Scott Rotondo wrote: > On the other hand, I really do think this could be considered a detail > of the implementation. Under my proposal, sulogin will still prompt for > a username and password. Even if the spec is silent on the issue, the > implementation has to have *some* deterministic behavior when the > username is a null string; I'm just suggesting what that behavior should > be.
But that behaviour should be documented otherwise we are free to change it and that means people can't depend on it (but probably would anyway). If they can't depend on it why bother with it implying root rather than just giving an error message and asking for the username again. This might seem like nits but it is stuff like this that customers actually notice and get annoyed about when we change it. I know this from the praise and flak we got for changing the exact dialog that passwd(1) output (and passwd wasn't actually changed but PAM modules). -- Darren J Moffat
