bug#3619: Replace Rectangle can't do empty after first replace anymore

[Stefan Monnier]

>> Actually, C-x r d is the equivalent (since C-x r k additionally stores
>> the killed text on the kill ring).  And yes, this was a known tradeoff
>> when introducing the new behavior in Emacs-22.
> I looked briefly at this, and found it surprising that `read-string'
> offers no way to reject the default value in favour of an empty
> string. `read-from-minibuffer', on the other hand, does.

I don't know what you're referring to: in which way does
read-from-minibuffer let you distinguish between "the user wanted the
empty string" and "the user wanted the default value"?


        Stefan