look forward, can design and implement code
if necessary - instead of all this whining.
From: Christopher R. Barry
Subject: Re: Lisp and Perl. (was Re: Book suggestions)
Date:
Message-ID: <871zbw7l6j.fsf@2xtreme.net>
······@lavielle.com (Rainer Joswig) writes:
> In article <···············@lostwithiel.tfeb.org>, Tim Bradshaw <···@tfeb.org> wrote:
>
> > * Christopher R Barry wrote:
> > > Name one connectivity feature any Lisp system provides -- even the
> > > Lisp Machine -- that is not provided by Perl.
> >
> > A generic filesystem that hides multiple underlying filesystems in a
> > completely transparent way. (And is extensible of course).
>
> It is cruel to compare the Lisp machine with a bunch
> of PERL hacks. Especially when you know that the Lisp machine
> pioneered things like generic networking (which nowbody has - even
> nowadays),
Please define the term "generic networking".
> distributed objects,
No one in this thread ever disputed that the Lisp Machine pioneered a
great many things, myself included. And that said, I don't think the
Lisp Machine has any support for modern distributed objects, though
other commercial Lisp systems have support through Corba (at a nice
add-on price).
> has been used to control all kinds of devices (from capture boards
> to massively parrallel computers like the connection machine), had
> stuff like remote 3d graphics rendering a decade ago,
C, C++, and Perl have all been used to do this.
> If people think they cannot use Lisp to connect things or even
> that Lisp is bad at connecting things, then they are just
> losers or liars
Rainer, how about backing up your talk with some facts? Lisp systems
like Allegro, Harlequin, and MCL currently do not have nearly the
connectivity that Perl has, and what connectivity they provide usually
comes as expensive layered products. Perl is a fast-to-use and cheap
choice.
> - it's a kind of brutal revisionism that always amazes me. It's the
> ignorance of people like Christopher Barry who have not a single
> clue what is going on. I mean Lisp systems are being used in
> commercial environments where they are connected to mainframes (like
> at Swissair) or they are handling the whole luggage transport at
> airports or they controlling large painting systems for cars (at
> Daimler Benz) or ...
And Perl and C++ and many other things are probably vital components
to those systems as well. (Not that any C++ components wouldn't be
better off written in Lisp, but my point all along that started this
thread is that rewriting the Perl components in Lisp may be more than
you bargained for....)
> If you give up early, you may better use PERL, be happy
> and spare us with your negativism...
> I'd prefer people who can look forward, can design and implement code
> if necessary - instead of all this whining.
Would you like it if those of us Lispers that are smart enough to use
Perl in the situations that it is in fact the better choice were to
post some of our scripts and explain what they do and ask how to even
_begin_ to make Lisp do some of the things these scripts do in a
reasonable ammount of code?
You used strong language and were very negative when replying to my
original post, and you not surprisingly haven't discussed any
technical aspects of real-world connectivity. I haven't posted to
comp.lang.lisp in a while because I've been too busy with
DHTML/CGI/Perl crap and I just don't have the time for this crap right
now like I used to. Sorry.
Christopher
In article <··············@2xtreme.net>, ······@2xtreme.net (Christopher R. Barry) wrote:
> Please define the term "generic networking".
Read the Lispm manual yourself.
> > distributed objects,
>
> No one in this thread ever disputed that the Lisp Machine pioneered a
> great many things, myself included. And that said, I don't think the
> Lisp Machine has any support for modern distributed objects,
Read the Lispm manual.
> > has been used to control all kinds of devices (from capture boards
> > to massively parrallel computers like the connection machine), had
> > stuff like remote 3d graphics rendering a decade ago,
>
> C, C++, and Perl have all been used to do this.
And Lisp, too.
> Rainer, how about backing up your talk with some facts? Lisp systems
> like Allegro, Harlequin, and MCL currently do not have nearly the
> connectivity that Perl has,
So you have compared MCL and PERL on the Mac? You are a genius.
> comes as expensive layered products. Perl is a fast-to-use and cheap
> choice.
Use PERL and go away.
> Would you like it if those of us Lispers
You are a Lisper? I haven't seen any indication of that.
> You used strong language and were very negative when replying to my
> original post,
You were making another bold claim, without knowing that much
about Lisp and its use. In your case again, why not just ask
what other people think, instead of postulating something?
I haven't posted to
> comp.lang.lisp in a while because I've been too busy with
> DHTML/CGI/Perl crap
Just what I say. That's the difference.
I'm busy with Lisp. We are writing our CGIs in Scheme
(integrating thinks like Shockwave, JavaScript, mSQL, ..).
But then we probably should use PERL because it
has better "connectivity" (hey, PERL 5.6 now gets Unicode support,
but still no usable compiler in sight.... Bitter.).
From: Christopher R. Barry
Subject: Re: Lisp and Perl. (was Re: Book suggestions)
Date:
Message-ID: <87vh98626f.fsf@2xtreme.net>
······@lavielle.com (Rainer Joswig) writes:
> In article <··············@2xtreme.net>, ······@2xtreme.net (Christopher R. Barry) wrote:
>
> > Please define the term "generic networking".
>
> Read the Lispm manual yourself.
>
> > > distributed objects,
> >
> > No one in this thread ever disputed that the Lisp Machine pioneered a
> > great many things, myself included. And that said, I don't think the
> > Lisp Machine has any support for modern distributed objects,
>
> Read the Lispm manual.
Okay, I have about 50 lbs (maybe more) of printed Symbolics
documentation. Even more in DocEx form on the machine itself. Could
you be a bit more specific than "read the Lispm manual", Rainer?
(Which I doubt you or anyone else has read the entirety of.)
> I haven't posted to
> > comp.lang.lisp in a while because I've been too busy with
> > DHTML/CGI/Perl crap
>
> Just what I say. That's the difference.
> I'm busy with Lisp. We are writing our CGIs in Scheme
> (integrating thinks like Shockwave, JavaScript, mSQL, ..).
Scheme isn't Lisp. And how are the Scheme CGIs being invoked? Are you
having Apache (or whatever) start a new Scheme process for each CGI
invocation? Do you have a Scheme system embedded in the server? (If
so, which server?) And since you specifically mention Shockwave and
JavaScript, how are you using Scheme with these? Are you generating
Shockwave content on the fly? If yes, then details please. (I'm
actually curious.) And if you are generating JavaScript on the fly (no
special feat), does your Scheme system give you special support for
JavaScript stuff? (These are all real, not rhetorical, questions.)
If you're just generating HTML that embeds JavaScript and links to
Shockwave content, and invoking a new process for each request, then
I'm naturally not too impressed.
> But then we probably should use PERL because it
> has better "connectivity" (hey, PERL 5.6 now gets Unicode support,
> but still no usable compiler in sight.... Bitter.).
Perl compiles to bytecode on the fly and there is also a Perl->C
compiler that is supposedly "usable".
Christopher
In article <··············@2xtreme.net>, ······@2xtreme.net (Christopher R. Barry) wrote:
> > Read the Lispm manual.
>
> Okay, I have about 50 lbs (maybe more) of printed Symbolics
> documentation. Even more in DocEx form on the machine itself. Could
> you be a bit more specific than "read the Lispm manual", Rainer?
Hmm, I thought it was pretty obvious that you'd need
to read books like "Networks", "Symbolics IP/TCP Software Package",
"Site operations" or the Statice docs.
> (Which I doubt you or anyone else has read the entirety of.)
Have you read some of them at all?
> If you're just generating HTML that embeds JavaScript and links to
> Shockwave content, and invoking a new process for each request, then
> I'm naturally not too impressed.
Yes, I was expecting this sentence. Stay in the darkness.
From: Christopher R. Barry
Subject: Re: Lisp and Perl. (was Re: Book suggestions)
Date:
Message-ID: <87so4c59qi.fsf@2xtreme.net>
······@lavielle.com (Rainer Joswig) writes:
> > If you're just generating HTML that embeds JavaScript and links to
> > Shockwave content, and invoking a new process for each request, then
> > I'm naturally not too impressed.
>
> Yes, I was expecting this sentence. Stay in the darkness.
Right. ;-)
Christopher
In article <·······················@194.163.195.67>,
······@lavielle.com says...
> Just what I say. That's the difference.
> I'm busy with Lisp. We are writing our CGIs in Scheme
> (integrating thinks like Shockwave, JavaScript, mSQL, ..).
I also write CGI code in Scheme - but in my own time. We're not
competing with Perl, but with tools like Cold Fusion. [sigh]
Anyway, I'm forwarding your article to my boss. Thanks.
--
Remove insect from address | You can never browse enough
will write code that writes code that writes code for food
In article <··························@news.demon.co.uk>, ···@wildcard.butterfly.demon.co.uk (Martin Rodgers) wrote:
> In article <·······················@194.163.195.67>,
> ······@lavielle.com says...
>
> > Just what I say. That's the difference.
> > I'm busy with Lisp. We are writing our CGIs in Scheme
> > (integrating thinks like Shockwave, JavaScript, mSQL, ..).
>
> I also write CGI code in Scheme - but in my own time. We're not
> competing with Perl, but with tools like Cold Fusion. [sigh]
>
> Anyway, I'm forwarding your article to my boss. Thanks.
You can tell him that the 15k lines of Scheme + 1k lines of CL
have earned > DM 100k (until now).
In article <·······················@194.163.195.67>,
······@lavielle.com says...
> You can tell him that the 15k lines of Scheme + 1k lines of CL
> have earned > DM 100k (until now).
I certainly will, thanks.
--
Remove insect from address | You can never browse enough
will write code that writes code that writes code for food
In article <··············@2xtreme.net>,
Christopher R. Barry <······@2xtreme.net> wrote:
>······@lavielle.com (Rainer Joswig) writes:
>
>> In article <···············@lostwithiel.tfeb.org>, Tim Bradshaw
><···@tfeb.org> wrote:
>>
>> > * Christopher R Barry wrote:
>> > > Name one connectivity feature any Lisp system provides -- even the
>> > > Lisp Machine -- that is not provided by Perl.
>> >
>> > A generic filesystem that hides multiple underlying filesystems in a
>> > completely transparent way. (And is extensible of course).
>>
>> It is cruel to compare the Lisp machine with a bunch
>> of PERL hacks. Especially when you know that the Lisp machine
>> pioneered things like generic networking (which nowbody has - even
>> nowadays),
>
>Please define the term "generic networking".
A protocol-independent networking API. The same terminal emulator is used
for TELNET over TCP/IP, RLOGIN over TCP/IP, TELNET over CHAOSnet,
host connections over async (direct connect or dialup), and whatever the
analogous DECnet remost login protocol is. You use the same API to access
files via FTP, NFILE, or NFS.
--
Barry Margolin, ······@bbnplanet.com
GTE Internetworking, Powered by BBN, Burlington, MA
*** DON'T SEND TECHNICAL QUESTIONS DIRECTLY TO ME, post them to newsgroups.
Please DON'T copy followups to me -- I'll assume it wasn't posted to the group.
In article <·······················@194.163.195.67>,
······@lavielle.com says...
> If you give up early, you may better use PERL, be happy
> and spare us with your negativism...
It's not giving up on Lisp to admit that there are domains and
environments that are hostile to Lisp. I call that pragmatism.
When time is short, I'll use whatever tools work fastest. Sometimes
that'll be Perl, so I sometimes use Perl. This isn't a "because I want
to" situation, it's "because I can".
In theory we should all use Lisp all the time. Alas, real life is more
demanding. This may indeed be due to negativism, but it ain't ours.
It's worth stressing this point: some of us use Lisp _and_ Perl.
--
Remove insect from address | You can never browse enough
will write code that writes code that writes code for food
···@wildcard.butterfly.demon.co.uk (Martin Rodgers) writes:
> In theory we should all use Lisp all the time. Alas, real life is more
> demanding. This may indeed be due to negativism, but it ain't ours.
>
> It's worth stressing this point: some of us use Lisp _and_ Perl.
You took the words right out of my mind. As I read through this
thread, I was thinking the same. It's funny how quickly my company
turned from a Lisp house to a Perl house for web-based programming.
When I use Perl, I say to myself, over and over again, that "if only
Lisp had the libraries". Perl really does come equipped with some
useful tools.
This guy, CB, is one of the most pestilent people I've encountered on
the newsgroups, and definitely on this group. But that doesn't mean
that he's always wrong.
I hate perl, but I must admit that part of that hatred is envy that it
gets the job done. I use it because it is as powerful as hell when it
comes to going from zero to solution, and now that I've become about
half as comfortable with it as Lisp, I'm less scared of it. Lisp
programmers are very often scared of Perl, because it's kinda hard to
get a grasp of after you've been hacking at Lisp for a decade or two.
But it's not so bad, and (as CB said) it does compile in its own way,
and it's actually quite quick.
Anyway, I wanted to address some of the various things that I read in
this thread:
1. Name some connectivity features that Lisp has and Perl doesn't.
There's no doubt that right now, Perl is the way to go for simple
problems. If you need to write a CGI script to do something simple,
then really, Perl is better than Lisp. There are lots of reasons for
this. And if anything is thinking CL-HTTP, then I say trash that
noise. CL-HTTP is ugly to me, and so is its license. It's not open
or free. I sent J. Mallery an email asking him about it a long time
ago, and got nothing back. At least Perl is free, and that's not
going to change.
Let's take SQL/ODBC connectivity. I I use LispWorks on MS Windows.
For not very much money, I have everything bundled. The SQL/ODBC
package there blows away anything that Perl has in DBI or elsewhere by
orders of magnitude. Now, it's very possible that with Perl, someone
can put together something like what LispWorks has, using Ties, DBI,
objects, etc, but it'll be a painful implementation, I doubt it will
make database programming nearly as expressing and simple as LispWorks
has. So there you go.
And, finally, what can most commercial Lisps do that Perl can't? How
about multiprocessing? ACL and LispWorks both have really nice MP
packages. Perl's is not quite there (yet). But, if you see what's in
the works, it doesn't look too bad:
use Thread;
$t = new Thread \&start_sub, @start_args;
$result = $t->join;
$t->detach;
And, when the threads finally do work in Perl, you can probably count
on them working on every OS that can support threads. Furthermore,
you get the additional packages, like Thread::Semaphore, Thread::Pool,
Thread::Queue. It's easy to implement this stuff in Lisp, but it's
nice when it's all just right there, like in Perl (and with Perl, you
can always see the source; can't say the same about what comes out of
these Commercial Lisp houses. I think in some cases, you can pay more
money to see the source, though I'm not too sure).
Unfortunately, though, you've gotta ask yourself if Perl was really
cut out for advanced programming features. I personally don't think
it is, and eventually, Perl will have to die. Lisp can easily scale
where I can't see Perl scaling quite as easily. For example, Perl
uses the global $_ magic variable all over the place. What's gonna
happen to this variable in the multi-threaded Perl? I'm sure they'll
address this, but still, it just goes to show that Perl wasn't really
supposed to be doing this stuff.
2. Libraries
I really came to love Lisp when I realized that all the things I
enjoyed doing were more fun and elegant to code up in Lisp than in any
other language. It wasn't even that Lisp was just better than the
others; it really made some tasks fun. But for extracting data from
text files, Lisp pales in comparison. This is why I sometimes wish:
1) Harlequin would have spent the Dylan money instead on simply
putting together massive libraries for their Lisp (like SMTP, FTP,
HTTP, Regexp, CGI, ...).
2) That Franz would get their CBIND utility to the point where it was
so robust that they could swallow efficient packages written in C,
like Spencer's regexp package, and make it easy to use those
libraries by writing some wrapper code around it. (instead, ACL
has a not-so-friendly regexp package, lacking functionality and
with a terrible syntax. Still, they're the only Lisp vendor that I
know of that has a functional package with good performance.)
Also worth mentioning is that Perl modules tend to have a pretty
simple and uniform structure to them. They're almost all
object-oriented, and if you read the synopses, they always look
similar, and quite simple:
example:
use Net::SMTP;
$smtp = Net::SMTP->new('smtp.mycorp.com', Timeout => 10);
$smtp->mail(·····@bu.edu');
$smtp->to(···@friend.net');
$smtp->data();
$smtp->datasend("yo. what's up? --dave");
$smtp->dataend();
$smtp->quit();
As you look through what Perl has, you see some lisp things like
anonymous functions, keyword args (see `Timeout' above), and such.
enough. Here's my point:
The world embraced Perl when it came out. It had a great author,
great maintainers, an amazing community, a generous license, and a
grand goal. People shoot it down all the time, but the world will
keep using it, and it'll keep doing what it does. If the world was
smarter, they would have chosen Lisp, and HTML would be LML, and then
XML would not even need to exist, and the list goes on and on. But
Lisp simply has too many obstacles in the way. Very few of them are
technical, though.
dave
In article <··············@lost-in-space.ne.mediaone.net>,
·····@bu.edu says...
> If the world was
> smarter, they would have chosen Lisp, and HTML would be LML, and then
> XML would not even need to exist, and the list goes on and on.
This reminds me of Online DSSSL. How could they do it? We've seen
countless examples of people who don't get it. Well, this is just one
more.
> But
> Lisp simply has too many obstacles in the way. Very few of them are
> technical, though.
Some obstacles are certainly political. Some might be the result of a
poor education (anyone here remember the "Peaceman" thread?").
Another thread, not too long ago, mentioned "marketing by stealth".
An image springs to mind, curtesy of the late Alan Clark: Lisp
programmers playing cricket while Perl programmers throw hand
granades. Alas, the Perl mob would probably take it as a compliment.
Perl is just the tip of the "worse is better" iceberg. While two-way
conflicts are an attractive problem to solve, reality is not so
conveniently simple. Gabriel also looked below the waterline.
--
Remove insect from address | You can never browse enough
will write code that writes code that writes code for food
From: Reini Urban
Subject: Re: Lisp and Perl. (was Re: Book suggestions)
Date:
Message-ID: <37e6052c.86643977@judy>
Dave Bakhash wrote:
>2) That Franz would get their CBIND utility to the point where it was
> so robust that they could swallow efficient packages written in C,
> like Spencer's regexp package, and make it easy to use those
> libraries by writing some wrapper code around it. (instead, ACL
> has a not-so-friendly regexp package, lacking functionality and
> with a terrible syntax. Still, they're the only Lisp vendor that I
> know of that has a functional package with good performance.)
But David, you forgot that spencer's original regexp library is almost
impossible to wrap automatically without manual corrections to the
interface (-> CLISP, python, perl) and it is almost impossible to
understand and maintain.
With the GNU regex it was cleaned up a bit, but it is still a mess.
Same with trying to extract and wrap the perl regex extensions to
henry's code, (search/replace instead of the simple search) by parsing
the sources and producing wrappers automatically.
CBIND is in my opinion one of the best implementations, because it went
the stupid way (as CLISP as well, btw): just invoke the GNU C
preprozessor, instead of trying to parse the C mess in LISP.
It swallows even the spencer regex, but the mess it produces is
spencer's fault, not CBIND's.
see CLISP/modules/regexp/regexp.lsp and regexpi.c (which is actually
quite simple, but hand-optimized)
BTW: I agree with the rest of this perl<->lisp posts by Martin and Dave.
--
Reini
another typical example of recent overreaction: NSAKEY
http://hyperarchive.lcs.mit.edu/telecom-archives/archives/back.issues/recent.single.issues/V19_%23379
······@xarch.tu-graz.ac.at (Reini Urban) writes:
> Dave Bakhash wrote:
> >2) That Franz would get their CBIND utility to the point where it was
> > so robust that they could swallow efficient packages written in C,
> > like Spencer's regexp package, and make it easy to use those
> > libraries by writing some wrapper code around it. (instead, ACL
> > has a not-so-friendly regexp package, lacking functionality and
> > with a terrible syntax. Still, they're the only Lisp vendor that I
> > know of that has a functional package with good performance.)
>
> But David, you forgot that spencer's original regexp library is almost
> impossible to wrap automatically without manual corrections to the
> [...]
> It swallows even the spencer regex, but the mess it produces is
> spencer's fault, not CBIND's.
I don't know what to say. I tried CBIND on spencer's code, and it
just couldn't deal at all. That was about 6 months ago, so it's
possible that CBIND has been improved. But my opinion is that if the
C code compiles, then CBIND should not fail.
btw, I didn't know that CLISP had a regexp library. I'll check it
out, since I recently built it. thanks.
dave
In article <··············@lost-in-space.ne.mediaone.net>, Dave Bakhash
<·····@bu.edu> wrote:
> ...And if anything is thinking CL-HTTP, then I say trash that
> noise. CL-HTTP is ugly to me, and so is its license. It's not open
> or free. I sent J. Mallery an email asking him about it a long time
> ago, and got nothing back. At least Perl is free, and that's not
> going to change.
I'm new to this, so I'd be really interested in what you find ugly
about CL-HTTP (apart from the license). I thought it looked pretty
interesting at first sight.
Thanks,
Alex
From: David Bakhash
Subject: Re: Lisp and Perl. (was Re: Book suggestions)
Date:
Message-ID: <cxjiu4aymqk.fsf@engc.bu.edu>
interesting is not the same as straightforward. CL-HTTP is not
simple. It's hard to use. I discovered that when I considered using
CL-HTTP for CGI, i.e. to dynamically generate HTML.
Last I remember, CL-HTTP did not have a web-based administration
mechanism, like Xitami has, and IIS has.
It's open-source, so just dive in. It's a judgment call -- some
people will like the style of CL-HTTP and some won't. certainly don't
take any one person's word for it.
I'm very happy that it exists, though, and I think it's overall design
is not bad.
From: Gareth McCaughan
Subject: Re: Lisp and Perl. (was Re: Book suggestions)
Date:
Message-ID: <864sgsbc6q.fsf@g.local>
Christopher R. Barry wrote:
>>> In
>>> other words, if you use Perl as a _scripting_language_ for writing
>>> _scripts_, and understand when a script is called for and when a
>>> program is, you can use Perl productively. It provides far more
>>> connectivity than all the add-on features of all commercial Lisp
>>> systems combined,
[Rainer Joswig:]
>> Bullshit.
[Christopher Barry again:]
> Name one connectivity feature any Lisp system provides -- even the
> Lisp Machine -- that is not provided by Perl. Oh, and don't try to be
> a lawyer and say something like "the Lisp Machine provides Statice
> connectivity", since for example SQL and DBM connectivity and
> Oracle/Informix/etc is all that really matters. (And commercial Lisp
> systems do provide good SQL connectivity as well as many other
> important kinds -- somtimes for a lot of extra money -- but I'm
> interested in you naming a connectivity feature that a Lisp system
> provides that is not provided by Perl, since you are calling me a
> liar.)
You didn't say "Perl provides all the connectivity features you
can find in Lisp systems"; you said "far more". Rainer's reply
was certainly intemperate, but if it turns out (say) that "all
the add-on features of all commercial Lisp systems combined"
provide exactly as much connectivity as Perl does then (1) your
original statement will be exploded, but (2) Rainer will be unable
to meet the challenge you have just posed. I don't think that's
fair.
--
Gareth McCaughan ················@pobox.com
sig under construction
In article <··············@g.local>, Gareth McCaughan <················@pobox.com> wrote:
> original statement will be exploded, but (2) Rainer will be unable
> to meet the challenge you have just posed. I don't think that's
> fair.
You got it wrong. He was claiming that PERL provides more
connectivity (blablabla). He has to prove that.
Then he talked about a system for which development
pretty much halted 8 (?) years ago - which at ***that*** time had
- generic object-oriented networking (pretty much unmatched today)
- NFS server and client
- X server (written in C) and client
- terminal client and server
- decnet support
- FTP server and client
- DNS server and client
- LPR support
- its own network protocol: CHAOS
- its own file sharing protocol: NFILE
- SMTP reader and server
- NNTP reader
- a client-server object-oriented database
- FINGER, TALK, ...
- SLIP
- IP routing
- netbooting
- site management
- X25 support
- RPC
- support for various VME bus cards
- support for SCSI disks and tapes
- an advanced file system
- an advanced window system
...
all written in Lisp.
Why are guys like he wasting their and our time at comp.lang.lisp?
Why isn't he writing his crap on comp.lang.perl?
From: Gareth McCaughan
Subject: Re: Lisp and Perl. (was Re: Book suggestions)
Date:
Message-ID: <86hfks9q60.fsf@g.local>
Rainer Joswig wrote:
[I said:]
>> original statement will be exploded, but (2) Rainer will be unable
>> to meet the challenge you have just posed. I don't think that's
>> fair.
>
> You got it wrong. He was claiming that PERL provides more
> connectivity (blablabla). He has to prove that.
That is *exactly* what I said. I think you misunderstood my article.
Note that the #2 you quoted there was part of a *hypothetical*
situation. In other words, I'm not saying that you will in fact
be unable to meet the challenge; I'm saying that your original
criticism could be right even if you were unable to meet the
challenge.
--
Gareth McCaughan ················@pobox.com
sig under construction
From: Christopher R. Barry
Subject: Re: Lisp and Perl. (was Re: Book suggestions)
Date:
Message-ID: <87puzg57uj.fsf@2xtreme.net>
······@lavielle.com (Rainer Joswig) writes:
> In article <··············@g.local>, Gareth McCaughan <················@pobox.com> wrote:
>
> > original statement will be exploded, but (2) Rainer will be unable
> > to meet the challenge you have just posed. I don't think that's
> > fair.
>
> You got it wrong. He was claiming that PERL provides more
> connectivity (blablabla). He has to prove that.
>
>
> Then he talked about a system for which development
> pretty much halted 8 (?) years ago - which at ***that*** time had
[...]
> - NNTP reader
[...]
Where is it and how do I use it? After much frustrating searching
through the Symbolics docs for any kind of NNTP support, I gave up and
asked at the SLUG-list and got no reply. I don't believe the Lisp
Machine has any NNTP connectivity. I was going to work around this by
sending news over a different transport from the Ivory to the Linux
box and then have the Linux box forward and retrieve using its NNTP
support but more pressing matters took my time instead.
Oh, and I've read pretty much _all_ of the Symbolics mail and
networking docs. I read the X Window System guide lord knows how many
times trying to get that damned Symbolics X Server working....
Christopher
In article <··············@2xtreme.net>, ······@2xtreme.net (Christopher R. Barry) wrote:
> I don't believe the Lisp
> Machine has any NNTP connectivity.
That's exactly your problem.
From "Usenet Software: History and Sources":
There is also an NNTP-based netnews reader for Symbolics Lisp Machines
(under Genera 7) available for anonymous FTP from ucbvax.berkeley.edu
[128.32.133.1] in pub/nntp-clients/lispm written by Ian Connolly
<········@coins.cs.umass.edu> and maintained by Richard Welty
<·····@lewis.crd.ge.com>. In addition, another NNTP-based news
browser is available running under Genera 7 and Genera 8. It provides
mouse driven hierarchic browsing of newsgroups and articles, with
support for X11 servers on remote machines. It is available for
anonymous FTP on flash.bellcore.com [128.96.32.20] in the directory
pub/lispm/news-reader/. It is written and maintained by Peter
Clitherow <··@bellcore.com>
I don't think the code is still there. The latter was called
"Hermes" and was written with CLIM 1.1 (I think) for the Lisp machine
(using Lisp machine specific networking code).
If anybody wants to port it to CLIM 2 (which is a bit of work)
and contribute the result - I could look in my archive for a copy.
From: Christopher R. Barry
Subject: Re: Lisp and Perl. (was Re: Book suggestions)
Date:
Message-ID: <87n1uj6h24.fsf@2xtreme.net>
······@lavielle.com (Rainer Joswig) writes:
> In article <··············@2xtreme.net>, ······@2xtreme.net (Christopher R. Barry) wrote:
>
> > I don't believe the Lisp
> > Machine has any NNTP connectivity.
>
> That's exactly your problem.
>
> From "Usenet Software: History and Sources":
>
> There is also an NNTP-based netnews reader for Symbolics Lisp Machines
> (under Genera 7) available for anonymous FTP from ucbvax.berkeley.edu
> [128.32.133.1] in pub/nntp-clients/lispm written by Ian Connolly
> <········@coins.cs.umass.edu> and maintained by Richard Welty
> <·····@lewis.crd.ge.com>. In addition, another NNTP-based news
> browser is available running under Genera 7 and Genera 8. It provides
> mouse driven hierarchic browsing of newsgroups and articles, with
> support for X11 servers on remote machines. It is available for
> anonymous FTP on flash.bellcore.com [128.96.32.20] in the directory
> pub/lispm/news-reader/. It is written and maintained by Peter
> Clitherow <··@bellcore.com>
>
> I don't think the code is still there. The latter was called
> "Hermes" and was written with CLIM 1.1 (I think) for the Lisp machine
> (using Lisp machine specific networking code).
> If anybody wants to port it to CLIM 2 (which is a bit of work)
> and contribute the result - I could look in my archive for a copy.
Neither of those servers exist anymore, and I get the feeling that the
latter client probably wouldn't compile/run on Genera 8.3 as-is. (I've
found plenty of old Symbolics code on the net that doesn't work as-is.
It's not always too hard to get it working, and it's fun to explore
the system in the process, but it can be tedious....)
Anyways, I guess it's nice to know that at one time there was an NNTP
client solution for Lisp Machines, though I'm not convinced there's
anything that I can actually use right now without putting in a good
bit of effort.
Christopher
On 13 Sep 1999 12:33:28 GMT, ···@unpkhswm04.bscc.bls.com (Mitchell Morris)
wrote:
> have anything that I recognized as this, but are there any books that discuss
> the philosophy of LISP and/or thinking in LISP more than the syntax or
> function library?
Check the Sep. 1991 special issue on Lisp of CACM. It also includes the
paper "The Philosophy of Lisp" by Kenneth Sinclair and David Moon.
Paolo
--
EncyCMUCLopedia * Extensive collection of CMU Common Lisp documentation
http://cvs2.cons.org:8000/cmucl/doc/EncyCMUCLopedia/
For uses of Lisp, I like
Peter Norvig.
Paradigms of Artificial Intelligence Programming: Case Studies in Common Lisp
ISBN: 1-55860-191-0
Morgan Kaufmann
--
Thomas A. Russ, USC/Information Sciences Institute ···@isi.edu