[dns-operations] Odd EDNS implementation bug

[Geoffrey Sisson]

Mark Andrews <marka at isc.org> wrote:

> In message <20100526232957.5B0C25406E at smtp.geoff.co.uk>, "Geoffrey Sisson" writ
> es:
> > marka at isc.org (Mark Andrews) wrote:
> > 
> > > In message <alpine.LSU.2.00.1005261924540.28459 at hermes-2.csi.cam.ac.uk>, To
> > ny Finch writes:
> > > > Some rr.com name servers put the answer RRset in the additional ssection
> > > > if you make an EDNS request:
> [snip]
> > > I suspect that it is a specialized server synthesizing responses
> > > on the fly and it hasn't been properly tested.  You can't just
> > > append answers to queries and have it work.
> > 
> > Looking at this on the wire, it seems the OPT RR is in the answer section
> > and the answer is in the additional section, so it looks like they just
> > have the RRs in the wrong order.
> > 
> > Geoff
>
> If you play around with the edns buffer size you will see it change
> in the response to match that of the query.

Yup, that seems to confirm your suspicion that they're simply appending.
I thought they might just be serializing the RRs incorrectly.

Geoff