For the last 16 years, it seems that nobody complained about Tie::StdHash no longer inheriting the useless new() from Tie::Hash. So I am very sure that we can also move that class safely out of Tie/Hash.pm.
Having a fresh look at the code of both Tie::StdHash and Tie::Hash explains why nobody complained:
- Tie::Hash::TIEHASH() is designed to call new() in an inheriting class, or to be reimplemented by the inheriting class. If the inheriting class implements neither new() nor TIEHASH(), Tie::Hash::TIEHASH() croak()s.
- (If the inheriting class implements new(), but not TIEHASH(), Tie::Hash::TIEHASH() warns that it calls new() because TIEHASH() is missing. So why even bother to implement new()?)
- Tie::Hash::new(), which is never called by tie, just calls TIEHASH() of the inheriting class. Implementing new() in an inheriting class does not make any sense at all, because you need a class name, not an object, for tie.
- Tie::StdHash implements its own TIEHASH(), as expected by Tie::Hash and tie. This prevents new() from being called indirectly from tie().
So, no class inheriting from Tie::StdHash needs a new() implementation.
Today I will gladly share my knowledge and experience, for there are no sweeter words than "I told you so". ;-)