Le 08/11/2011 09:31, William Stein a écrit :
On Tue, Nov 8, 2011 at 12:09 AM, Julien Puydt<julien.pu...@laposte.net>  wrote:
Sage developers shouldn't care. If you start down that road, you'll soon end
up putting in your own libc, your own libc++, your own editor, your own
C/C++/whatever compiler, etc.

I disagree with your prediction.  We started down that road over 6
years ago, and we don't include our own libc, libc++, or editor in
Sage.   (We include a Fortran compiler on OS X, since Fortran on OS X
is a mess.)

I have to stress that I know the past. What has been done, and especially by you, William Stein, was good. Indeed, making it easy to install anywhere, bringing the existing pieces of open source math software together and building on them : that is the reason why sage is where it is today.

With hindsight, one can say that it was necessary to bootstrap things. I think sage is past that boostrap stage.

Making sure the deps for the sage package they have built are available is a
packager's problem -- not a developer's.
[...]
You seem to think that you're arguing against my point, but I don't think
that's really the case : in fact, you explain well how important it is to
have sage handily distributed. And I fully agree, it's mango flesh. Then you
lightly conclude that sage=distribution+program is the only way out of. And
there I fully disagree : you've just swallowed the mango's kernel!

Having a sage=distribution organisation has the problems I already pointed
out ; let me remind them:
- developers use too much of their time worrying about deps instead of
making things better ;

In fact, it is the opposite.  If we didn't have our own distribution,
developers would spend way more time worying about deps.  Because of
having a distribution, developers usually don't have to worry about
deps.  That's Jon's point.

Again, this is about the past.

If you add "foo" to your deps, anybody using sage will know they have "foo" available. There's no need to bundle "foo" with sage to gain that.

Much more likely is that without our own distribution, there would be
no developers, so indeed developers wouldn't spend their time worrying
about deps (since Sage wouldn't exist).

Same remarks as above apply.

- sage is too big to fit in the newer smaller computing devices available
today and tomorrow ;

If that is true (and I don't believe it is), then the same is true for
all of Sage's direct competitors: Mathematica, Maple, and Matlab.

Yes. Definitely. Why wouldn't sage be better?

Many of us are absolutely 100% opposed to getting rid of what we
currently have.

I am too. I merely point the benefits there would have if sage were easier to install within the normal distributions, and not as a big (oh, so big...) source package bundling almost every deps it has.

However, I see no reason that you can't *also* make a version of Sage
that has the properties you want.   The Gentoo people do that, the
Mandriva people do that, Burcin with his Gentoo-prefix thing is doing
that, and it was even done for Debian once by Tim Abbot.   Do it.

Here I have to quote myself : <<<
I still want to remind that I'm not asking for an overnight switch or a revolution. What I want is people to :

1. be conscious of the matter ;
2. let little steps in the right direction happen ;
3. not go in the way by adding careless pitfalls...

so that when time is ripe, the switch will be easy. >>>

The fact that the gentoo&mandriva people are doing something is interesting ; the fact that for debian that was done but didn't stick is too.

In fact, it would be good if those people stepped in this thread to explain what they do, how well that works, and how things could be modified to ease things.

Perhaps I'm mistaken when I think it's anormally hard to package sage-as-a-program outside sage-as-a-distribution... After all, there's a reason why I'm not famous for being perfect and always right.

Snark on #sagemath

--
To post to this group, send an email to sage-devel@googlegroups.com
To unsubscribe from this group, send an email to 
sage-devel+unsubscr...@googlegroups.com
For more options, visit this group at http://groups.google.com/group/sage-devel
URL: http://www.sagemath.org

Reply via email to