>  1) How does Athena/MIT do this?  Or anyone else for that matter?

The Athena hesiod distribution has used the "uid" map since the
beginning of time, according to the RCS logs.  I didn't know about
this incompatibility with Ultrix until now.

>  2) If Athena/MIT (or anyone else) do it differently, where can we
>     specify which map to look up user id's in?  /etc/hesiod.conf would
>     seem to be the obvious place, but it's gen/getpwent.c that needs
>     to know the map and not the general hesiod routines.  Having
>     gen/getpwent.c also parsing /etc/hesiod.conf seems ... tacky.

There's always the runtime approach of looking for <uid>.uid and then
<uid>.passwd if that fails with ENOENT.  But of course it sucks not
being able to inform the library that only one lookup will be
necessary, like you can do with the IN/HS screw.