[bind10-dev] Non-absolute names for $ORIGIN

Michal 'vorner' Vaner michal.vaner at nic.cz
Mon Dec 17 10:46:11 UTC 2012


Hello

First of, the branch is not merged yet. So the discussion can change the
behaviour.

On Mon, Dec 17, 2012 at 10:33:32AM +0000, Stephen Morris wrote:
> Firstly if $ORIGIN is a relative name, then what is it relative to?

To the previous origin (either being origin of the name or from previous
$ORIGIN). Just like x = x + 1 in C.

> Secondly, the omission of a trailing "." in a $ORIGIN statement caused
> a ccTLD to go off-line a few years ago. Perhaps we could make
> acceptance of relative names for $ORIGIN a BIND 10 configuration
> option?  Some users may welcome the ability to catch a common error.

I don't know if the argument „Someone screwed up some years ago, so we'll
be afraid and have a workaround for eternity in around“ is very good approach ‒
that's why I always hate to see things like „if [ x$SOMETHING = xsomething ]“.

But, indeed, I the use of relative names in $ORIGIN may be questionable, just
because the semantics are not really clear. The current behaviour is what I've
seen in bind9 code ‒ but I think it allows them more due to code reuse than as a
feature.

So, we could do three things:
 * Consider the name absolute, even if it doesn't end with dot.
 * Fail if we see a name that does not end with dot.
 * Warn if we see a name that does not end with dot (I'd personally prefer this,
   if we consider the issue important enough to care, as it preserves backward
   compatibility, but it still tells the admin something may be wrong and how to
   fix it).

With regards

-- 
  I'm a dragon, and I will not be ignored, not even by a tree!
       Saphira Brightscales

Michal 'vorner' Vaner
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 198 bytes
Desc: Digital signature
URL: <https://lists.isc.org/pipermail/bind10-dev/attachments/20121217/2cb0e778/attachment.bin>


More information about the bind10-dev mailing list