RE: IPv6 automatic reverse DNS

Mailman List Mirror (Read Only)
1

There are two competing drafts for synthetic rule-based PTR responses
for IPv6 rDNS:

Howard Lee, Time Warner Cable (now Charter)
draft-howard-isp-ip6rdns-08

J. Woodworth, CenturyLink
draft-woodworth-bulk-rr-09 - BULK DNS Resource Records

Nominum and Xerocole/Akamai also have proprietary solutions to this
in their Vantio AuthServ and AuthX products, respectively.

It seems to me that it is still an open question whether the
recommendations in RFC-1912 that any IP address that accesses the
Internet should have a PTR and matching forward record. My personal
thoughts are that the best solution would be an OPTIONAL standards-based
method of generating DNS responses based on a ruleset if a specific zone
record is not present, and that implementation of that requirement
should be left to the developers of the auth nameserver software.

Greetings Andrew,

I am new to the group but one of the authors referenced above. My
colleagues and I are glad to see the discussion around this issue
see some recent movement.

As indicated by one of our esteemed WG chairs elsewhere in this thread,
I am currently working to provide additional clarity for some of the
more difficult concepts in the draft and have not yet requested the
next step. Once these changes are complete we will enthusiastically
move forward with this request.

As I am new to this forum, for the moment I wanted to simply state:
synthesized records based on the proposed "bulk rr" method can
_only_exist_where_zone_records_do_not_already_. One critical goal of
the draft is to make the "intent" of synthesized records easy to
transfer between nameservers in authoritative roles. Examples for
implementing the draft using fairly straightforward regex
manipulation are included but are more of a guideline for making
the pattern substitution easier for the implementor and provide
a reference for the accompanying examples. Ultimately, as you
recommend, the auth nameserver software vendor would be free to
provide their own pattern substitution logic (so long as the
intent is not lost).

DNSSEC for synthesized records also poses its own obvious set of…
complications for which we've outlined a number of solutions to
help satisfy this challenge.

Admittedly, it is a bit of a hefty read but we would love the
feedback (directly or in the IETF DNSOP mailing list of course).

Thanks,
John Woodworth

Andrew

Caveat: These thoughts are mine personally and do not represent
any official position of Charter Communications.

Ληdrеw Whiте
Charter Network Operations - DAS DNS
Desk: 314-394-9594 ? Cell: 314-452-4386
andrew.white2@charter.com

-- THESE ARE THE DROIDS TO WHOM I REFER:
This communication is the property of CenturyLink and may contain confidential or privileged information. Unauthorized use of this communication is strictly prohibited and may be unlawful. If you have received this communication in error, please immediately notify the sender by reply e-mail and destroy all copies of the communication and any attachments.

2

Hi John,

Thanks for the info and background.

One operational suggestion I have is … why link synthesis rules to a specific DNS zone?

Most larger operators of auth DNS use an IP management tool, like BT Diamond IPAM, BlueCat, or Infoblox. Oftentimes, allocations of IP space will not be on classful boundaries, yet most often reverse DNS zones are on classful boundaries.

What may be more operationally useful would be an (optional) feature in auth DNS software that would process an incoming PTR request as follows:

1. Answer the PTR with an entry in the corresponding ip6.arpa or in-addr.arpa zone file if the PTR exists

2. Otherwise, examine a rule set of synthetic PTR responses and answer by the rule set (e.g. 10.0.0.128 matches rule for “10.0.0.128/27” and returns PTR of 10-0-0-128.dhcp.example.com.)

3. Otherwise, return NXDOMAIN or NOANSWER/NOERROR as appropriate

Such a ruleset could apply to forward zones as well to create the matching forward lookup.

Just my two cents! Caveat: personal opinion and not the official position of Charter.

Andrew

Ληdrеw Whiте
Charter Network Operations - DAS DNS
Desk: 314-394-9594 - Cell: 314-452-4386
andrew.white2@charter.com<mailto:andrew.white2@charter.com>

3

Hi John,

Thanks for the info and background.

One operational suggestion I have is … why link synthesis rules to a
specific DNS zone?

Most larger operators of auth DNS use an IP management tool, like BT
Diamond IPAM, BlueCat, or Infoblox. Oftentimes, allocations of IP space
will not be on classful boundaries, yet most often reverse DNS zones
are on classful boundaries.

What may be more operationally useful would be an (optional) feature
in auth DNS software that would process an incoming PTR request as
follows:

1. Answer the PTR with an entry in the corresponding ip6.arpa
            or in-addr.arpa zone file if the PTR exists
2. Otherwise, examine a rule set of synthetic PTR responses and
            answer by the rule set (e.g. 10.0.0.128 matches rule for
            “10.0.0.128/27” and returns PTR of 10-0-0-128.dhcp.example.com.)
3. Otherwise, return NXDOMAIN or NOANSWER/NOERROR as appropriate

Such a ruleset could apply to forward zones as well to create the
matching forward lookup.
Just my two cents! Caveat: personal opinion and not the official
position of Charter.

Andrew,

Excellent question. Out of necessity we have an in-house federated
solution for DNS/DHCP/IP/etc. which solves part of the problem.
However, not all data can be managed this way; some more tech-savvy
customers expect to manage their own data and transfer it directly
to our nameservers for the higher availability, lower latency,
tighter security, etc. This then becomes a shared burden at the
zone level where, from our perspective, the intent should be easily
transferable. I suspect if/when the draft is adopted, other IP
management tools may offer the capability of automatically
generating the associated "BULK" resource records for the various
DNS zones allowing for better interoperability (i.e. "transferability").

One of the draft's features I am most proud of is the concept of
superimposed records. This can scale to really huge levels where
for example: the RIR could provide patterns for all unclaimed records
under "10.in-addr.arpa." which could be overridden by more specific
patterns for records under "255.0.10.in-addr.arpa." The DNS ownership
now follows the intent of the expected DNS zone owner. If one
follows this logic through the ipv6 tree, this concept of ownership
becomes even more pronounced.

I guess in short, the answer is to maintain the concept of zone
ownership :slight_smile:

Thanks,
John Woodworth

Andrew

Ληdrеw Whiте
Charter Network Operations - DAS DNS
Desk: 314-394-9594 - Cell: 314-452-4386
andrew.white2@charter.com

-- THESE ARE THE DROIDS TO WHOM I REFER:
This communication is the property of CenturyLink and may contain confidential or privileged information. Unauthorized use of this communication is strictly prohibited and may be unlawful. If you have received this communication in error, please immediately notify the sender by reply e-mail and destroy all copies of the communication and any attachments.