* Adam Kennedy <[EMAIL PROTECTED]> [2007-04-20 18:30]:
> Module::Build's main attitude has always been to make things
> easier for authors at the expense of users, in part by
> implementing things that the rest of the tool-chain simply
> doesn't support.

And Module::Install makes things easy for users at the expense of
authors. Imagine the churn caused my new M::I releases if half
the CPAN was using M::I and all authors were active all the time.
Imagine the problems if half the CPAN was using M::I and a
serious bug was found in it while many authors, as is the case in
reality, are intermittently AWOL.

That doesn’t scale either.

As far as I’m concerned, `configure_requires` is the solution
that should have been implemented IN THE FIRST PLACE.

Not M::B and not M::I.

It isn’t just that “well we had to do `configure_requires`
because M::B is too widespread now and M::B needs it” – as if the
rest of the toolchain didn’t have any real issues. It’s the other
way around – the most serious issue in the toolchain was that
there is no tenable way forward in the absence of a mechanism
like `configure_requires`.

(Personally I would have favoured a solution wherein the Build.PL
or Makefile.PL can signal to the installer that the process needs
to start over after installing modules X, Y and Z, as that would
provide actual Turing completeness.)

*After* fixing that, we could have let a thousand attempts to fix
EU::MM’s shortcomings blossom. *Without* overly complicating the
installation process for users.

> This STILL doesn't fix the problem with current or old default
> Perl installations without upgraded CPAN clients, but then you
> REALLY can't fix a circular dependency that only occurs due to
> events that occured in the past.

That should be possible to fix, shouldn’t it? What I’m thinking
is Module::Build::Installer which implements just enough of
CPAN.pm to recursively download, install and run Module::Build
using the existing CPAN::Config, so that you could depend on
Module::Build::Installer in a sort-of passthru Makefile.PL.

Something like that; not sure of the specifics. In any case it
seems eminently solvable to me.

Regards,
-- 
Aristotle Pagaltzis // <http://plasmasturm.org/>

Reply via email to