draft-manning-dnssvr-criteria-00.txt: 2 additions, 1 question

Aaron Leonard <AARON@tgv.com> Sat, 13 April 1996 18:40 UTC

Received: from ietf.cnri.reston.va.us by IETF.CNRI.Reston.VA.US id aa14377; 13 Apr 96 14:40 EDT
Received: from CNRI.Reston.VA.US by IETF.CNRI.Reston.VA.US id aa14373; 13 Apr 96 14:40 EDT
Received: from ietf.cnri.reston.va.us by CNRI.Reston.VA.US id aa09962; 13 Apr 96 14:40 EDT
Received: from ietf.cnri.reston.va.us by IETF.CNRI.Reston.VA.US id aa14344; 13 Apr 96 14:40 EDT
Received: from CNRI.Reston.VA.US by IETF.CNRI.Reston.VA.US id aa14267; 13 Apr 96 14:30 EDT
Received: from Dr-Crippen.TGV.COM by CNRI.Reston.VA.US id aa09851; 13 Apr 96 14:30 EDT
Received: from TGV.COM by TGV.COM (PMDF V5.0-6 #16141) id <01I3EPYH6HYOHSJRDS@TGV.COM>; Sat, 13 Apr 1996 11:30:30 -0700 (PDT)
Date: Sat, 13 Apr 1996 11:11:56 -0700
X-Orig-Sender: ietf-request@IETF.CNRI.Reston.VA.US
Sender: ietf-archive-request@IETF.CNRI.Reston.VA.US
From: Aaron Leonard <AARON@tgv.com>
Subject: draft-manning-dnssvr-criteria-00.txt: 2 additions, 1 question
To: ietf@CNRI.Reston.VA.US
Cc: aaron@cisco.com, bmanning@isi.edu
Reply-to: Aaron@cisco.com
Message-id: <01I3HP94Z3F6HSJRDS@TGV.COM>
MIME-version: 1.0
Content-transfer-encoding: 7bit
Comments: TGV Software Inc. Technical Support
Source-Info: From (or Sender) name not authenticated.

A proposed addtion to "Technical Criteria for Root and TLD Servers":

  UDP and TCP TTL >= 64
        DNS response packets must be emitted with IP TTLs greater
        than or equal to 64, to ensure that they can reach all
        regions of the Internet.

  Comment: the last time I checked, one of the root nameservers
  was still sending out UDP responses with a TTL of 32 or so,
  rendering it useless to some of the hinterlands.

Another proposed addition:

  RP RR for each ?.ROOT-SERVERS.NET
        Each host in ?.ROOT-SERVERS.NET shall have an RP
        ("Responsible Person") resource record entered for
        it in the DNS, with a valid mailbox name.

  Comment: what with the generic "?.ROOT-SERVERS.NET", and
  with the new proposed requirement that all PTR records
  for root servers return the "?.ROOT-SERVERS.NET" name,
  it's pretty unobvious who exactly is really responsible
  for a given root nameserver.

A question:

> 4. Singly homed (only one interface).

Uh, why?  Modern BINDs on multihomed hosts manage correctly
to emit UDP responses with the correct source IP address.
For reliability's sake, it may well make sense for a given
root server to have multiple IP addresses in separate routing
domains, would it not?

If there is some technical reason why a given NS should not
have multiple A records, then would it be adequate for the
box itself to be multihomed, but for only one of its 
interface addresses to be mapped to the NS name?

Cheers,

Aaron

Aaron Leonard			aaron@cisco.com / aaron@tgv.com
cisco MultiNet for OpenVMS Support