Re: CVS commit: src/bin/hostname

[David Holland <dholland-tech%netbsd.org@localhost>]

On Tue, Jul 30, 2013 at 01:28:04AM -0400, Mouse wrote:
 > > If my laptop moves back and forth between multiple sites, [...] does
 > > its "real" hostname change every time it moves?
 > 
 > > Ultimately the "real" hostname [...] is how you refer to the computer
 > > and distinguish it from others you deal with.
 > 
 > That's one reasonable definition of a machine's hostname: it's a
 > human-layer name for the machine.
 > 
 > But it's far from the only reasonable definition.  Another one is "some
 > name, resolvable, with respect to the context the name is used in, to
 > some nonempty set of addresses at which the machine can be reached
 > from within that context" - ie, loosely, a name at which it can be
 > reached.

That one basically doesn't work for mobile hosts; that's part of the
problem I was citing. For a mobile host you can in practice either
have a name that can currently be reached, or a permanent name, but
not both.

Most mobile hosts do not have a fixed IP even in their native
environment, because fixed IPs are expensive and there's no need for
it. So even in a particular context, there isn't necessarily a
resolvable way to reach the host.

(You can set up dynamic DNS to work around this problem but again,
most of the time people don't because there's typically no real
point.)

 > Which one you prefer is, of course, likely to affect your stance on
 > some, possibly all, of these hostname(1) flags.

Not in my case -- having hostname(1) do DNS lookups is problematic and
potentially insecure and working around these issues via /etc/hosts is
brittle. So I don't think it should be done. You can always run
nslookup(1) on what hostname(1) returns.

(What would be useful is a standard tool for doing DNS lookups in the
shell that produced easily parsed output. But I digress...)

-- 
David A. Holland
dholland%netbsd.org@localhost