You cut off the quote at an unfortunate place. I see plenty of reasons for scalars to distinguish between empty (string) and undefined. What I see no logical reason for is that scalars make that distinction and arrays don't. It makes obvious sense to either have undef for both or have undef for neither, but what's the reasoning behind only doing it for one and not the other?

Anyhow, as you said, you can definitely work around it using references, no problem there. The way Perl implements it is perfectly usable. It just seems strange to me that scalars have a special "missing value" value, but arrays don't.