From: Kenny Tilton
Subject: CLIM looks cool, but...a new CL community poll
Date:
Message-ID: <3BF5E9AF.DDDC3A42@nyc.rr.com>
...it has occurred to me that sharing my dataflow hack is my real goal
and interest, and I approached it as "make a great CL GUI" because
providing that would get newbies into my stuff more quickly. It also
occurred to me that a standard CL GUI would help make CL more
approachable, but the CL GUI game looks pretty fractured right now and
perhaps rather than try to unify that with a killer GUI layer -- a tough
sell against CG, CAPI and MCL's GUI (forget if it had a moniker) --
well, the fact is my stuff can be mixed in if you will with any
windowing system. I did it on MCL and then did it in ACL/CG. CLIM and X
look like fits as well.
So rather than go CLIM or go native (isolating os-dependent stuff in a
package folks could implement for any substrate they like), I am
thinking take one familiar (to me) platform (probably ACL/CG) and
release a framework for that (CG specifics isolated in a package). Let
CAPI or CLIM or GTk or MCL users port to their substrates if they like
(and they would do it better already knowing their chosen systems). And
then everyone gets a new Garnet-like environment, not just CLIMers.
One big factor here is that this exercise has to be easy on me since it
is a part-time thing and I am coding all day anyway. The ACL/CG stuff is
pretty much ready to go, just have to clean it up some. The dataflow
hack itself can grow in many directions, and I would rather work on that
to support user requirements.
The one problem with this plan is: how will anyone know if they like my
stuff if they do not have ACL? :) Several years ago the free version of
ACL was too crippled to run the dataflow engine. Well, OK, I gave up
after one try. :) Maybe the LispWorks freebee is less crippled, or maybe
the new ACL freebee is juicier. Also, I think I can drastically cut the
stack usage of the engine, maybe the freebees will Just Work.
Poll Question #1: suppose for the sake of argument my hack is an
Insanely Great hack producing not just a Garnet-on-steroids GUI layer
but a wholly different and more productive dataflow paradigm. /Iff/ you
would want to check such a creature out, what platform would you need to
see it on even to check it out? I actually have no idea of the
distribution of CL users across diff vendors/OSes/hardware/GUI.
I have MCL and ACL/Win32 platforms sitting here, but last I looked MCL
did not support the MOP and my current implementation uses the MOP. But
the first imp was on MCL hence MOP-free so I guess I could grind out a
MOP-free imp if pressed (it's grown a little under ACL and would have to
be ported back).
Poll Question #2: suppose I get the thing cooking under free ACL/win32.
I guess that reaches a lot of the possibly interested? And I just
checked out the free MCL, I see it does not have any heap limitation,
tho you are limited to 15min (yikes!) or four weeks unlimited... well,
this is dumb, if yer on the Mac you probably have an MCL license.
Anyway, would any of you Linux/CMUCL/CLIM types fire up a win32
partition to check it out on a free CL?
kenny
clinisys
From: Kenny Tilton
Subject: Re: CLIM looks cool, but...a new CL community poll
Date:
Message-ID: <3BF61B88.7F553F6D@nyc.rr.com>
Kenny Tilton wrote:
>
>
> Poll Question #2: suppose I get the thing cooking under free ACL/win32.
OK, good news is it does run under the trial ACL version, the bad news
is the way their license reads. I'll have to get in touch with them. Use
for evaluation of my stuff by a university to decide if they wanted to
port it to, say, CLIM looks to be disallowed.
Anyway, I can certainly share as is with ACL/win32 owners. Still mulling
the CLIM thing.
kenny
clinisys
······@hushmail.com wrote:
> On Sat, 17 Nov 2001 08:08:04 GMT, Kenny Tilton <·······@nyc.rr.com> wrote:
>
>>
>>
>>Kenny Tilton wrote:
>>>
>>>
>>> Poll Question #2: suppose I get the thing cooking under free ACL/win32.
>>
>>OK, good news is it does run under the trial ACL version, the bad news
>>is the way their license reads. I'll have to get in touch with them. Use
>>for evaluation of my stuff by a university to decide if they wanted to
>>port it to, say, CLIM looks to be disallowed.
>
> I wouldn't worry too much. If your code is useful, someone will port it.
> Look at AllegroServe: there are already LW and Corman ports of it.
and CMUCL
ciao,
Jochen
--
http://www.dataheaven.de
From: Kenny Tilton
Subject: Re: CLIM looks cool, but...a new CL community poll
Date:
Message-ID: <3BF6B5BD.D5927371@nyc.rr.com>
······@hushmail.com wrote:
>
> On Sat, 17 Nov 2001 08:08:04 GMT, Kenny Tilton <·······@nyc.rr.com> wrote:
>
> >
> >
> >Kenny Tilton wrote:
> >>
> >>
> >> Poll Question #2: suppose I get the thing cooking under free ACL/win32.
> >
> >OK, good news is it does run under the trial ACL version, the bad news
> >is the way their license reads. I'll have to get in touch with them. Use
> >for evaluation of my stuff by a university to decide if they wanted to
> >port it to, say, CLIM looks to be disallowed.
>
> I wouldn't worry too much. If your code is useful, someone will port it. Look
> at AllegroServe: there are already LW and Corman ports of it.
One issue is that while I would GPL or LGPL the GUI I have not yet
decided 100% to share the dataflow engine source, and until I do I have
to provide the FASLs for that. I can do ACL now and I would buy LW and
Corman CL if anyone cared about those, and of course an MCL version if I
re-port back to a MOP-free or MOP-lite implementation. I would also
break down and get Linux installed and do CMUCL versions for the right
partner (someone who would come over and install that stuff <g>).
kenny
clinisys
From: Dr. Edmund Weitz
Subject: Re: CLIM looks cool, but...a new CL community poll
Date:
Message-ID: <m3y9l4ping.fsf@duke.agharta.de>
Kenny Tilton <·······@nyc.rr.com> writes:
> One issue is that while I would GPL or LGPL the GUI I have not yet
> decided 100% to share the dataflow engine source, and until I do I
> have to provide the FASLs for that. I can do ACL now and I would buy
> LW and Corman CL if anyone cared about those, and of course an MCL
> version if I re-port back to a MOP-free or MOP-lite
> implementation. I would also break down and get Linux installed and
> do CMUCL versions for the right partner (someone who would come over
> and install that stuff <g>).
Hmmm, if you don't want to release the source, why not distribute
stand-alone executables for the people who are willing to test your
code? AFAIK, ACL, LW, and MCL can do that, and you wouldn't need to
ask someone to install a specific (trial) implementation. And, if
you're already thinking about buying LW: You could buy the Linux
version of LW and thus be able to provide executables to the
Linux/FreeBSD users out there. (This of course implies that you
install a Linux distro on one of your machines - or someone else does
it for you - to be able to deliver the executable.)
Edi.
From: Kenny Tilton
Subject: Re: CLIM looks cool, but...a new CL community poll
Date:
Message-ID: <3BF78986.1F565B12@nyc.rr.com>
"Dr. Edmund Weitz" wrote:
>
> Hmmm, if you don't want to release the source, why not distribute
> stand-alone executables for the people who are willing to test your
> code?
Well, I want folks to be able to play with tutorial examples and then go
nuts on the GUI once they get the hang of things, meaning they would
need the compiler and even the IDE, and that is not allowed by the
licenses I have seen for commercial CLs.
kenny
clinisys
From: Dr. Edmund Weitz
Subject: Re: CLIM looks cool, but...a new CL community poll
Date:
Message-ID: <m366884c3z.fsf@duke.agharta.de>
Kenny Tilton <·······@nyc.rr.com> writes:
> "Dr. Edmund Weitz" wrote:
>>
>> Hmmm, if you don't want to release the source, why not distribute
>> stand-alone executables for the people who are willing to test your
>> code?
>
> Well, I want folks to be able to play with tutorial examples and then go
> nuts on the GUI once they get the hang of things, meaning they would
> need the compiler and even the IDE, and that is not allowed by the
> licenses I have seen for commercial CLs.
From the LispWorks FAQ:
Q. What's the royalty for delivering my own application?
A. End-user applications delivered from the Professional Edition are
royalty free, including those using CLIM 2.0! Xanalys believes it
is the only company in the world to offer such an unrestrictive
commercial Lisp deployment model.
I haven't found anything that prevents you from delivering the
compiler or the IDE with your application. Please correct me if I've
missed something.
Edi.
From: Kenny Tilton
Subject: Re: CLIM looks cool, but...a new CL community poll
Date:
Message-ID: <3BF80D7B.D73D5EB4@nyc.rr.com>
"Dr. Edmund Weitz" wrote:
>
> From the LispWorks FAQ:
>
> Q. What's the royalty for delivering my own application?
>
> A. End-user applications delivered from the Professional Edition are
> royalty free, including those using CLIM 2.0! Xanalys believes it
> is the only company in the world to offer such an unrestrictive
> commercial Lisp deployment model.
>
> I haven't found anything that prevents you from delivering the
> compiler or the IDE with your application. Please correct me if I've
> missed something.
LW is not one of the product licenses I have examined and I will not
guess at the details of their licensing, but the above Q&A is about
distributing /my/ application, not XAnalysis's (which is pretty much
what I do if I distribute their compiler and IDE).
So what /is/ cool is that if I develop, say, a chess application with LW
I can distrib without paying XAnalysis royalties, not so with ACL. Not
sure why they say "only". MCL used to be that way, but hey last I looked
was five years ago, so whaddoIknow?
The key phrase is "applications delivered /from/ [emphasis added] the
Professional Edition". But check with XAnalysis, I am just speculating.
kenny
clinisys
Kenny Tilton wrote:
>
>
> "Dr. Edmund Weitz" wrote:
>>
>> From the LispWorks FAQ:
>>
>> Q. What's the royalty for delivering my own application?
>>
>> A. End-user applications delivered from the Professional Edition are
>> royalty free, including those using CLIM 2.0! Xanalys believes it
>> is the only company in the world to offer such an unrestrictive
>> commercial Lisp deployment model.
>>
>> I haven't found anything that prevents you from delivering the
>> compiler or the IDE with your application. Please correct me if I've
>> missed something.
>
> LW is not one of the product licenses I have examined and I will not
> guess at the details of their licensing, but the above Q&A is about
> distributing /my/ application, not XAnalysis's (which is pretty much
> what I do if I distribute their compiler and IDE).
>
> So what /is/ cool is that if I develop, say, a chess application with LW
> I can distrib without paying XAnalysis royalties, not so with ACL. Not
> sure why they say "only". MCL used to be that way, but hey last I looked
> was five years ago, so whaddoIknow?
>
> The key phrase is "applications delivered /from/ [emphasis added] the
> Professional Edition". But check with XAnalysis, I am just speculating.
Taken from the license:
3. Distribution of Runtimes. You may distribute Runtimes solely to
end-users as part of an application developed using the Software
("Application"), except that you may not distribute any part of the
Software as a general purpose Lisp development tool. Any Runtimes
distributed as part of the Application will continue to be subject to
the terms and conditions of this Agreement. You agree to license the
Application to your customers under a written license agreement
containing terms and conditions with regard to the Software and Runtimes
that are at least as restrictive as those contained herein.
So your application must not have a part that can be considered a "general
purpose Lisp development tool". So I think delivering IDE and compiler can
lead very fast to such situation.
Another issue I have read somewhere in the documentation is that the
delivered application contains no file-compiler - so adding the compiler to
the application means having COMPILE but not COMPILE-FILE (as far as I
understand it)
Please take note that I recite from my memory so it is certainly a good
idea to simply ask Xanalys Support for the exact details.
ciao,
Jochen
--
http://www.dataheaven.de
Kenny Tilton <·······@nyc.rr.com> writes:
> I have MCL and ACL/Win32 platforms sitting here, but last I looked MCL
> did not support the MOP and my current implementation uses the MOP. But
> the first imp was on MCL hence MOP-free so I guess I could grind out a
> MOP-free imp if pressed (it's grown a little under ACL and would have to
> be ported back).
MCL has supported parts of the MOP back to version 2.0. If you have a
copy of the hard-copy documentation, look at the implementation notes
in Appendix A for some specifics. IMHO, one notable missing feature
is SLOT-VALUE-USING-CLASS. Fortunately, there is some contributed
code which gives you that method at the expense of the standard
slot-access optimizations. I seem to recall parts of the
documentation as being less than up-to-date, so it is definitely worth
trying out what you have and just seeing what breaks.
From: Kenny Tilton
Subject: Re: CLIM looks cool, but...a new CL community poll
Date:
Message-ID: <3BF69B1D.6BAA73FB@nyc.rr.com>
Carl Shapiro wrote:
>
> Kenny Tilton <·······@nyc.rr.com> writes:
>
> > I have MCL and ACL/Win32 platforms sitting here, but last I looked MCL
> > did not support the MOP
>
> MCL has supported parts of the MOP back to version 2.0. If you have a
> copy of the hard-copy documentation, look at the implementation notes
> in Appendix A for some specifics. IMHO, one notable missing feature
> is SLOT-VALUE-USING-CLASS.
Right. My current implementation beats on SVUC like it was its daddy.
But I might try an SVUC-free approach, even MOP-free possibly. Either
way in a pinch I could resurrect the original MCL, MOP-free approach.
kenny
clinisys
Kenny Tilton <·······@nyc.rr.com> writes:
> Right. My current implementation beats on SVUC like it was its daddy.
Heh :-)
> But I might try an SVUC-free approach, even MOP-free possibly. Either
> way in a pinch I could resurrect the original MCL, MOP-free approach.
ftp://ftp.digitool.com/pub/mcl/contrib/slot-value-using-class.lisp
Kenny Tilton <·······@nyc.rr.com> writes:
> Poll Question #1: suppose for the sake of argument my hack is an
> Insanely Great hack producing not just a Garnet-on-steroids GUI layer
> but a wholly different and more productive dataflow paradigm. /Iff/ you
> would want to check such a creature out, what platform would you need to
> see it on even to check it out? I actually have no idea of the
> distribution of CL users across diff vendors/OSes/hardware/GUI.
Well, it depends. Your billing of it as Garnet-on-steroids makes me
curious, given that I happily use Garnet. I would definately check it
out if it ran on CMUCL or CLISP on Debian. If it came with
documentation [ ;-) ] that made me understand how much better than
Garnet it is, I'd consider installing the copy of NT I have sitting
around. But I probably want to do that even less than you want to
install Debian (I ended out with Debian because a friend promised me
that if I installed it once, I'd never have to reinstall for as long
as I have this machine -- I *hate* system administration with a passion).
Your billing of this as Garnet-on-steroids makes me wonder, why don't
you just port Garnet to ACL/Win32? Or make an actual
Garnet-on-steroids, by feeding steroids to the actual Garnet code base?
> Poll Question #2: suppose I get the thing cooking under free ACL/win32.
> I guess that reaches a lot of the possibly interested? And I just
> checked out the free MCL, I see it does not have any heap limitation,
> tho you are limited to 15min (yikes!) or four weeks unlimited... well,
> this is dumb, if yer on the Mac you probably have an MCL license.
> Anyway, would any of you Linux/CMUCL/CLIM types fire up a win32
> partition to check it out on a free CL?
Well, I'm a Linux/CMUCL/Garnet type, but I might. This is pretty much
the same as Question #1, btw :-)
--
/|_ .-----------------------.
,' .\ / | No to Imperialist war |
,--' _,' | Wage class war! |
/ / `-----------------------'
( -. |
| ) |
(`-. '--.)
`. )----'
From: Kenny Tilton
Subject: Re: CLIM looks cool, but...a new CL community poll
Date:
Message-ID: <3BF76CA5.B4508FBE@nyc.rr.com>
"Thomas F. Burdick" wrote:
>
> Kenny Tilton <·······@nyc.rr.com> writes:
>
> > Poll Question #1: suppose for the sake of argument my hack is an
> > Insanely Great hack producing not just a Garnet-on-steroids GUI layer
> > but a wholly different and more productive dataflow paradigm. /Iff/ you
> > would want to check such a creature out, what platform would you need to
> > see it on even to check it out? I actually have no idea of the
> > distribution of CL users across diff vendors/OSes/hardware/GUI.
>
> Well, it depends. Your billing of it as Garnet-on-steroids makes me
> curious, given that I happily use Garnet. I would definately check it
> out if it ran on CMUCL or CLISP on Debian.
OK. I gave up pretty quick, if it comes down to it I suppose I could
slog thru all that.
> If it came with
> documentation [ ;-) ] ...
<heh-heh>
>...that made me understand how much better than
> Garnet it is,...
well I should make clear that I am speaking specifically about the
constraint aspect and overall model-building aspect (aggregates in
garnetese). looking at the Garnet doc for the first time in years just
now I see a lot of other stuff in there like Postscript output, none of
which I do.
>... I'd consider installing the copy of NT I have sitting
> around. But I probably want to do that even less than you want to
> install Debian
Agreed, installing NT sounds worse than Debian from what I hear. I just
buy systems and leave them alone.
>
> Your billing of this as Garnet-on-steroids makes me wonder, why don't
> you just port Garnet to ACL/Win32? Or make an actual
> Garnet-on-steroids, by feeding steroids to the actual Garnet code base?
Not sure if this is a point of confusion I created, but
"Garnet-on-steroids" was just a shorthand way of conveying the idea of
"GUI+constraints". I did not even know about Garnet until my framework
was done and I and my cohorts were so delighted with it I concluded
someone else must have done something similar before, went searching on
the web.
> Well, I'm a Linux/CMUCL/Garnet type, but I might.
Understood. But if someone such as yourself took a look at the tool
under win32 and expressed interest in working with it, that would give
me that much more incentive to slug it out with Debian. btw, did I see
somewhere that Mandrake or Suse have better installs, tho Debian is
superior overall? (Not to start a distro war.)
> This is pretty much
> the same as Question #1, btw :-)
Close, but in #2 I was casting the net wider trying to pull in those
Linux/CMUCL users who could get access to win32 long enough to see if
they liked my stuff. Given enough interest from folks such as yourself I
would then either get Linux installed or find a Linux/CMUCL person to
build a fasl distro of the dataflow engine under NDA.
So you are all Linux all the time? Or is the Mac your other OS?
kenny
clinisys
Kenny Tilton <·······@nyc.rr.com> writes:
> >...that made me understand how much better than
> > Garnet it is,...
>
> well I should make clear that I am speaking specifically about the
> constraint aspect and overall model-building aspect (aggregates in
> garnetese). looking at the Garnet doc for the first time in years just
> now I see a lot of other stuff in there like Postscript output, none of
> which I do.
Okay, now I see what you meant by that.
> > Your billing of this as Garnet-on-steroids makes me wonder, why don't
> > you just port Garnet to ACL/Win32? Or make an actual
> > Garnet-on-steroids, by feeding steroids to the actual Garnet code base?
>
> Not sure if this is a point of confusion I created, but
> "Garnet-on-steroids" was just a shorthand way of conveying the idea of
> "GUI+constraints". I did not even know about Garnet until my framework
> was done and I and my cohorts were so delighted with it I concluded
> someone else must have done something similar before, went searching on
> the web.
Well, okay, so you've got this Really Cool constraints system that
kick's KR's ass. I'd definately be interested in playing with such a
beast. I can understand your wanting to keep the source to yourself,
but that would make me less likely to replace Garnet with a
(whatever)-based system. I don't expect anyone but universities to
release software under do-whatever-you-want-with-it licenses, but once
I do have software under such a license, I'm not very likely to give
it up. Of course, if you did something like Qt and released a GPL
version and sold a commercial version, that would be close enough.
Then again, I'm completely full of shit because it's been something
like 2 years since I released a piece of Free Software :-)
> > Well, I'm a Linux/CMUCL/Garnet type, but I might.
>
> Understood. But if someone such as yourself took a look at the tool
> under win32 and expressed interest in working with it, that would give
> me that much more incentive to slug it out with Debian. btw, did I see
> somewhere that Mandrake or Suse have better installs, tho Debian is
> superior overall? (Not to start a distro war.)
They're easier to install if the installation goes well. If anything
is problematic, they're more difficult (caveat: this is second-hand
knowledge). Setting up a working CL environment on Debian is trivial.
It's something of a pain on any other distro.
--
/|_ .-----------------------.
,' .\ / | No to Imperialist war |
,--' _,' | Wage class war! |
/ / `-----------------------'
( -. |
| ) |
(`-. '--.)
`. )----'
From: Dr. Edmund Weitz
Subject: Re: CLIM looks cool, but...a new CL community poll
Date:
Message-ID: <m3d72f3ge5.fsf@duke.agharta.de>
···@apocalypse.OCF.Berkeley.EDU (Thomas F. Burdick) writes:
> They're easier to install if the installation goes well. If
> anything is problematic, they're more difficult (caveat: this is
> second-hand knowledge). Setting up a working CL environment on
> Debian is trivial. It's something of a pain on any other distro.
[Standard Disclaimer: I don't want to start a distro flame war
either.]
With the recent versions of SuSE you just install the CLISP RPM from
the DVD or CD-ROM and that's it - AFAIK this is also true for newer
Mandrakes. Both of them still don't have any support for CMUCL[1] (or
SBCL) and that's a pity. If you want CMUCL 'out-of-the-box' you have
to go with Debian or FreeBSD.
Best regards,
Edi.
[1] While I was writing this I thought that this might be due to the
fact that both SuSE and Mandrake are available for a couple of
different platforms including PowerPC. CLISP is available for all
of these platforms and thus makes it easier for the maintainers of
the distro. (Yeah, I know that Debian supports at least as much
platforms. But given the fact that Lisp probably has a priority
next to zero for the commercial distros one shouldn't be surprised
that they've chosen not to include CMUCL.)
···@agharta.de (Dr. Edmund Weitz) writes:
> ···@apocalypse.OCF.Berkeley.EDU (Thomas F. Burdick) writes:
>
> > They're easier to install if the installation goes well. If
> > anything is problematic, they're more difficult (caveat: this is
> > second-hand knowledge). Setting up a working CL environment on
> > Debian is trivial. It's something of a pain on any other distro.
>
> [Standard Disclaimer: I don't want to start a distro flame war
> either.]
>
> With the recent versions of SuSE you just install the CLISP RPM from
> the DVD or CD-ROM and that's it - AFAIK this is also true for newer
> Mandrakes. Both of them still don't have any support for CMUCL[1] (or
> SBCL) and that's a pity. If you want CMUCL 'out-of-the-box' you have
> to go with Debian or FreeBSD.
Wow, that's good news. But, as far as I know, neither of them have
ILISP packaged. If they had CLISP, Emacs, and ILISP packaged so you
could just install the packages, then they'd join Debian in making the
setting up of a *working* CL *environment* trivial. Somehow, I don't
think this is one of their commercial priorities, though :-). Last I
saw, there weren't RPMs for ILISP nor CMUCL on rpmfind.net (I think
that's the place where RPM-based folks keep their
community-contributed stuff, right?). Hopefully this has or will
change. Of course, that's still not as easy as Debian, but users of
commercial distros get used to getting their software from more than
their distro, and pretty quickly.
--
/|_ .-----------------------.
,' .\ / | No to Imperialist war |
,--' _,' | Wage class war! |
/ / `-----------------------'
( -. |
| ) |
(`-. '--.)
`. )----'
From: Friedrich Dominicus
Subject: Re: CLIM looks cool, but...a new CL community poll
Date:
Message-ID: <873d3b1ak3.fsf@frown.here>
···@agharta.de (Dr. Edmund Weitz) writes:
> ···@apocalypse.OCF.Berkeley.EDU (Thomas F. Burdick) writes:
>
> > They're easier to install if the installation goes well. If
> > anything is problematic, they're more difficult (caveat: this is
> > second-hand knowledge). Setting up a working CL environment on
> > Debian is trivial. It's something of a pain on any other distro.
>
> [Standard Disclaimer: I don't want to start a distro flame war
> either.]
>
> With the recent versions of SuSE you just install the CLISP RPM from
> the DVD or CD-ROM and that's it - AFAIK this is also true for newer
> Mandrakes. Both of them still don't have any support for CMUCL[1] (or
> SBCL) and that's a pity. If you want CMUCL 'out-of-the-box' you have
> to go with Debian or FreeBSD.
Hm looking on a Debian site reveals:
unstable
clisp-doc 1:2.27-0.1 (277k)
GNU CLISP, a Common Lisp implementation (documentation)
stable
clisp 1999-07-22-5 (2940.1k)
a Common Lisp implementation
testing
clisp 2000-03-06-2 (4765.2k)
a Common Lisp implementation
unstable
clisp 1:2.27-0.1 (3188.8k)
GNU CLISP, a Common Lisp implementation
Seems as if you got .deb Package for both CMUCL and CLisp.
Regards
Friedrich
From: Dr. Edmund Weitz
Subject: Re: CLIM looks cool, but...a new CL community poll
Date:
Message-ID: <m3wv0nf85b.fsf@duke.agharta.de>
Friedrich Dominicus <·····@q-software-solutions.com> writes:
> ···@agharta.de (Dr. Edmund Weitz) writes:
>
>> If you want CMUCL 'out-of-the-box' you have to go with Debian or
>> FreeBSD.
>
> [...]
>
> Seems as if you got .deb Package for both CMUCL and CLisp.
That's correct of course. I didn't want to imply that CLISP isn't
supported by Debian (or FreeBSD). I somehow took it for granted that
it is and that everybody knows... :)
Thanks for clearing this up,
Edi.
From: Kenny Tilton
Subject: Re: CLIM looks cool, but...a new CL community poll
Date:
Message-ID: <3BF84B70.89168342@nyc.rr.com>
"Thomas F. Burdick" wrote:
>
> Well, okay, so you've got this Really Cool constraints system that
> kick's KR's ass.
Nah, the constraints crowd turns their nose up at the linear dataflow
model I have implemented. They are hot for multi-way dependencies, and I
can understand the ambition, but I am just a humble application
programmer so I just focused on getting the app done and have yet to
come across a requirement for that in the problems I have addressed. I
do have some tolerance of circularities and in the first implementation
there was some hairy stuff which worked out a solution given partial
information and some rules, but after a long time I realized it was
unnecessary for my needs. Could easily make a comeback, tho.
Oops. Just dug into the KR doc and I am not sure /they/ do any more than
linear dataflow. Anyway...
One advantage I think I have over Garnet (might have missed it) is that
(in garnetese) the parts of an aggregate can be defined by a rule. that
is huge.
Another more fundamental difference is that mine is an eager evaluation
model, and without that you do not have dataflow. The KR talk of not
re-evaluating a formula "until it is needed" means the application as a
whole is still making the world turn. with my stuff some OS event just
kicks off a cascade of re-evaluation and (again in garnetese) demon
dispatching and then yer done, next event please.
Also my stuff is pure CLOS, obvious advantages there. Also I have spent
a /lot/ of effort on optimizing the stuff and performance is dandy.
Finally, while the dataflow is merely linear I have a little more fun
with it than did KR. I have varieties of what I call Semaphors
(constrained slots, no relation to Unix semaphores) such as deltas,
aggregates, ephemerals, and I have Synapses which can mediate the
dataflow between two semaphors in interesting ways, especially to
improve performance. For example, where A depends on B, a Sensitivity
synapse can be used to defer re-evaluation of A until B has changed by a
certain amount.
> I'd definately be interested in playing with such a
> beast. I can understand your wanting to keep the source to yourself,
well, I think the technology is interesting enough that it should be
opened up for others to explore. OTOH, hell, there are lots of systems
like mine out there, no one really needs my source. If Semaphors catch
on anyone could re-implement from scratch the same functionality. eg,
the COSI system already has a lot of similarities, including having the
parts of an aggregate determined by a rule.
That said, sharing the source is probably the way to go, I just need to
stare at my navel a little longer before going that route.
> Of course, if you did something like Qt and released a GPL
> version and sold a commercial version, that would be close enough.
You know, it just dawned on me that the GPL (which I never liked because
I like to sell stuff) has that excellent virtue of letting commercial
code be shared.
> They're easier to install if the installation goes well. If anything
> is problematic, they're more difficult (caveat: this is second-hand
> knowledge). Setting up a working CL environment on Debian is trivial.
thx for the info. i was also worried about getting CMUCL cooking after
getting Linux up, so if I have to have a hassle it may as well be with
the Debian install.
kenny
clinisys
I spent a fair amount of time working on Garnet when I was using it
for project work. I found that it could do a lot of things pretty
well and some other things not so well. One thing I really wanted was
to be able to create an interactive stream (and thus make listeners,
little text windows that the application can create as streams etc.).
I found that the Garnet Multitext component had serious efficiency
problems. It wasn't really a serious attempt at a text component.
If someone wants to look at the Garnet interactive stream stuff I
wrote, you can get it at
ftp://ftp.csl.sri.com/pub/users/gilham/garnet/example/interactive-streams.tar.gz
This runs under CMUCL. It includes the TL code from Rick Taube.
Compile and load it and then type (tl::my-tl).
I wrote interactive streams that at least partly worked for Garnet,
CLUE and EW. I think that might be a record. :-)
Another limitation I found in Garnet was lack of a `canvas' component
that you could just draw into using either Garnet-level calls or CLX
level calls, that would handle updating etc. in some reasonable way.
This is not the Garnet model, since it backs each drawn element with
an actual object, creating a lot of overhead.
I ended up using Garnet pixmaps to do things like this but I wasn't
completely happy with it --- one can only set individual pixels, not
draw lines or circles or whatever. An example of code I wrote using
this is the fmand.lisp program in the above FTP directory. I never
finished the stuff to set the colors that the thing uses to draw the
mandelbrot set. The .xpm files in the same directory were drawn using
this program (with different color settings and size).
One can compile and run this and type (start). It uses the CMUCL
multiprocess stuff if it is present.
I also tried implementing the unit coversion program I found in the
Java Tutorial book. I got excited about using constraints to do it,
then found out that the constraint model was limited and I really
needed multi-way constraints. That code is the converter.lisp code in
the above directory. One can compile and load this and then run it
with (cv::start). One will note that *almost* everything works, but
there was one relationship I didn't know how to propagate using
constraints.
--
Fred Gilham ······@csl.sri.com
"...If reason is no longer a faculty enabling us to discern some
cosmic design, then it is no longer clear what claim reason has upon
us, or upon our politics. On the contrary, the sorts of habits that we
associate with `reasoning' --- that is, reflecting and discussing and
trying to achieve some kind of coherence among our various opinions
--- can come to seem like nothing more than the mental preferences (or
prejudices) of a particular class of people --- the dogmatic
proceduralism of an educated elite." --- Steven D. Smith
From: Kenny Tilton
Subject: Re: Garnet sans steroids (was Re: CLIM looks cool, but...a new CL community poll)
Date:
Message-ID: <3BF99795.246E8D8@nyc.rr.com>
Fred Gilham wrote:
>
> I spent a fair amount of time working on Garnet when I was using it
> for project work. I found that it could do a lot of things pretty
> well and some other things not so well. One thing I really wanted was
> to be able to create an interactive stream (and thus make listeners,
> little text windows that the application can create as streams etc.).
> I found that the Garnet Multitext component had serious efficiency
> problems. It wasn't really a serious attempt at a text component.
We executed a multitext component using Semaphors, seemed pretty fast,
but maybe we were not as feature-rich as Garnet's. I would have to
re-execute that, though, that bit is proprietary to CliniSys.
> Another limitation I found in Garnet was lack of a `canvas' component
> that you could just draw into using either Garnet-level calls or CLX
> level calls, that would handle updating etc. in some reasonable way.
> This is not the Garnet model, since it backs each drawn element with
> an actual object, creating a lot of overhead.
I have an object behind each drawn element as well, but nothing would
prevent an object from internally representing subobjects if you will in
a very lightweight fashion. The invalidation scheme would have to be
enhanced a whisker to let such an object optimize invalidation, but the
redraw mechanism already supplies the bitmap and invalid region, so the
object's paint mechanism could repaint subobjects only as needed.
>
> I ended up using Garnet pixmaps to do things like this but I wasn't
> completely happy with it --- one can only set individual pixels, not
> draw lines or circles or whatever.
I guess that has to do with trying to have a cross-platform pixmap
abstracted from the native OS capabilities? Maybe cross-platform is a
dangerous design goal. [See "Java".] Or maybe the issue is how one goes
after that goal.
> I also tried implementing the unit coversion program I found in the
> Java Tutorial book. I got excited about using constraints to do it,
> then found out that the constraint model was limited and I really
> needed multi-way constraints. That code is the converter.lisp code in
> the above directory.
I faced the same problem in a couple of places. One is the thumb of a
scrollbar. It both affects and reflects the scroll position. It has
occurred to me that a "see-saw" semaphor to handle such a thing would be
unambiguous hence not as nasty as something like a = b + c where the
engine does not know what to do when any one value changes so you have
to go to great lengths not to get unexpected outcomes. But I was in a
hurry so I gave the thumb and scroll positions echo functions (Garnet
says "demon", I say "echo") to setf the other, then resurrected the
Semaphor "circular propagation" detector so propagation stops when the
semantic wave gets back to someone already changing.
In another place, something like a text-only converter (without
sliders), I did the same thing except avoided the circularity even
arising by not changing the field with the keyboard focus. ie, that's
the value I want converted, don't calculate it!
kenny
clinisys
From: Kenny Tilton
Subject: Re: CLIM looks cool, but...a new CL community poll
Date:
Message-ID: <3BF805F6.83DE72CA@nyc.rr.com>
yes, vmware is great (i am told) and that's what i did, DLed the trial
VMWARE and then tried to install Debian. Got CDs from a book and also
from one of the vendors selling it. Both seemed to be the potato
release, but someone at debian said use woody. that was one depressing
event--why did the vendor send me potato? the book i can understand.
linux install got old so i punted, but vmware was a joy, utterly
painless to deal with.
kenny
clinisys
······@hushmail.com wrote:
>
> On Sun, 18 Nov 2001 08:06:26 GMT, Kenny Tilton <·······@nyc.rr.com> wrote:
>
> >Agreed, installing NT sounds worse than Debian from what I hear. I just
> >buy systems and leave them alone.
>
> This may help then:
> http://www.vmware.com/
From: Bulent Murtezaoglu
Subject: Re: CLIM looks cool, but...a new CL community poll
Date:
Message-ID: <87d72e1oj5.fsf@nkapi.internal>
>>>>> "KT" == Kenny Tilton <·······@nyc.rr.com> writes:
KT> ... that would give me that much more incentive to slug it out
KT> with Debian. btw, did I see somewhere that Mandrake or Suse
KT> have better installs, tho Debian is superior overall? (Not to
KT> start a distro war.) ...
What exactly is the problem you were having with the Debian install?
While this seems OT in cll, I think enough people run/recommend Debian
as the distribution of choice for Common Lisp on Linux that we'd be
interested in finding out about problems people are having with the
install.
cheers,
BM
From: Kenny Tilton
Subject: Re: CLIM looks cool, but...a new CL community poll
Date:
Message-ID: <3BF98222.37E33DC9@nyc.rr.com>
Bulent Murtezaoglu wrote:
>
> >>>>> "KT" == Kenny Tilton <·······@nyc.rr.com> writes:
>
> KT> ... that would give me that much more incentive to slug it out
> KT> with Debian. btw, did I see somewhere that Mandrake or Suse
> KT> have better installs, tho Debian is superior overall? (Not to
> KT> start a distro war.) ...
>
> What exactly is the problem you were having with the Debian install?
Insufficient motivation. :)
First I fought thru the partitioning mystery thanks to help I found on
the Debian site, but I thought it was a little harsh that to get past an
install step I had to go poking around online.
Another issue was that my graphics card was not recognized. I posted a
question on the Debian site and got an answer right away, viz. that that
card's manufacturer had a Linux driver I could DL. Sounded easy, as long
as I could get Linux to reach the Web thru my cable modem. That had me
worried since I had not been able to give a good answer when the install
asked me what number to dial to reach my ISP. But i did not get that
far, the next thing I tried was getting the X configuration to recognize
my mouse. The mouse is IntelliMouse compatible, there was an
IntelliMouse option, no dice. Nor with any other mouse option.
Also I had gotten a pair of cds from a vendor (instead of a book) and it
was still potato, while the bloke who answered my message said I should
use woody. so why did a vendor send me potato?
They say the difference between an adventure and a disastrous experience
is the attitude you take into it, and my attitude was "this better be
like an app install". But this was an OS install, so I had the wrong
attitude. The problem is, I do not want to install Linux, I want to
share my stuff.
So nothing insurmountable, just the usual BS, but enough resistance to
make me put the project off until a solid motive materialized, such as
someone interested in my stuff who was on that platform.
kenny
clinisys
From: Bulent Murtezaoglu
Subject: Re: CLIM looks cool, but...a new CL community poll
Date:
Message-ID: <87adxi1ili.fsf@nkapi.internal>
>>>>> "KT" == Kenny Tilton <·······@nyc.rr.com> writes:
[...]
KT> First I fought thru the partitioning mystery thanks to help I
KT> found on the Debian site, but I thought it was a little harsh
KT> that to get past an install step I had to go poking around
KT> online.
I can see why this would happen with the installer dropping you right
into cfdisk.
KT> Another issue was that my graphics card was not recognized. I
KT> posted a question on the Debian site and got an answer right
KT> away, viz. that that card's manufacturer had a Linux driver I
KT> could DL. [...]
I used to recommend to people who had some networking know-how
available to them to use an X server on their Windows box and run
linux pretty much headless on a separate machine with possibly samba
providing some file sharing. I haven't had occasion to recommend and
follow this through recently, but I think it is still a good
recommendation especially if the Windows box has to be up all the time
and is getting the good monitor. I understand X over VNC is pretty
snappy now, so that might be worth looking into also especially since
it will ensure that your X context will survive a Windows reboot. The
only problem that I can see with the set-up I am outlining is that
Windows tends to grab meta(alt)-tab which I and I suspect most others
use for completion under ILISP.
Under the scheme above, you can just do a minimal text-based install of
Debian, make sure it can network and not deal with the X issues at all.
I think this is a good solution even if bringing up X/sound etc. were
painless since you get to use your favorite chair/desk/keyboard/monitor
and such w/o disturbing your main system.
I would like to hear others' opinions and experiences on this as I think
there are now quite a few good lispers using Windows who would like to at
least test their stuff on Linux.
[...]
KT> They say the difference between an adventure and a disastrous
KT> experience is the attitude you take into it, and my attitude
KT> was "this better be like an app install". But this was an OS
KT> install, so I had the wrong attitude. The problem is, I do not
KT> want to install Linux, I want to share my stuff.
[...]
Please consider what I am saying above. If that looks like too much
hassle still, maybe we can cook-up some documentation and/or images for
that kind of a set-up.
cheers,
BM
From: Kenny Tilton
Subject: Re: CLIM looks cool, but...a new CL community poll
Date:
Message-ID: <3BF99C4F.96671718@nyc.rr.com>
Bulent Murtezaoglu wrote:
> I used to recommend to people who had some networking know-how
> available to them to use an X server on their Windows box and run
> linux pretty much headless on a separate machine with possibly samba
> providing some file sharing.
Are you saying I would have an easier time installing Linux as the only
OS on a box? I do have a spare box (a 450mhz dog) sitting here under my
desk and I already have two monitors in front of me, so I 'spose I could
just lose the NT setup altogether on the old box. I had been wondering
if going native would help in some way.
But if it don't like my mouse or graphics card or cable modem under
VMWare it won't like them native, I reckon. Or did I misunderstand you
completely? I am in way over my head. :)
>
> Please consider what I am saying above. If that looks like too much
> hassle still, maybe we can cook-up some documentation and/or images for
> that kind of a set-up.
ok, thx for the thought. right now I am starting to get more realistic
with my objectives, and I might just focus on sharing the dataflow
engine with everyone and win32 GUI framework with other EEers. I would
maybe release the source under NDA to individuals on diff platforms to
build FASLs for sharing.
even if i get Semaphors working under CMUCL it would still be a big job
to port the GUI to CLX. Not /huge/, but since this is a side pursuit it
would likely never get done. that can wait until I see if anyone else
even likes Semaphors.
kenny
clinisys
From: Hartmann Schaffer
Subject: Re: CLIM looks cool, but...a new CL community poll
Date:
Message-ID: <3bf9ab9a@news.sentex.net>
In article <·················@nyc.rr.com>,
Kenny Tilton <·······@nyc.rr.com> writes:
> ...
> Are you saying I would have an easier time installing Linux as the only
> OS on a box? I do have a spare box (a 450mhz dog) sitting here under my
> desk and I already have two monitors in front of me, so I 'spose I could
> just lose the NT setup altogether on the old box. I had been wondering
> if going native would help in some way.
if you have a second machne i would always recommend a sperate
install, but once you have made room for (an) extra partition(s), a
multiboot install should also be straightforward. Check out the HOWTO
section on http://www.linuxdoc.org. it contains pretty detailed
advice how to prepare a window system so that you can install linux as
a second system (follow links -> HOWTOs -> miniHOWTOs, then look for
"Linux+NT")
> But if it don't like my mouse or graphics card or cable modem under
> VMWare it won't like them native, I reckon. Or did I misunderstand you
> completely? I am in way over my head. :)
There is also a Hardware HOWTO that gives an overview over supported
hardware, including some pointers where to fnd more information
hs
--
Apart from the obvious disagreement about who the good guys are, what
is the difference between "You are either with us or against us" and
"There are only good muslim and infidels"?
From: Bulent Murtezaoglu
Subject: Re: CLIM looks cool, but...a new CL community poll
Date:
Message-ID: <87y9l2xpoh.fsf@nkapi.internal>
>>>>> "KT" == Kenny Tilton <·······@nyc.rr.com> writes:
KT> [...] Are you saying I would have an easier time installing Linux as
KT> the only OS on a box?
Yes I am saying that + under the scheme I propose you interface with
fewer pieces hardware under linux. You just need to get networking up
and install a few packages.
KT> I do have a spare box (a 450mhz dog)
KT> sitting here under my desk and I already have two monitors in
KT> front of me,
I was thinking that once configured, the linux machine could be hidden
somewhere and run headless.
KT> But if it don't like my mouse or graphics card or cable modem
KT> under VMWare it won't like them native, I reckon. Or did I
KT> misunderstand you completely? I am in way over my head. :) [...]
Not completely, partially. You don't need it to see your mouse,
graphics card etc. since it is not using those. I am assuming your
cable modem is being shared through the network or that you will be
able to share it (possibly with one of those 4-5 port
hub/gateway/NAT/router devices).
As Bruce Hoult indicated people prefer reaching their Linux box
through X or VNC for other good reasons, but in your case I think
small steps with almost immediate positive results might provide
additional motivational benefits.
cheers,
BM
In article <··············@nkapi.internal>, Bulent Murtezaoglu
<··@acm.org> wrote:
> I used to recommend to people who had some networking know-how
> available to them to use an X server on their Windows box and run
> linux pretty much headless on a separate machine with possibly samba
> providing some file sharing. I haven't had occasion to recommend and
> follow this through recently, but I think it is still a good
> recommendation especially if the Windows box has to be up all the time
> and is getting the good monitor. I understand X over VNC is pretty
> snappy now, so that might be worth looking into also especially since
> it will ensure that your X context will survive a Windows reboot.
I've never had a problem doing an install on my Linux box, but that is,
all the same, the way I normally use it, because I simply don't have rom
on my desk for as many machines as I'm using simultaneously!! I mean,
here at home I've got an Athlon 700 running Linux, a PIII/1000 running
W2K, and a G4/867 running OS X. Oh, and the old 266 MHz powerbook I'm
typing this on...
The P3 has a 17" IBM Trinitron on it, and the G4 has a 17" LCD, which
together take up quite enough desk space. So the Linux box is accessed
using VNC from the Windows machine, or using XDarwin or VNC from the Mac
-- or ssh from anything (PuTTY/NiftyTelnet).
I'm developing some cross-platform software (using C++, alas...) and
it's pretty handy to be able to either have things from two different
machines (in more or less any combination, thanks to VNC) on two
different screens, but also to be able to alt-tab between two machines
on the same screen.
Of course everything is kept synched using a CVS repository on the Linux
machine.
Performance of VNC is just fine on 10baseT, but as it happens all the
machines except the PowerBook have 100baseT anyway and my DSL modem is a
10/100 switch...
-- Bruce
From: Juliusz Chroboczek
Subject: Linux without hassles [was: CLIM looks cool, but...]
Date:
Message-ID: <87elmqej1o.fsf_-_@pps.jussieu.fr>
>> What exactly is the problem you were having with the Debian install?
KT> Insufficient motivation. :)
I would like to recommend DemoLinux for this sort of experimentation.
http://www.demolinux.org/
DemoLinux was designed to provide a simple, no-installation Linux
environment without requiring complex tools such as VMware.
Please note that by default, DemoLinux runs from the CD decompressing
itself on the fly. It may, therefore, require a reasonably fast
machine with a spiffy CD-ROM drive.
(Somebody really should work out a simple way of installing Clisp and
CMUCL under DemoLinux, and share his experiences with the ng as well
as with Roberto Di Cosmo, DL's main author.)
Juliusz
Forgive me jumping in here, but (i.m.h.o., of course) the biggest thing that
Lisp GUIs need is *not*:
- more expressive and powerful constraint systems,
- better rendering and graphics models,
- spiffier wysiwyg program-framework builders,
- super-optimizing methods of incremental redisplay.
What they need most is much much clearer modularity boundaries so that all
the goodies above can be provided on an incremental and interchangeable
basis.
If you've got an idea for a cool window-layout-constraint solver, my
challenge to you is: implement it so that it can run on top of CLIM's sheets
or Garnet's equivalents, or raw X windows.. Measure your success not only
by how much functionality your constraint system provides, but also by how
*little* glue code is required for attaching it to each underlying system.
Help us all by teasing out the protocol between a constraint solver and the
rest of a GUI.
If you've got an idea for a wizzy input editor, my challenge to you is:
implement it so that it can run with or without a presentation system
producing all sorts of ptyped gestures in the middle of the input. (Or with
a different sort of context providing non-character input that depends on
the state of parsing.)
The goal is to break apart the huge monolithic choice that one has to make
when selecting GUI technology. Look at the number of posters asking which
toolkit to try for their next project... why don't they just try a few, and
form their own opinion? Because the barrier to entry, or the penalty for
changing your mind, is huge. Let's reduce that penalty.
John
From: Kenny Tilton
Subject: Re: CLIM looks cool, but...a new CL community poll
Date:
Message-ID: <3BFEA43D.BA7AF3D2@nyc.rr.com>
I should clarify that I am going on about a CL GUI because I think we
need one and more importantly because no one will get anywhere with
Semaphors if they have to start by doing a ton of work to get back to
where they already are happily programming toolkits procedurally.
The dataflow approach really is a paradigm shift. It is hard to mix with
a procedural approach to high-level application management. So I want to
give everyone a running start by providing a fun GUI to play with. It
will just use a native OS window and event stream and take it from there
(tho also manage OS widgets as a side project). If folks dig that we can
look at standardizing on X or something, but for now I just want to get
the ball rolling.
John Aspinall wrote:
> If you've got an idea for a cool window-layout-constraint solver
It's not just for window layout, it (the dataflow engine) works between
any slots defined to handle dataflow. (I have toyed with an approach
that works without touching SVUC, hence would not require slots be
defined specially, but it does not look as correct as the current
MOP-based approach.)
Also, just a reminder, to me a constraint solver is something that works
with multi-way dependencies and accepts rules only partially specifiying
outcomes. I just do dataflow, something akin to a spreadsheets for clos
instances.
> , my
> challenge to you is: implement it so that it can run on top of CLIM's sheets
> or Garnet's equivalents, or raw X windows..
I have already done so on the MCL and under ACL, gluing the system to
native OS (MCL) and framework (ACL/CG) widgets. I wasn't crazy about the
latter, but I plan to dig up that code and include it in the distro so
anyone who wants to add Semaphors to CLIM or X or whatever can do so.
> Help us all by teasing out the protocol between a constraint solver and the
> rest of a GUI.
That's easy. Basically you build a hieararchy of instances all connected
by dataflow. Some of them are wrappers for widgets from your targeted
GUI toolkit. What I call "echo" functions the programmer supplies are
called whenever a slot changes (and what it means to change is all user
definable). Anyway, when the model widget moves, an echo function simply
moves the wrapped CG or X or whatever widget.
kenny
clinisys