>>>>> On Thu, 3 Jul 2003, Jeremy C. Reed wrote:

>> Can you tell me more (or point me to some references) about it being
>> harmful if domainname is set when NIS is not used?

Some (old?) implementations of NIS open a TCP socket to access ypbind
at every call of name-service-related functions (e.g. getpwent(),
gethostbyname()), even if the access failed every time due to absence
of ypbind and YP servers. This takes significant cost especially on
heavy loaded servers.
So, I don't think it's good thing to set `domainname' in an
environment without NIS in general.

Our NIS library doesn't take such significant cost, because our libc
checks existence of /var/run/ypbind.lock at first. But I still don't
think it's good thing to set `domainname' by default installation,
because it may confuse users about the meaning of the `domainname',
and it has slight, but certain performance penalty.

If one uses convention like `domainname` == DNS domain in one's
environment, it may be fine. But I don't think it should be standard
on NetBSD.
--
soda