Why Ice-9?

Why Ice-9?

[Noah Lavine]

I am not completely sure this is the right place to ask this, but why
do many of the module names in Guile start with 'ice-9'?

I can tell you that as a newcomer this is quite unintuitive.

Noah

Re: Why Ice-9?

[Andy Wingo]

On Fri 09 Jul 2010 19:59, Noah Lavine <noah.b.lavine@gmail.com> writes:

> I am not completely sure this is the right place to ask this, but why
> do many of the module names in Guile start with 'ice-9'?
>
> I can tell you that as a newcomer this is quite unintuitive.

Heh, I thought that too, once. 

  http://en.wikipedia.org/wiki/Ice-nine

The idea at the time (1998 or so) was that Guile's module system would
crystallize the mass of Scheme code out there. It didn't happen exactly
like that; in practice ice-9 is Guile's namespace.

Now, there is a deeper issue here -- a global Scheme namespace is
starting to emerge, and Guile is a bit all over the map. To me it's fine
to have e.g. statprof have the toplevel module, (statprof), even though
it's part of Guile; but in a way I feel that instead of having ice-9 and
system, we should just have (guile) as our module prefix, in the same
way that e.g. ikarus and chez scheme have (ikarus ...) and (chezscheme
...), respectively.

It's too late to do that this cycle, but perhaps during 2.0 or for 2.2
we could provide (ice-9 popen) as an alias to (guile popen), and
eventually for 2.4 deprecate both ice-9 and system. Just a thought,
though...

Andy
-- 
http://wingolog.org/

Re: Why Ice-9?

[Andreas Rottmann]

Andy Wingo <wingo@pobox.com> writes:

> On Fri 09 Jul 2010 19:59, Noah Lavine <noah.b.lavine@gmail.com> writes:
>
>> I am not completely sure this is the right place to ask this, but why
>> do many of the module names in Guile start with 'ice-9'?
>>
>> I can tell you that as a newcomer this is quite unintuitive.
>
> Heh, I thought that too, once. 
>
>   http://en.wikipedia.org/wiki/Ice-nine
>
> The idea at the time (1998 or so) was that Guile's module system would
> crystallize the mass of Scheme code out there. It didn't happen exactly
> like that; in practice ice-9 is Guile's namespace.
>
> Now, there is a deeper issue here -- a global Scheme namespace is
> starting to emerge, and Guile is a bit all over the map. To me it's fine
> to have e.g. statprof have the toplevel module, (statprof), even though
> it's part of Guile; but in a way I feel that instead of having ice-9 and
> system, we should just have (guile) as our module prefix, in the same
> way that e.g. ikarus and chez scheme have (ikarus ...) and (chezscheme
> ...), respectively.
>
> It's too late to do that this cycle, but perhaps during 2.0 or for 2.2
> we could provide (ice-9 popen) as an alias to (guile popen), and
> eventually for 2.4 deprecate both ice-9 and system. Just a thought,
> though...
>
+1 for that idea from me.

Rotty
-- 
Andreas Rottmann -- <http://rotty.yi.org/>

Re: Why Ice-9?

[Ludovic Courtès]

Hello,

Andy Wingo <wingo@pobox.com> writes:

> Now, there is a deeper issue here -- a global Scheme namespace is
> starting to emerge, and Guile is a bit all over the map. To me it's fine
> to have e.g. statprof have the toplevel module, (statprof), even though
> it's part of Guile; but in a way I feel that instead of having ice-9 and
> system, we should just have (guile) as our module prefix, in the same
> way that e.g. ikarus and chez scheme have (ikarus ...) and (chezscheme
> ...), respectively.

OTOH, ‘ice-9’ is unlikely to ever collide with anything else, so it’s
roughly equivalent to ‘guile’, just a little bit more obscure.  :-)

‘system’ may be more risky in that respect.  ‘guile’ would have been a
better choice, but I’m afraid it’s too late (in theory it isn’t,
but...).

Thoughts?

Thanks,
Ludo’.

Re: Why Ice-9?

[Thien-Thi Nguyen]

() ludo@gnu.org (Ludovic Courtès)
() Sat, 10 Jul 2010 16:17:07 +0200

   ‘system’ may be more risky in that respect.  ‘guile’ would have been a
   better choice, but I’m afraid it’s too late (in theory it isn’t, but...).

   Thoughts?

Perhaps we can start discussing "directory aliases", a feature whereby
(foo bar baz) can be specified to resolve (perhaps in multiple steps)
to (actual baz).  We alias only directories, not full module names, to
avoid having to deal with ‘%load-extensions’.

Design decisions to be made:

- Module name directory part (of aliases, of fully-resolved) 
  must be non-empty?

- Will dealing with ‘%load-extensions’ be in the future plans?

- Where/how to store alias info.  Can this mechanism be unified
  with one for resolved modules (e.g., Guile 1.4.x module catalogs)?

Anyway, just some thoughts.

thi

Re: Why Ice-9?

[Andy Wingo]

On Mon 12 Jul 2010 10:07, Thien-Thi Nguyen <ttn@gnuvola.org> writes:

> () ludo@gnu.org (Ludovic Courtès)
> () Sat, 10 Jul 2010 16:17:07 +0200
>
>    ‘system’ may be more risky in that respect.  ‘guile’ would have been a
>    better choice, but I’m afraid it’s too late (in theory it isn’t, but...).
>
>    Thoughts?
>
> Perhaps we can start discussing "directory aliases", a feature whereby
> (foo bar baz) can be specified to resolve (perhaps in multiple steps)
> to (actual baz).  We alias only directories, not full module names, to
> avoid having to deal with ‘%load-extensions’.

There are also submodule binders, in git guile, that can be used to
implement this.

> - Module name directory part (of aliases, of fully-resolved) 
>   must be non-empty?
>
> - Will dealing with ‘%load-extensions’ be in the future plans?
>
> - Where/how to store alias info.  Can this mechanism be unified
>   with one for resolved modules (e.g., Guile 1.4.x module catalogs)?

I don't understand the problem solved by these things; I would need more
details to comment, I think.

Andy
-- 
http://wingolog.org/

Re: Why Ice-9?

[Thien-Thi Nguyen]

() Andy Wingo <wingo@pobox.com>
() Tue, 13 Jul 2010 23:20:35 +0200

   There are also submodule binders, in git guile, that can be used to
   implement this.

Cool, i'll take a look at submodule binders.

   > - Module name directory part (of aliases, of fully-resolved) 
   >   must be non-empty?
   >
   > - Will dealing with ‘%load-extensions’ be in the future plans?
   >
   > - Where/how to store alias info.  Can this mechanism be unified
   >   with one for resolved modules (e.g., Guile 1.4.x module catalogs)?

   I don't understand the problem solved by these things; I would need more
   details to comment, I think.

The first and third are issues analogous to design of symbolic links on a
filesystem.  The ‘%load-extensions’ question is orthogonal.

thi