Favorite DNS?

Michael L Torrie torriem at chem.byu.edu
Fri Jul 27 13:55:16 MDT 2007

Jonathan Duncan wrote:
> To quote DJB:
> Don't use [CNAME] if there are any other records for [the fully  
> qualified domain name]. Don't use [CNAME] for common aliases; use  
> [the fully qualified domain name] instead. Remember the wise words of  
> Inigo Montoya: ``You keep using CNAME records. I do not think they  
> mean what you think they mean.''  (http://cr.yp.to/djbdns/tinydns- 
> data.html)

Good advice I think.  By definition there can be only one canonical name
for an IP address.

There are exceptions to this rule of thumb, of course.  For example, I
have my www and ns1 names both being A records for my IP address (same
ip address).  Hence I have two FQDNs for one IP address.  This is
because usually you don't want ns1 (the name server) to be a CNAME; also
I might move it to another IP address later.  Everything else is an
aliase for www (www is the CNAME of everything else).  This violates
Dan's rules, but for good reason (or maybe not).

But generally, if you're ever wondering if you should do an A record or
a CNAME record, you should probably do a CNAME record (an alias).

> The above record would be written like this in TinyDNS:
> +athena.domain.com:
> Not to start a war, but doesn't that look much easier?

No, not really.  It's more concise, in terms of characters, but it's not
more clear to the end user.  It still requires that you understand how
to parse the line.  So it's certainly not worse.  But not better either.
 It's also not in standard BIND format, which is often used to read
zones into other DNS programs (IE everyone should understand BIND zone
format; not everyone will understand DJB zone format).

What bugs me about djbdns is that he bills it as a BIND replacement.  If
it was a BIND replacement, it should be able to natively read BIND zone
files.  Oh well.

> Jonathan
> /*
> PLUG: http://plug.org, #utah on irc.freenode.net
> Unsubscribe: http://plug.org/mailman/options/plug
> Don't fear the penguin.
> */

More information about the PLUG mailing list