The short answer: for consistency. I mean, if a user chooses to write wide logs, he probably knows that national characters may appear in log messages and attributes. Attribute names may also be put into log and, in fact, may be composed by the application and therefore may contain national characters. Another reason for this decision is the library interface unification. For instance, in many cases users won't have to remember to write either attr or wattr, but to always write attr, whatever character type they are using. This also helps in generic programming.