From: Simon Cozens <simon@...>
Date: 2002-10-23T07:18:01+09:00
Subject: Re: RAA.succ?

why the lucky stiff <ruby-talk@whytheluckystiff.net> writes:
> I've spent some time reading the CPAN modules to get an idea of how
> they do mirroring and it just seems to involve FTP sites with metadata
> and modules.  I'd love to see us shack up with CPAN because we could
> certainly build a cool interface on top of the existing infrastructure.
> But I don't think there's enough inertia on our part or theirs to make
> this happen.

Some ideas from a Perler:

    I'm not sure that shacking up with CPAN is a great idea, since you'd
basically have to split the thing down the middle - otherwise you're in
a mess of name clashes. A better idea might be to grab the code which runs
the CPAN indexes and search.cpan.org and clone off a CRAN or RAA.succ or
something using the same codebase. (Maybe replacing it over time with
a Ruby version if you're unhappy running Ruby's central repository on
Perl code...)

    A single but distributed source for modules is brilliant. If I want a
Perl module, I know exactly where it'll be; I don't want to be going half
way around the Internet for things. Some hard-core Perlers keep a private
mirror of CPAN themselves so that they know that if at any point they need
additional modules, they've already got them.

   Indexing and categorizing is good, but one of the biggest problems with
CPAN has been the "official module list". 
(http://www.cpan.org/modules/00modlist.long.html) CPAN's advantage is that
it's a commons; anyone with an account can upload stuff, and so they do.
You get a lot of junk, but you get a lot of code. The idea of the module
list is to sort out some of the good code from the junk. It doesn't work
because it's a bureaucratic nightmare. A better idea (which is basically
what's happened over time) is to use social pressure to ensure that the
modules are self-categorizing by careful use of naming.

   I haven't looked at rpkg yet, but an automated package downloading
and installing tool is a really big win. Look at Perl's CPANPLUS.pm for
that.

Also, look to the Python Vaults of Parnassus for an idea of how *not* to
do things.

Finally:
    I think some structured, highly available, distributed repository
of Ruby libraries with easy client-side installation is crucial to the
success of Ruby as a mainstream language; it's certainly a key to Perl 
5's success.

-- 
"If you do not wish your beer to be served without the traditional head,
please ask for a top-up". With the subtext: "Your traditional head will 
then exit via the traditional window. Arsehole".
    - Mark Dickerson