Russ Cox has some nice examples of the issues with this: Otherwise the meaning of &f().x is different for f() returning pointer-to-struct and f() returning

Russ Cox has some nice examples of the issues with this: Otherwise the meaning of &f().x is different for f() returning pointer-to-struct and f() returning struct. Similarly &m["x"] is a compile error today but would silently make a copy tomorrow rather than produce a pointer to the value in a map. All of that would be incredibly confusing and the source of many subtle bugs.