On Mon, Apr 21, 2008 at 3:22 AM, Ondrej Certik <[EMAIL PROTECTED]> wrote:
>
>
> On Mon, Apr 21, 2008 at 9:25 AM, mabshoff
> <[EMAIL PROTECTED]> wrote:
> >
> > On Apr 21, 9:09 am, Martin Rubey <[EMAIL PROTECTED]> wrote:
> > > Just a quick note now, maybe more later:
> >
> > <SNIP>
> >
> > Hi Martin,
> >
> >
> > > > "I know of three open source implementations of lisp that do not need
> to
> > > > bootstrap themselves"
> > >
> > > What's wrong with bootstrapping, and in particular with
> sbcl?http://www.sbcl.org/platform-table.html
> >
> > I like sbcl and it works well.
> >
> >
> > > Do you insist on building the lisp from scratch?
> >
> > Yes.
> >
> > > If so, why?
> >
> > Shipping binaries is not an option due to size constraints. Requiring
> > some external version of lisp is tricky and Maxima for example behaves
> > differently, i.e. default precision of floats is different, with
> > different lisp implementations. And Sage is supposed to be a compile
> > from scratch & self hosting experience and while it usually isn't a
> > problem to find or install some version of lisp on Linux it is a pain
> > to do so on Windows or OSX. And Sage is supposed to be "easy" to
> > install and run and every external requirement makes it more
> > complicated. There are also important Sage users who greatly
> > appreciate the fact that every bit in Sage can be delivered in source.
Just to emphasize -- the fact that Sage can be built from source
on the supported platforms is perhaps the most basic requirement
at the core of the Sage project.
Sage is three things: (1) an easily buildable-from-source distribution
of math software; (2) a new library; (3) interfaces. Part of the
definition of part (1) is that Sage be buildable from source
with only the following potentially nonstandard dependencies:
gcc, g++, make, m4, perl, ranlib, and tar
This has not changed since day 1 of Sage and it won't, except
that eventually we may include gfortran in the list.
Note -- we currently ship some Fortran binaries with the Sage
source tarball, but they can easily be deleted if one has gfortran,
which is fairly standard.
> >
> >
> > > In any case, I think that clisp is not a very good choice for FriCAS
> except
> > > that it is available almost everwhere. SBCL based FriCAS is *a lot*
> faster. I
> > > do not have a clisp version handy, but SBCL vs. GCL makes a *factor* of
> 2 for,
> > > eg., guessing. CLISP is still slower.
> >
> > Yes, we are well aware that clisp is quite bad performance wise
> > compared to gcl and especially sbcl, but we prefer a buildable lisp
> > over a potentially faster version that creates even more build
> > problems.
Hopefully the table here is wrong:
http://www.sbcl.org/platform-table.html
It says the only supported platform for the most recently
released version of SBCL is Linux MIPS. The most
recent before that -- version 1.0.12 ? -- says the only supported
platforms are:
Linux x86, Linux x86_64, and Mac OS X x86,
and that is it! There is no version that has ever supported Microsoft Windows
according to that chart.
So unless the Sage developers want to start spending our time
porting sbcl, then SBCL is a non-option. This is because we're
not going to start messing with using SBCL on one architecture, another
SBCL version on another architecture, GCL on another,
and clisp on yet a third. The more I think about the mess
that is lisp compilers, the more I wish we didn't have any lisp
in Sage.
> Well, you can first build clisp and then sbcl using clisp. :)
The *show stopper* problem with sbcl isn't building it but that
it doesn't support nearly enough platforms, and that nobody
seriously involved in Sage wants to work on porting lisp
compilers (as far as I know -- maybe I'm wrong?).
-- William
--~--~---------~--~----~------------~-------~--~----~
To post to this group, send email to sage-devel@googlegroups.com
To unsubscribe from this group, send email to [EMAIL PROTECTED]
For more options, visit this group at http://groups.google.com/group/sage-devel
URLs: http://www.sagemath.org
-~----------~----~----~----~------~----~------~--~---
