DHC Working Group M. Stapp
Internet-Draft Cisco Systems, Inc.
Expires: August 31, 2001 March 2, 2001
Resolution of DNS Name Conflicts Among DHCP Clients
<draft-ietf-dhc-ddns-resolution-01.txt>
Status of this Memo
This document is an Internet-Draft and is in full conformance with
all provisions of Section 10 of RFC2026.
Internet-Drafts are working documents of the Internet Engineering
Task Force (IETF), its areas, and its working groups. Note that
other groups may also distribute working documents as
Internet-Drafts.
Internet-Drafts are draft documents valid for a maximum of six
months and may be updated, replaced, or obsoleted by other documents
at any time. It is inappropriate to use Internet-Drafts as reference
material or to cite them other than as "work in progress."
The list of current Internet-Drafts can be accessed at
http://www.ietf.org/ietf/1id-abstracts.txt.
The list of Internet-Draft Shadow Directories can be accessed at
http://www.ietf.org/shadow.html.
This Internet-Draft will expire on August 31, 2001.
Copyright Notice
Copyright (C) The Internet Society (2001). All Rights Reserved.
Abstract
DHCP provides a powerful mechanism for IP host configuration.
However, the configuration capability provided by DHCP does not
include updating DNS(RFC1034[1], RFC1035[2]), and specifically
updating the name to address and address to name mappings maintained
in the DNS.
The "Client FQDN Option"[13] specifies the client FQDN option,
through which DHCP clients and servers can exchange information
about client FQDNs. This document describes techniques for the
resolution of DNS name conflicts among DHCP clients.
Stapp Expires August 31, 2001 [Page 1]
Internet-Draft Resolution of Name Conflicts March 2001
Table of Contents
1. Terminology . . . . . . . . . . . . . . . . . . . . . . . . 3
2. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 3
3. Issues with DDNS in DHCP Environments . . . . . . . . . . . 3
3.1 Name Conflicts . . . . . . . . . . . . . . . . . . . . . . . 4
3.2 Multiple DHCP servers . . . . . . . . . . . . . . . . . . . 5
3.3 Use of the DHCID RR . . . . . . . . . . . . . . . . . . . . 5
3.3.1 Format of the DHCID RRDATA . . . . . . . . . . . . . . . . . 5
3.4 DNS RR TTLs . . . . . . . . . . . . . . . . . . . . . . . . 7
4. Procedures for performing DNS updates . . . . . . . . . . . 7
4.1 Adding A RRs to DNS . . . . . . . . . . . . . . . . . . . . 7
4.2 Adding PTR RR Entries to DNS . . . . . . . . . . . . . . . . 8
4.3 Removing Entries from DNS . . . . . . . . . . . . . . . . . 9
4.4 Updating other RRs . . . . . . . . . . . . . . . . . . . . . 9
5. Security Considerations . . . . . . . . . . . . . . . . . . 9
6. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . 10
References . . . . . . . . . . . . . . . . . . . . . . . . . 11
Author's Address . . . . . . . . . . . . . . . . . . . . . . 12
Full Copyright Statement . . . . . . . . . . . . . . . . . . 13
Stapp Expires August 31, 2001 [Page 2]
Internet-Draft Resolution of Name Conflicts March 2001
1. Terminology
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",
"SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this
document are to be interpreted as described in RFC 2119[6].
2. Introduction
"The Client FQDN Option"[13] includes a description of the operation
of DHCP[3] clients and servers that use the client FQDN option.
Through the use of the client FQDN option, DHCP clients and servers
can negotiate the client's FQDN and the allocation of responsibility
for updating the DHCP client's A RR. This document identifies
situations in which conflicts in the use of FQDNs may arise among
DHCP clients, and describes a strategy for the use of the DHCID DNS
resource record[11] in resolving those conflicts.
In any case, whether a site permits all, some, or no DHCP servers
and clients to perform DNS updates into the zones which it controls
is entirely a matter of local administrative policy. This document
does not require any specific administrative policy, and does not
propose one. The range of possible policies is very broad, from
sites where only the DHCP servers have been given credentials that
the DNS servers will accept, to sites where each individual DHCP
client has been configured with credentials which allow the client
to modify its own domain name. Compliant implementations MAY support
some or all of these possibilities. Furthermore, this specification
applies only to DHCP client and server processes: it does not apply
to other processes which initiate DNS updates.
3. Issues with DDNS in DHCP Environments
There are two DNS update situations that require special
consideration in DHCP environments: cases where more than one DHCP
client has been configured with the same FQDN, and cases where more
than one DHCP server has been given authority to perform DNS updates
in a zone. In these cases, it is possible for DNS records to be
modified in inconsistent ways unless the updaters have a mechanism
that allows them to detect anomolous situations. If DNS updaters can
detect these situations, site administrators can configure the
updaters' behavior so that the site's policies can be enforced. We
use the term "Name Conflict" to refer to cases where more than one
DHCP client has been associated with a single FQDN. This
specification describes a mechanism designed to allow updaters to
detect these situations, and requires that DHCP implementations use
this mechanism by default.
Stapp Expires August 31, 2001 [Page 3]
Internet-Draft Resolution of Name Conflicts March 2001
3.1 Name Conflicts
How can the entity updating an A RR (either the DHCP client or DHCP
server) detect that a domain name has an A RR which is already in
use by a different DHCP client? Similarly, should a DHCP client or
server update a domain name which has an A RR that has been
configured by an administrator? In either of these cases, the
domain name in question would either have an additional A RR, or
would have its original A RR replaced by the new record. Either of
these effects may be considered undesirable by some sites. Different
authority and credential models have different levels of exposure to
name conflicts.
1. Client updates A RR, uses Secure DNS Update with credentials
that are associated with the client's FQDN, and exclusive to the
client. Name conflicts in this scenario are unlikely (though not
impossible), since the client has received credentials specific
to the name it desires to use. This implies that the name has
already been allocated (through some implementation- or
organization-specific procedure) to that client.
2. Client updates A RR, uses Secure DNS Update with credentials
that are valid for any name in the zone. Name conflicts in this
scenario are possible, since the credentials necessary for the
client to update DNS are not necessarily name-specific. Thus,
for the client to be attempting to update a unique name requires
the existence of some administrative procedure to ensure client
configuration with unique names.
3. Server updates the A RR, uses a name for the client which is
known to the server. Name conflicts in this scenario are likely
unless prevented by the server's name configuration procedures.
See Section 5 for security issues with this form of deployment.
4. Server updates the A RR, uses a name supplied by the client.
Name conflicts in this scenario are highly likely, even with
administrative procedures designed to prevent them. (This
scenario is a popular one in real-world deployments in many
types of organizations.) See Section 5 for security issues with
this type of deployment.
Scenarios 2, 3, and 4 rely on administrative procedures to ensure
name uniqueness for DNS updates, and these procedures may break
down. Experience has shown that, in fact, these procedures will
break down at least occasionally. The question is what to do when
these procedures break down or, for example in scenario #4, may not
even exist.
Stapp Expires August 31, 2001 [Page 4]
Internet-Draft Resolution of Name Conflicts March 2001
In all cases of name conflicts, the desire is to offer two modes of
operation to the administrator of the combined DHCP-DNS capability:
first-update-wins (i.e., the first updating entity gets the name) or
most-recent-update-wins (i.e., the last updating entity for a name
gets the name).
3.2 Multiple DHCP servers
If multiple DHCP servers are able to update the same DNS zones, or
if DHCP servers are performing A RR updates on behalf of DHCP
clients, and more than one DHCP server may be able to serve
addresses to the same DHCP clients, the DHCP servers should be able
to provide reasonable and consistent DNS name update behavior for
DHCP clients.
3.3 Use of the DHCID RR
A solution to both of these problems is for the updating entities
(both DHCP clients and DHCP servers) to be able to detect that
another entity has been associated with a DNS name, and to offer
administrators the opportunity to configure update behavior.
Specifically, a DHCID RR, described in DHCID RR[11] is used to
associate client identification information with a DNS name and the
A RR associated with that name. When either a client or server adds
an A RR for a client, it also adds a DHCID RR which specifies a
unique client identity (based on a "client specifier" created from
the client's client-id or MAC address). In this model, only one A
RR is associated with a given DNS name at a time.
By associating this ownership information with each A RR,
cooperating DNS updating entities may determine whether their client
is the first or last updater of the name (and implement the
appropriately configured administrative policy), and DHCP clients
which currently have domain names may move from one DHCP server to
another without losing their DNS names.
The specific algorithms utilizing the DHCID RR to signal client
ownership are explained below. The algorithms only work in the case
where the updating entities all cooperate -- this approach is
advisory only and is not substitute for DNS security, nor is it
replaced by DNS security.
3.3.1 Format of the DHCID RRDATA
The DHCID RR used to hold the DHCP client's identity is formatted as
follows:
The name of the DHCID RR is the name of the A or PTR RR which refers
Stapp Expires August 31, 2001 [Page 5]
Internet-Draft Resolution of Name Conflicts March 2001
to the DHCP client.
The RDATA section of a DHCID RR in transmission contains RDLENGTH
bytes of binary data. From the perspective of DHCP clients and
servers, the DHC resource record consists of a 16-bit identifier
type, followed by one or more bytes representing the actual
identifier. There are two possible forms for a DHCID RR - one that
is used when the client's link-layer address is being used to
identify it, and one that is used when some DHCP option that the
DHCP client has sent is being used to identify it.
DISCUSSION:
Implementors should note that the actual identifying data is
never placed into the DNS directly. Instead, the client-identity
data is used as the input into a one-way hash algorithm, and the
output of that hash is then used as DNS RRDATA. This has been
specified in order to avoid placing data about DHCP clients that
some sites might consider sensitive into the DNS.
When the updater is using the client's link-layer address, the first
two bytes of the DHCID RRDATA MUST be zero. To generate the rest of
the resource record, the updater MUST compute a one-way hash using
the MD5[12] algorithm across a buffer containing the client's
network hardware type and link-layer address. Specifically, the
first byte of the buffer contains the network hardware type as it
appears in the DHCP htype field of the client's DHCPREQUEST message.
All of the significant bytes of the chaddr field in the client's
DHCPREQUEST message follow, in the same order in which the bytes
appear in the DHCPREQUEST message. The number of significant bytes
in the chaddr field is specified in the hlen field of the
DHCPREQUEST message.
When the updater is using a DHCP option sent by the client in its
DHCPREQUEST message, the first two bytes of the DHCID RR MUST be the
option code of that option, in network byte order. For example, if
the DHCP client identifier option is being used, the first byte of
the DHCID RR should be zero, and the second byte should be 61
decimal. The rest of the DHCID RR MUST contain the results of
computing a one-way hash across the payload of the option being
used, using the MD5 algorithm. The payload of a DHCP option consists
of the bytes of the option following the option code and length.
The two byte identifier code 0xffff is reserved for future
assignment.
In order for independent DHCP implementations to be able to use the
DHCID RR as a prerequisite in dynamic DNS updates, each updater must
be able to reliably choose the same identifier that any other would
Stapp Expires August 31, 2001 [Page 6]
Internet-Draft Resolution of Name Conflicts March 2001
choose. To make this possible, we specify a prioritization which
will ensure that for any given DHCP client request, any updater will
select the same client-identity data. All updaters MUST use this
order of prioritization by default, but all implementations SHOULD
be configurable to use a different prioritization if so desired by
the site administrators. Because of the possibility of future
changes in the DHCP protocol, implementors SHOULD check for updated
versions of this draft when implementing new DHCP clients and
servers which can perform DDNS updates, and also when releasing new
versions of existing clients and servers.
DHCP clients and servers should use the following forms of client
identification, starting with the most preferable, and finishing
with the least preferable. If the client does not send any of these
forms of identification, the DHCP/DDNS interaction is not defined by
this specification. The most preferable form of identification is
the Globally Unique Identifier Option [TBD]. Next is the DHCP
Client Identifier option. Last is the client's link-layer address,
as conveyed in its DHCPREQUEST message. Implementors should note
that the link-layer address cannot be used if there are no
significant bytes in the chaddr field of the DHCP client's request,
because this does not constitute a unique identifier.
3.4 DNS RR TTLs
RRs associated with DHCP clients may be more volatile than
statically configured RRs. DHCP clients and servers which perform
dynamic updates should attempt to specify resource record TTLs which
reflect this volatility, in order to minimize the possibility that
there will be stale records in resolvers' caches. A reasonable basis
for RR TTLs is the lease duration itself: TTLs of 1/2 or 1/3 the
expected lease duration might be reasonable defaults. Because
configured DHCP lease times vary widely from site to site, it may
also be desirable to establish a fixed TTL ceiling. DHCP clients and
servers MAY allow administrators to configure the TTLs they will
supply, possibly as a fraction of the actual lease time, or as a
fixed value.
4. Procedures for performing DNS updates
4.1 Adding A RRs to DNS
When a DHCP client or server intends to update an A RR, it first
prepares a DNS UPDATE query which includes as a prerequisite the
assertion that the name does not exist. The update section of the
query attempts to add the new name and its IP address mapping (an A
RR), and the DHCID RR with its unique client-identity.
If this update operation succeeds, the updater can conclude that it
Stapp Expires August 31, 2001 [Page 7]
Internet-Draft Resolution of Name Conflicts March 2001
has added a new name whose only RRs are the A and DHCID RR records.
The A RR update is now complete (and a client updater is finished,
while a server might proceed to perform a PTR RR update).
If the first update operation fails with YXDOMAIN, the updater can
conclude that the intended name is in use. The updater then
attempts to confirm that the DNS name is not being used by some
other host. The updater prepares a second UPDATE query in which the
prerequisite is that the desired name has attached to it a DHCID RR
whose contents match the client identity. The update section of
this query deletes the existing A records on the name, and adds the
A record that matches the DHCP binding and the DHCID RR with the
client identity.
If this query succeeds, the updater can conclude that the current
client was the last client associated with the domain name, and that
the name now contains the updated A RR. The A RR update is now
complete (and a client updater is finished, while a server would
then proceed to perform a PTR RR update).
If the second query fails with NXRRSET, the updater must conclude
that the client's desired name is in use by another host. At this
juncture, the updater can decide (based on some administrative
configuration outside of the scope of this document) whether to let
the existing owner of the name keep that name, and to (possibly)
perform some name disambiguation operation on behalf of the current
client, or to replace the RRs on the name with RRs that represent
the current client. If the configured policy allows replacement of
existing records, the updater submits a query that deletes the
existing A RR and the existing DHCID RR, adding A and DHCID RRs that
represent the IP address and client-identity of the new client.
DISCUSSION:
The updating entity may be configured to allow the existing DNS
records on the domain name to remain unchanged, and to perform
disambiguation on the name of the current client in order to
attempt to generate a similar but unique name for the current
client. In this case, once another candidate name has been
generated, the updater should restart the process of adding an A
RR as specified in this section.
4.2 Adding PTR RR Entries to DNS
The DHCP server submits a DNS query which deletes all of the PTR RRs
associated with the lease IP address, and adds a PTR RR whose data
is the client's (possibly disambiguated) host name. The server also
adds a DHCID RR specified in Section 3.3.
Stapp Expires August 31, 2001 [Page 8]
Internet-Draft Resolution of Name Conflicts March 2001
4.3 Removing Entries from DNS
The most important consideration in removing DNS entries is be sure
that an entity removing a DNS entry is only removing an entry that
it added, or for which an administrator has explicitly assigned it
responsibility.
When a lease expires or a DHCP client issues a DHCPRELEASE request,
the DHCP server SHOULD delete the PTR RR that matches the DHCP
binding, if one was successfully added. The server's update query
SHOULD assert that the name in the PTR record matches the name of
the client whose lease has expired or been released.
The entity chosen to handle the A record for this client (either the
client or the server) SHOULD delete the A record that was added when
the lease was made to the client.
In order to perform this delete, the updater prepares an UPDATE
query which contains two prerequisites. The first prerequisite
asserts that the DHCID RR exists whose data is the client identity
described in Section 3.3. The second prerequisite asserts that the
data in the A RR contains the IP address of the lease that has
expired or been released.
If the query fails, the updater MUST NOT delete the DNS name. It
may be that the host whose lease on the server has expired has moved
to another network and obtained a lease from a different server,
which has caused the client's A RR to be replaced. It may also be
that some other client has been configured with a name that matches
the name of the DHCP client, and the policy was that the last client
to specify the name would get the name. In this case, the DHCID RR
will no longer match the updater's notion of the client-identity of
the host pointed to by the DNS name.
4.4 Updating other RRs
The procedures described in this document only cover updates to the
A and PTR RRs. Updating other types of RRs is outside the scope of
this document.
5. Security Considerations
Unauthenticated updates to the DNS can lead to tremendous confusion,
through malicious attack or through inadvertent misconfiguration.
Administrators should be wary of permitting unsecured DNS updates to
zones which are exposed to the global Internet. Both DHCP clients
and servers SHOULD use some form of update request origin
authentication procedure (e.g., Secure DNS Dynamic Update[10]) when
performing DNS updates.
Stapp Expires August 31, 2001 [Page 9]
Internet-Draft Resolution of Name Conflicts March 2001
Whether a DHCP client may be responsible for updating an FQDN to IP
address mapping, or whether this is the responsibility of the DHCP
server is a site-local matter. The choice between the two
alternatives may be based on the security model that is used with
the Dynamic DNS Update protocol (e.g., only a client may have
sufficient credentials to perform updates to the FQDN to IP address
mapping for its FQDN).
Whether a DHCP server is always responsible for updating the FQDN to
IP address mapping (in addition to updating the IP to FQDN mapping),
regardless of the wishes of an individual DHCP client, is also a
site-local matter. The choice between the two alternatives may be
based on the security model that is being used with dynamic DNS
updates. In cases where a DHCP server is performing DNS updates on
behalf of a client, the DHCP server should be sure of the DNS name
to use for the client, and of the identity of the client.
Currently, it is difficult for DHCP servers to develop much
confidence in the identities of its clients, given the absence of
entity authentication from the DHCP protocol itself. There are many
ways for a DHCP server to develop a DNS name to use for a client,
but only in certain relatively unusual circumstances will the DHCP
server know for certain the identity of the client. If DHCP
Authentication[9] becomes widely deployed this may become more
customary.
One example of a situation which offers some extra assurances is one
where the DHCP client is connected to a network through an MCNS
cable modem, and the CMTS (head-end) of the cable modem ensures that
MAC address spoofing simply does not occur. Another example of a
configuration that might be trusted is one where clients obtain
network access via a network access server using PPP. The NAS itself
might be obtaining IP addresses via DHCP, encoding a client
identification into the DHCP client-id option. In this case, the
network access server as well as the DHCP server might be operating
within a trusted environment, in which case the DHCP server could be
configured to trust that the user authentication and authorization
procedure of the remote access server was sufficient, and would
therefore trust the client identification encoded within the DHCP
client-id.
6. Acknowledgements
Many thanks to Mark Beyer, Jim Bound, Ralph Droms, Robert Elz, Peter
Ford, Edie Gunter, Andreas Gustafsson, R. Barr Hibbs, Kim Kinnear,
Stuart Kwan, Ted Lemon, Ed Lewis, Michael Lewis, Josh Littlefield,
Michael Patton, and Glenn Stump for their review and comments.
References
Stapp Expires August 31, 2001 [Page 10]
Internet-Draft Resolution of Name Conflicts March 2001
[1] Mockapetris, P., "Domain names - Concepts and Facilities", RFC
1034, Nov 1987.
[2] Mockapetris, P., "Domain names - Implementation and
Specification", RFC 1035, Nov 1987.
[3] Droms, R., "Dynamic Host Configuration Protocol", RFC 2131,
March 1997.
[4] Marine, A., Reynolds, J. and G. Malkin, "FYI on Questions and
Answers to Commonly asked ``New Internet User'' Questions",
RFC 1594, March 1994.
[5] Vixie, P., Thomson, S., Rekhter, Y. and J. Bound, "Dynamic
Updates in the Domain Name System", RFC 2136, April 1997.
[6] Bradner, S., "Key words for use in RFCs to Indicate
Requirement Levels", RFC 2119, March 1997.
[7] Vixie, P., "Extension Mechanisms for DNS (EDNS0)", RFC 2671,
August 1999.
[8] Vixie, P., Gudmundsson, O., Eastlake, D. and B. Wellington,
"Secret Key Transaction Authentication for DNS (TSIG)", RFC
2845, May 2000.
[9] Droms, R. and W. Arbaugh, "Authentication for DHCP Messages
(draft-ietf-dhc-authentication-*)", June 1999.
[10] Wellington, B., "Secure DNS Dynamic Update", RFC 3007,
November 2000.
[11] Stapp, M., Gustafsson, A. and T. Lemon, "A DNS RR for Encoding
DHCP Information (draft-ietf-dnsext-dhcid-rr-*)", November
2000.
[12] Rivest, R., "The MD5 Message Digest Algorithm", RFC 1321,
April 1992.
[13] Stapp, M. and Y. Rekhter, "The DHCP Client FQDN Option
(draft-ietf-dhc-fqdn-option-*.txt)", July 2000.
Stapp Expires August 31, 2001 [Page 11]
Internet-Draft Resolution of Name Conflicts March 2001
Author's Address
Mark Stapp
Cisco Systems, Inc.
250 Apollo Dr.
Chelmsford, MA 01824
USA
Phone: 978.244.8498
EMail: mjs@cisco.com
Stapp Expires August 31, 2001 [Page 12]
Internet-Draft Resolution of Name Conflicts March 2001
Full Copyright Statement
Copyright (C) The Internet Society (2001). All Rights Reserved.
This document and translations of it may be copied and furnished to
others, and derivative works that comment on or otherwise explain it
or assist in its implementation may be prepared, copied, published
and distributed, in whole or in part, without restriction of any
kind, provided that the above copyright notice and this paragraph
are included on all such copies and derivative works. However, this
document itself may not be modified in any way, such as by removing
the copyright notice or references to the Internet Society or other
Internet organizations, except as needed for the purpose of
developing Internet standards in which case the procedures for
copyrights defined in the Internet Standards process must be
followed, or as required to translate it into languages other than
English.
The limited permissions granted above are perpetual and will not be
revoked by the Internet Society or its successors or assigns.
This document and the information contained herein is provided on an
"AS IS" basis and THE INTERNET SOCIETY AND THE INTERNET ENGINEERING
TASK FORCE DISCLAIMS ALL WARRANTIES, EXPRESS OR IMPLIED, INCLUDING
BUT NOT LIMITED TO ANY WARRANTY THAT THE USE OF THE INFORMATION
HEREIN WILL NOT INFRINGE ANY RIGHTS OR ANY IMPLIED WARRANTIES OF
MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE.
Acknowledgement
Funding for the RFC editor function is currently provided by the
Internet Society.
Stapp Expires August 31, 2001 [Page 13]
