Not to detract from your excellent point, parent allows for (and documents) this case:

package MyHash;
use Tie::Hash;
use parent -norequire, 'Tie::StdHash';
[download]

I understand that use parent -norequire, 'Some::Class'; can be helpful, but it requires even more typing than our @ISA=('Some::Class');. I've considered for a few seconds if parent could fix what Tie::Hash broke, without forcing me to write more code.

Adding a few special cases for core modules like Tie::StdHash that exist in files where they should not be. A simple hash mapping broken classes to their containing files ({ 'Tie::StdHash' => 'Tie::Hash', 'Tie::ExtraHash' => 'Tie::Hash', ... }), plus a few lines of code to use that hash. Perhaps only if require failed to load the base class.

It would have solved the problem that use parent 'Tie::StdHash'; does not work out of the box. But the same would have to be added to base, and to every other module that manipulates @ISA for other modules. And it would require updating the workaround hash from time to time, whenever someone blindly copies the Tie::Hash anti-pattern.

Another way could have been adding a way to load a class and inherit from a different class to parent, eleminating the need to manually load the file containing the base class. Something like use parent 'Tie::StdHash', -from => 'Tie::Hash'; or use parent { 'Tie::StdHash' => 'Tie::Hash' };.

But again, why should parent fix what is wrong in other modules?

After all, parent's history is quite clear:

This module was forked from base to remove the cruft that had accumulated in it.

So, after about three seconds, I discarded that idea and did not write about it in the meditation.

Alexander