Re: Last Call: Root Name Server Operational Requirements to BCP

[Robert Elz <kre@munnari.OZ.AU>]

    Date:        Mon, 28 Feb 2000 10:49:35 -0800
    From:        Randy Bush <randy@psg.com>
    Message-ID:  <E12PVEd-000Gpl-00@rip.psg.com>

  | i suspect that was the intent.  why respond to garbage?

If you don't respond, you, or some other root server, is just going to get the
query again, just the same (most of the time).

There are two situations - either the packet has been mangled somehow,
and the lack of the checksum means you can't detect that, or the packet
is correct but just comes from someone who doesn't bother to checksum
udp packets.   In the former case, the reply is likely wasted (even then
not necessarily, the corruption might be to some part that would be irrelevant
to a reply from a root server) - there will need to be a retransmit anyway,
and a new reply.   Fortunately, that shouldn't happen often (packets that
udp checksums catch, which link level checksums haven't caught, are fairly
rare).   In the latter case, the reply will be just fine, and communications
continue, and the root server stops being bothered by this query.

  | perls before swine?

I doubt that the root servers are the proper place to evangalise for the
use of udp checksums...

kre