The issue is basically, as you'll see in my comments on the updated
bug, Perl virtuals are not very trivial, and every perl release
results in a nightmare of fixes we have to apply to every Perl virtual
in tree, so we don't , at present, want more virtuals than necessary.

I'd personally love to have a virtual for each thing that is in perl
itself, as a virtual, because that would absolve me from a few
headaches with depending on certain core modules, but I won't request
it because the added complexity is just not really worth it.

And its not made easier by the fact upstream declare dependencies on
Modules, not Distributions , and Module<=>Distribution is resolved
dynamically at install time, and we basically have to make assumptions
downstream that modules wont jump around too much between releases .

If we were to have a virtual for each *module* that was possible to
depend on, there would be ... 668 virtuals, *just* for perl.

I volunteer to be somewhere else if that ever happens. Not maintaining
that, no thanks =)



-- 
Kent

Reply via email to