Help! Configuring BIND 8.22 on domain-less system

[Kevin Darcy]

Caching.
    As I understand your setup, your clients are using a Linux server for
DNS lookups, and the Linux server is configured to forward to your ISP's
nameservers. If this is the case, then you should *already* be caching
lookups for performance. What makes you think you aren't? Dump your
database -- with "ndc dumpdb" or an INT signal to the "named" process, the
default file for the dump is "named_dump.db", found in "named"'s working
directly -- sometime: that'll show you what your nameserver has cached.

Private Namespace
    You can achieve this easily by just defining a master zone on your
server. In the absence of your own *real* domain name, you could pick some
"impossible" TLD like ".internal". E.g. "potato.internal",
"onion.internal". Then just set that as your default domain on the clients
so that they can use short names. You can forbid external clients from
querying those names by putting an "allow-query" restriction on the zone
(in fact, for security reasons, it might be a good idea to restrict queries
to just your internal clients *globally* until and unless you start hosting
a domain to the outside).

Getting Your Own Domain
    Should you or shouldn't you? Depends. Are you going to be providing
services to external entities? If so, then you certainly want them to use a
domain name to access your server. But the fact that you don't have a
static IP is a bit of a showstopper -- it means that you can't actually
host the domain on your server. You'd have to get someone else to host the
domain. Also, since your server's address is dynamic, you would want to get
hooked up with something like dyndns.org , tzo.org (I've never used them;
only heard about them second-hand, so don't take that as an endorsement) or
some other service that allows you to dynamically update a DNS database as
your address changes. You could have, say, for a web server,
www.example.com (your domain name) be an alias to {whatever}.dyndns.org,
and then dynamically update {whatever}.dyndns.org when your address changes
-- this will mean that, notwithstanding possible caching issues, people can
always access your server via the same, "friendly" name. I don't know for
sure if anyone is offering an integrated Dynamic DNS and domain hosting
package -- I'd have to believe that there is plenty of demand for such a
thing.

One other benefit of having your own domain name, even if you're not
offering services directly on the Internet, is that you can use that for
your private namespace instead of something bogus, without fear of ever
colliding with anyone else's domain name. This becomes important if you
want to build a robust email infrastructure, for instance, where global
uniqueness of names is important.


- Kevin


Robert La Ferla wrote:

> X-no-archive: yes
>
> Dear DNS Gurus,
>
> I need some help configuring bind 8.22.  I have a Linux server that
> connects to the Internet via a cable modem.  The server (onion) receives
> it's IP address via DHCP.  This server acts as a router/gateway for
> Windows 98 clients (potato and carrot).  The clients use the gateway for
> DNS.  The server uses the ISP's DNS servers in turn.  This all works.
> What I can't figure out is how to add local hostnames (visible only to
> the LAN)  For example, I'd like the hosts "potato" and "carrot" to be
> able to access "onion" by name.  They currently have to have this set in
> a hosts text file.  I want this to happen via DNS.  Lastly, I'd like the
> DNS server to cache previous host lookups for performance.  The other
> thing I can't figure out is whether I really need a domain name or
> not.   If I do, which one should I use?  I have registered domain names
> at my disposal but the gateway doesn't have a static IP address.  Should
> I use the ISP's domain?  Use a fake domain?  Does it matter?  Help!
>
> Robert