Subject: [resolv.conf] domain vs. search
To: None <current-users@NetBSD.ORG>
From: VaX#n8 <>
List: current-users
Date: 12/19/1994 21:30:42
I found the documentation on resolver somewhat confusing.  The resolver(3)
manpage suggests that the domainname is broken into component domains, and
each successive parent domain is tried until you reach the last one (.EDU or
something similar).  However, in the resolv.conf(5) manpage, it appears that
you can set the domain name (in case your hostname is deceptive or not set
yet) to whatever you want.

In my case, I set my hostname to and the domain
name in /etc/resolv.conf to the idea being that I could
"telnet ccwf", for example, and get "".  This worked fine,
but "telnet dimebox.cs" did not get me "" - that is,
it was not trying each successive parent domain, only the one listed.

I eventually figured out that I had to set the search path to the parent
domains I wanted searched.  It appears that the "search" keyword is not
only a superset of "domain", but would entirely replace it, given the same

So, my question is; is there anything wrong with either (1) modifying the
domain keyword to give the intuitive usage (e.g. it simply provides the domain
name, instead of calling hostname, and is otherwise the same algorithm) or
(2) getting rid of it entirely in favor of search (marking it as deprecated,
but perhaps retaining it for a while for sdrowkcab compatibility)

One more question: the distinction between bind/resolver and other things
is kind of blurry to me.  Where can I get a description of how names are
converted into IP addresses in a monolithic or sequential document?
The manpage for bind(2) doesn't indicate anything related to resolv.conf's
bind keyword or named for that matter.

PS: did anyone else know "man -a 7 \*" works to list the whole section 7?
VaX#n8 (vak-sa-nate) - n, CS senior++ and Unix junkie -
Just the vax-man.  Read my MIPS, no new VAXes!        - PGP key on request