[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]


To: Matt Larson <mlarson@verisign.com>
Cc: Michael Richardson <mcr@sandelman.ottawa.on.ca>, dnssec@cafax.se
From: Miek Gieben <miekg@atoom.net>
Date: Wed, 12 May 2004 12:03:40 +0200
Content-Disposition: inline
In-Reply-To: <20040511154445.GR31336@chinook.corppc.vrsn.com>
Mail-Followup-To: Matt Larson <mlarson@verisign.com>,Michael Richardson <mcr@sandelman.ottawa.on.ca>, dnssec@cafax.se
Sender: owner-dnssec@cafax.se
User-Agent: Vim/Mutt/Linux
Subject: Re: dnssec: resolver - application communication

[On 11 May, @17:44, Matt wrote in "Re: dnssec: resolver - applica ..."]
> have been in the "AD bit, SERVFAIL sucks" camp for some time and a
> voice crying for something else.  I think some additional RCODEs (or

I tend to agree that having only SERVFAIL to signal "something" is not
enough. But I also want to ask the following: aren't we optimizing the
least used code-path?

How often does one get SERVFAIL of a respected site (known to have
it's DNS in order)? I don't get it that often. Why can a secure aware
application/resolver that receives a SERVFAIL not _always_ follow up
with a requery + CD=1? 
And what do we expect the lecacy application/resolver to do when it
get a NOTVAL error code?

So two options can be pursued:

1)
Either we add a NOTVAL error code to signal that "yes, we've got the
data, and no it didn't verify". So that the thing at the other end
might requery with CD=1 if it requires so.

2)
Or we say SERVFAIL is enough, and if you get SERVFAIL the thing at the
other end might requery with CD=1 if it requires so.

regards,
Miek

Home | Date list | Subject list