What's Wrong with CNAMEs? (was Re: Prefixless domains)

Kevin Darcy kcd at daimlerchrysler.com
Sat Mar 23 00:26:13 UTC 2002


Will Yardley wrote:

> In article <a7ajfo$kan at pub3.rc.vix.com>, Barry Margolin wrote:
>
> > Since my recommendation is "Use CNAMEs unless they're prohibited by
> > the protocol", I might agree with that.  The test should be on when
> > CNAMEs are not allowed.
>
> well the difference is that you know how / when to use them properly.
>
> personally i avoid them *usually*, but i think it's probably good to
> encourage people not to use them in general. until they have a situation
> that actually calls for a CNAME, or have an understanding of the
> implications of using one.
>
> errors involving CNAMEs are very common, so while it's well and good for
> you to use them (where appropriate), i personally think it's best to
> suggest that people avoid them.

I have a question for all of you snobs who avoid CNAMEs like the plague:
how do you deal with reverse records when multiple names must resolve to
the same IP address? Multiple PTR records don't work, of course. What
reasonable alternative is there, besides making one of the names be an
A record, with the other (n-1) names being aliases to that name? Can't do
that with zone-apex names, of course...

I have no problem with CNAME use. The current versions of BIND forbid some
of the most egregious CNAME abuses; the maintenance system that my users
use stops even more abuses and/or questionable CNAME practices. So what's
so bad about them? We use CNAMEs extensively in our internal DNS, including
some zones that contain CNAMEs *exclusively*. Works fine for us.


- Kevin





More information about the bind-users mailing list