It has always worked like that and rightfully so imo; by storing the entire line in $1, multiple spaces are preserved (they wouldn't be if the line was tokenized to $1-).