Comment 12 for bug 228345

On 28/05/08 at 08:39 -0000, Neil Wilson wrote:
> 2008/5/27 Lucas Nussbaum <email address hidden>:
> > I'm not sure you understand the rationale for splitting ruby1.9.
>
> So that other packages don't pull in stuff they don't need and reduce
> the perennial dependency problem. However those packages are a
> minority application. The majority want ruby1.x to include gems, to
> provide an executable called 'ruby', and supply all the subsiduary
> tools and they will be (and indeed are) confused if that isn't the
> case.

Please refrain from using "a minority" or "the majority" if you can't
point to real data about that.

But there's some real data: popcon. 264545 Ubuntu users have libruby1.8
installed (ie about >50% of all ubuntu users, since the most installed
packages have 574294). And 9007 users (1.5% of all Ubuntu users, 3.4% of
users having libruby1.8 installed) have libgems-ruby1.8 installed.

So the data is pretty clear: most users of ruby in Ubuntu are interested
in running ruby apps, not in developing ruby apps. (compare with
rdoc1.8: 16493 insts)

And the package would still be listed if the user only installed
rubygems to do a 'gem update --system', so there's no bias introduced by
that.

> > It already exists, it's called ruby-full:
> > Package: ruby-full
> > [..]
> > Depends: irb, libdbm-ruby, libgdbm-ruby, libopenssl-ruby, libreadline-ruby, rdoc, ri, ruby, ruby1.8-dev
> > Recommends: libtcltk-ruby, ruby-elisp
> >
> > But currently, it pulls ruby1.8, since that's the default version, not
> > ruby1.9.
>
> Yes, but that's the problem - one of marketing. Names matter. People
> expect ruby1.x to provide what they need to use ruby. They don't
> expect it to be just core ruby. If packagers need smaller packages
> then they should be the one handling the 'what is the package called'
> problem, not the poor end user who just wants to develop in ruby.
>
> For me ruby1.x should be user facing, pull in everything required to
> develop in ruby (including gems which is notably missing from your
> list above). We should have ruby1.x-core for those packagers who need
> the lighter less fattening alternative (with ruby1.x-core handling the
> 'alternatives' system properly and providing the 'ruby' symlink).

That's your opinion.

> The 'ruby' package needs to be a virtual that pulls the 'current'
> version, but is satisfied by all the other options (which will include
> jruby and rubinus before too long).

That's hard to do, unfortunately, because of compiled extensions. (you
can't just install jruby and expect every ruby app to work)

> So it is really just a package renaming issue, which is a bit of a
> pain in the backside, but it will improve the standing of
> Debian/Ubuntu ruby in the community enormously if it is done. And it
> sets us up to handle the various interpreter combinations going
> forward.

Such a renaming is not an option for lenny. We can discuss it again for
lenny+1.
--
| Lucas Nussbaum
| <email address hidden> http://www.lucas-nussbaum.net/ |
| jabber: <email address hidden> GPG: 1024D/023B3F4F |