Sitefinder II, the sequel...

Patrick W. Gilmore patrick at
Thu Jul 13 12:08:01 UTC 2006

On Jul 13, 2006, at 3:39 AM, Simon Waters wrote:

> Most of those I know try to deploy recursive services as close as  
> possible to
> the client, avoiding where possible alternative views of the DNS, and
> forwarding.

Would that everyone did what the people you know do.

Unfortunately, there are a few providers doing things like  
outsourcing their recursive service to, say, their upstream, or  
having one "node" of recursive servers anywhere in the world for all  
their end users.  These providers violate the first part of your  

The second part doesn't make any sense to me.  It seems that having  
multiple, geographically disparate recursive name servers would be  
more likely to present an "alternative [view] of the DNS".  (In fact,  
I can prove that's true in at least some cases. :)  So you are  
actually arguing -against- your first point.

That said, no one has yet said why it is necessary, or even  
desirable, to have a completely homogenous view of the world.

> Perhaps time to ask Brad, Paul and Cricket what they think, and  
> have answers
> to their comments.

Perhaps.  However, in the last DNS related thread, Paul made a pretty  
strong claim (violating a protocol) and showed exactly _ZERO_ facts  
to back it up, despite being asked at least five times (by my count).

> With automated responses to "bad things", it is usually best to  
> minimise the
> scope of the change. Similarly typo correction makes sense for  
> URLs, but not
> for most other uses of the DNS (hence the proviso you make to  
> switch it off
> if you use RBL, although I'd say switch it off for all email  
> servers less you
> start correcting spambot crud, our email servers make a DNS check  
> on the
> senders domain, that doesn't want correcting either), so the answer is
> probably browser plug-in (although most browsers already try to  
> guess what
> you meant to some extent).

Perhaps something as simple as a preference only 'correcting' queries  
that begin with "www"?


More information about the NANOG mailing list