Arguments against newlisp (pro-SBCL)

Started by statik, August 05, 2005, 11:49:10 AM

Previous topic - Next topic

statik

Taken from a debate in the #newlisp channel on freenode, here was this guy's complaints/arguments:



<someguy> well, newlisp's memory manager really sucks

<someguy> because it doesn't handle shared structure

<someguy> it copies objects when you pass parameters

<someguy> because its inefficient

<someguy> also it means you can't have first-class arrays or hash tables

<someguy> you have to name all arrays and hashes by symbols, and pass symbols, you can't just pass an anonymous hash

<someguy> it severely limits the language

<someguy> performance?

<someguy> sbcl is much faster

<someguy> except the design of newlisp is quite poor

<someguy> and i noticed the flawed memory management, and lack of first-class arrays and hashes

<someguy> more problems in newlisp than high-quality common lisps like sbcl

<someguy> in fact, i don't see anything newlisp is good for

<someguy> i've also looked at the implementation code, its a mess

<someguy> in that sbcl is better in every way

<someguy> it has first class arrays, hashes, real gc, a better object system

<someguy> newlisp is a strict subset of common lisp, except the really broken parts (copying all values when passing parameters, 'contexts')

<someguy> also there's no number tower

<someguy> so integer math overflows

<someguy> that's really bad

<someguy> you can have a file larger than 2gb, but newlisp cannot represent its size

<someguy> algorithms like md5 and sha1 hashing, and rsa crypto all use bignums



I could argue all day with this guy, but in the end, I'm really only interested in Lutz's commentary. So if you wouldn't mind...



Thanks.
-statik

newdep

#1
Well ..whatever they say ;-)



It only points out that Newlisp is a real competitor to some.. ;-)



Ooo baiiy The HighWay.. What is sbcl ??

If its something intresting i would have known...
-- (define? (Cornflakes))

statik

#2
Steel Bank Common Lisp -- http://www.sbcl.org/">http://www.sbcl.org/
-statik

newdep

#3
Well i had a quick look but it sure isnt something i would use...



I like Lisp ..

...I like Lisp dialects..

........But i dont like Lookalike Lisp dialacts..







And let me explain that..



If you reinvent the wheel then do something about syntax,

most programmers dies days dont even know what that is.



A programming Language is judged by its elegance

within a 10th of a second by looking at some source code.

(...That means life or /dev/null...)
-- (define? (Cornflakes))

PaipoJim

#4
Quote from: "statik"Taken from a debate in the #newlisp channel on freenode, here was this guy's complaints/arguments:



<someguy> well, newlisp's memory manager really sucks

<someguy> because it doesn't handle shared structure


Put them in a context.


Quote
<someguy> it copies objects when you pass parameters

<someguy> because its inefficient

<someguy> also it means you can't have first-class arrays or hash tables


Put them in a context.


Quote
<someguy> you have to name all arrays and hashes by symbols, and pass symbols, you can't just pass an anonymous hash

<someguy> it severely limits the language


Not really.  The "S" in LISP stands for "Symbolic".  LISP is all about symbols.


Quote
<someguy> performance?

<someguy> sbcl is much faster

<someguy> except the design of newlisp is quite poor

<someguy> and i noticed the flawed memory management,


A memory management system not to his taste perhaps.  NewLISP imemory management is certainly well described though.


Quote
 and lack of first-class arrays and hashes


Redundant as well as incorrect.  Again, use contexts if you must.


Quote
<someguy> more problems in newlisp than high-quality common lisps like sbcl

<someguy> in fact, i don't see anything newlisp is good for


Try: http://newlisp.org/index.cgi?page=Applications">http://newlisp.org/index.cgi?page=Applications


Quote
<someguy> i've also looked at the implementation code, its a mess

<someguy> in that sbcl is better in every way

<someguy> it has first class arrays, hashes, real gc, a better object system

<someguy> newlisp is a strict subset of common lisp, except the really broken parts (copying all values when passing parameters, 'contexts')


No, contexts are passed by reference.


Quote
<someguy> also there's no number tower

<someguy> so integer math overflows

<someguy> that's really bad

<someguy> you can have a file larger than 2gb, but newlisp cannot represent its size

<someguy> algorithms like md5 and sha1 hashing, and rsa crypto all use bignums


You've got me there.  Aren't there bignum C libraries available?  Can NewLISP not interface to them?


Quote
I could argue all day with this guy


I think you should.

Lutz

#5
Development version 8.6.1 contains a module gmp.lsp for interfacing to the GNU Multiple Precision Library. This file also works with the 8.6.0 release.



Regarding the other comments, you can point <someguy> to the FAQ:



http://newlisp.org/index.cgi?page=FAQ">http://newlisp.org/index.cgi?page=FAQ



which has pointers to benchmarks:



http://newlisp.org/benchmarks/">http://newlisp.org/benchmarks/



at the debian benchmarks site:



http://shootout.alioth.debian.org/great/benchmark.php?test=all&lang=newlisp&sort=cpu">http://shootout.alioth.debian.org/great ... p&sort=cpu">http://shootout.alioth.debian.org/great/benchmark.php?test=all&lang=newlisp&sort=cpu



he can see memory usage comparisons in a variety of benchmarks, most of them favor newLISP.



To learn about differences to other LISPs:



http://newlisp.org/index.cgi?page=Differences_to_Other_LISPs">http://newlisp.org/index.cgi?page=Diffe ... ther_LISPs">http://newlisp.org/index.cgi?page=Differences_to_Other_LISPs



and about memory management:



http://newlisp.org/MemoryManagement.html">http://newlisp.org/MemoryManagement.html



Finally her some intersting discussions about various aspects of LISPs:



http://newlisp.org/index.cgi?page=Links">http://newlisp.org/index.cgi?page=Links



Lutz

statik

#6
Just the response I expected. By the way, the author of these accusations, slanderings, and or misunderstandings is the creator of the Factor programming language. And yes, I would have loved nothing more than to argue with this guy, but the stubborness was obvious. He wasn't there to understand Newlisp, rather, he was there to troll (promoting his language along the way).
-statik

nigelbrown

#7
My problem with sbcl and what made newlisp a winner is how newlisp runs equally on windows and linux (windows OS shortcomings excluded).



To me the common lisp competition to newlisp comes from clisp - but I prefer newlisp-tk over remembering emacs/slime.



Nigel