RFC 9103




Internet Engineering Task Force (IETF)                         W. Toorop
Request for Comments: 9103                                    NLnet Labs
Updates: 1995, 5936, 7766                                   S. Dickinson
Category: Standards Track                                     Sinodun IT
ISSN: 2070-1721                                                 S. Sahib
                                                          Brave Software
                                                                 P. Aras
                                                               A. Mankin
                                                              Salesforce
                                                             August 2021


                       DNS Zone Transfer over TLS

Abstract



   DNS zone transfers are transmitted in cleartext, which gives
   attackers the opportunity to collect the content of a zone by
   eavesdropping on network connections.  The DNS Transaction Signature
   (TSIG) mechanism is specified to restrict direct zone transfer to
   authorized clients only, but it does not add confidentiality.  This
   document specifies the use of TLS, rather than cleartext, to prevent
   zone content collection via passive monitoring of zone transfers: XFR
   over TLS (XoT).  Additionally, this specification updates RFC 1995
   and RFC 5936 with respect to efficient use of TCP connections and RFC
   7766 with respect to the recommended number of connections between a
   client and server for each transport.

Status of This Memo



   This is an Internet Standards Track document.

   This document is a product of the Internet Engineering Task Force
   (IETF).  It represents the consensus of the IETF community.  It has
   received public review and has been approved for publication by the
   Internet Engineering Steering Group (IESG).  Further information on
   Internet Standards is available in Section 2 of RFC 7841.

   Information about the current status of this document, any errata,
   and how to provide feedback on it may be obtained at
   https://www.rfc-editor.org/info/rfc9103.

Copyright Notice



   Copyright (c) 2021 IETF Trust and the persons identified as the
   document authors.  All rights reserved.

   This document is subject to BCP 78 and the IETF Trust's Legal
   Provisions Relating to IETF Documents
   (https://trustee.ietf.org/license-info) in effect on the date of
   publication of this document.  Please review these documents
   carefully, as they describe your rights and restrictions with respect
   to this document.  Code Components extracted from this document must
   include Simplified BSD License text as described in Section 4.e of
   the Trust Legal Provisions and are provided without warranty as
   described in the Simplified BSD License.

Table of Contents



   1.  Introduction
   2.  Terminology
   3.  Threat Model
   4.  Design Considerations for XoT
   5.  Connection and Data Flows in Existing XFR Mechanisms
     5.1.  AXFR Mechanism
     5.2.  IXFR Mechanism
     5.3.  Data Leakage of NOTIFY and SOA Message Exchanges
       5.3.1.  NOTIFY
       5.3.2.  SOA
   6.  Updates to Existing Specifications
     6.1.  Update to RFC 1995 for IXFR over TCP
     6.2.  Update to RFC 5936 for AXFR over TCP
     6.3.  Updates to RFCs 1995 and 5936 for XFR over TCP
       6.3.1.  Connection Reuse
       6.3.2.  AXFRs and IXFRs on the Same Connection
       6.3.3.  XFR Limits
       6.3.4.  The edns-tcp-keepalive EDNS(0) Option
       6.3.5.  Backwards Compatibility
     6.4.  Update to RFC 7766
   7.  XoT Specification
     7.1.  Connection Establishment
     7.2.  TLS Versions
     7.3.  Port Selection
     7.4.  High-Level XoT Descriptions
     7.5.  XoT Transfers
     7.6.  XoT Connections
     7.7.  XoT vs. ADoT
     7.8.  Response RCODES
     7.9.  AXoT Specifics
       7.9.1.  Padding AXoT Responses
     7.10. IXoT Specifics
       7.10.1.  Condensation of Responses
       7.10.2.  Fallback to AXFR
       7.10.3.  Padding of IXoT Responses
     7.11. Name Compression and Maximum Payload Sizes
   8.  Multi-primary Configurations
   9.  Authentication Mechanisms
     9.1.  TSIG
     9.2.  SIG(0)
     9.3.  TLS
       9.3.1.  Opportunistic TLS
       9.3.2.  Strict TLS
       9.3.3.  Mutual TLS
     9.4.  IP-Based ACL on the Primary
     9.5.  ZONEMD
   10. XoT Authentication
   11. Policies for Both AXoT and IXoT
   12. Implementation Considerations
   13. Operational Considerations
   14. IANA Considerations
   15. Security Considerations
   16. References
     16.1.  Normative References
     16.2.  Informative References
   Appendix A.  XoT Server Connection Handling
     A.1.  Listening Only on a Specific IP Address for TLS
     A.2.  Client-Specific TLS Acceptance
     A.3.  SNI-Based TLS Acceptance
     A.4.  Transport-Specific Response Policies
       A.4.1.  SNI-Based Response Policies
   Acknowledgements
   Contributors

   Authors' Addresses



1.  Introduction



   DNS has a number of privacy vulnerabilities, as discussed in detail
   in [RFC9076].  Query privacy between stub resolvers and recursive
   resolvers has received the most attention to date, with Standards
   Track documents for both DNS over TLS (DoT) [RFC7858] and DNS over
   HTTPS (DoH) [RFC8484] and a proposal for DNS over QUIC
   [DPRIVE-DNSOQUIC].  There is ongoing work on DNS privacy requirements
   for exchanges between recursive resolvers and authoritative servers
   and some suggestions for how signaling of DoT support by
   authoritative name servers might work.  However, there is currently
   no RFC that specifically defines recursive-to-authoritative DNS over
   TLS (ADoT).

   [RFC9076] establishes that a stub resolver's DNS query transactions
   are not public and that they need protection, but, on zone transfer
   [RFC1995] [RFC5936], it says only:

   |  Privacy risks for the holder of a zone (the risk that someone gets
   |  the data) are discussed in [RFC5155] and [RFC5936].

   In what way is exposing the full contents of a zone a privacy risk?
   The contents of the zone could include information such as names of
   persons used in names of hosts.  Best practice is not to use personal
   information for domain names, but many such domain names exist.  The
   contents of the zone could also include references to locations that
   allow inference about location information of the individuals
   associated with the zone's organization.  It could also include
   references to other organizations.  Examples of this could be:

   *  Person-laptop.example.org

   *  MX-for-Location.example.org

   *  Service-tenant-from-another-org.example.org

   Additionally, the full zone contents expose all the IP addresses of
   endpoints held in the DNS records, which can make reconnaissance and
   attack targeting easier, particularly for IPv6 addresses or private
   networks.  There may also be regulatory, policy, or other reasons why
   the zone contents in full must be treated as private.

   Neither of the RFCs mentioned in [RFC9076] contemplate the risk that
   someone gets the data through eavesdropping on network connections,
   only via enumeration or unauthorized transfer, as described in the
   following paragraphs.

   Zone enumeration is trivially possible for DNSSEC zones that use
   NSEC, i.e., queries for the authenticated denial-of-existence records
   allow a client to walk through the entire zone contents.  [RFC5155]
   specifies NSEC3, a mechanism to provide measures against zone
   enumeration for DNSSEC-signed zones (a goal was to make it as hard to
   enumerate a DNSSEC-signed zone as an unsigned zone).  Whilst this is
   widely used, it has been demonstrated that zone walking is possible
   for precomputed NSEC3 using attacks, such as those described in
   [NSEC3-attacks].  This prompted further work on an alternative
   mechanism for DNSSEC-authenticated denial of existence (NSEC5
   [NSEC5]); however, questions remain over the practicality of this
   mechanism.

   [RFC5155] does not address data obtained outside zone enumeration
   (nor does [NSEC5]).  Preventing eavesdropping of zone transfers (as
   described in this document) is orthogonal to preventing zone
   enumeration, though they aim to protect the same information.

   [RFC5936] specifies using TSIG [RFC8945] for authorization of the
   clients of a zone transfer and for data integrity but does not
   express any need for confidentiality, and TSIG does not offer
   encryption.

   Section 8 of the NIST document "Secure Domain Name System (DNS)
   Deployment Guide" [NIST-GUIDE] discusses restricting access for zone
   transfers using Access Control Lists (ACLs) and TSIG in more detail.
   It also discusses the possibility that specific deployments might
   choose to use a lower-level network layer to protect zone transfers,
   e.g., IPsec.

   It is noted that in all the common open-source implementations such
   ACLs are applied on a per-query basis (at the time of writing).
   Since requests typically occur on TCP connections, authoritative
   servers must therefore accept any TCP connection and then handle the
   authentication of each zone transfer (XFR) request individually.

   Because both AXFR (authoritative transfer) and IXFR (incremental zone
   transfer) are typically carried out over TCP from authoritative DNS
   protocol implementations, encrypting zone transfers using TLS
   [RFC8499] -- based closely on DoT [RFC7858] -- seems like a simple
   step forward.  This document specifies how to use TLS (1.3 or later)
   as a transport to prevent zone collection from zone transfers.

   This document also updates the previous specifications for zone
   transfers to clarify and extend them, mainly with respect to TCP
   usage:

   *  [RFC1995] (IXFR) and [RFC5936] (AXFR) are both updated to add
      further specification on efficient use of TCP connections.

   *  Section 6.2.2 of [RFC7766] ("DNS Transport over TCP -
      Implementation Requirements") is updated with a new recommendation
      about the number of connections between a client and server for
      each transport.

2.  Terminology



   The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",
   "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and
   "OPTIONAL" in this document are to be interpreted as described in BCP
   14 [RFC2119] [RFC8174] when, and only when, they appear in all
   capitals, as shown here.

   Privacy terminology is as described in Section 3 of [RFC6973].

   DNS terminology is as described in [RFC8499].  Note that, as in
   [RFC8499], the terms 'primary' and 'secondary' are used for two
   servers engaged in zone transfers.

   DoT:   DNS over TLS, as specified in [RFC7858]

   XFR over TCP:  Used to mean both IXFR over TCP [RFC1995] and AXFR
          over TCP [RFC5936]

   XoT:   XFR-over-TLS mechanisms, as specified in this document, which
          apply to both AXFR over TLS and IXFR over TLS (XoT is
          pronounced 'zot' since X here stands for 'zone transfer')

   AXoT:  AXFR over TLS

   IXoT:  IXFR over TLS

3.  Threat Model



   The threat model considered here is one where the current contents
   and size of the zone are considered sensitive and should be protected
   during transfer.

   The threat model does not, however, consider the existence of a zone,
   the act of zone transfer between two entities, nor the identities of
   the name servers hosting a zone (including both those acting as
   hidden primaries/secondaries or directly serving the zone) as
   sensitive information.  The proposed mechanism does not attempt to
   obscure such information.  The reasons for this include:

   *  much of this information can be obtained by various methods,
      including active scanning of the DNS, and

   *  an attacker who can monitor network traffic can rather easily
      infer relations between name servers simply from traffic patterns,
      even when some or all of the traffic is encrypted (in terms of
      current deployments).

   The model does not consider attacks on the mechanisms that trigger a
   zone transfer, e.g., NOTIFY messages.

   It is noted that simply using XoT will indicate a desire by the zone
   owner that the contents of the zone remain confidential and so could
   be subject to blocking (e.g., via blocking of port 853) if an
   attacker had such capabilities.  However, this threat is likely true
   of any such mechanism that attempts to encrypt data passed between
   name servers, e.g., IPsec.

4.  Design Considerations for XoT



   The following principles were considered in the design for XoT:

   Confidentiality:  Clearly using an encrypted transport for zone
      transfers will defeat zone content leakage that can occur via
      passive surveillance.

   Authentication:  Use of single or mutual TLS (mTLS) authentication
      (in combination with ACLs) can complement and potentially be an
      alternative to TSIG.

   Performance:
      *  Existing AXFR and IXFR mechanisms have the burden of backwards
         compatibility with older implementations based on the original
         specifications in [RFC1034] and [RFC1035].  For example, some
         older AXFR servers don't support using a TCP connection for
         multiple AXFR sessions or XFRs of different zones because they
         have not been updated to follow the guidance in [RFC5936].  Any
         implementation of XoT would obviously be required to implement
         optimized and interoperable transfers, as described in
         [RFC5936], e.g., transfer of multiple zones over one
         connection.

      *  Current usage of TCP for IXFR is suboptimal in some cases,
         i.e., connections are frequently closed after a single IXFR.

5.  Connection and Data Flows in Existing XFR Mechanisms



   The original specification for zone transfers in [RFC1034] and
   [RFC1035] was based on a polling mechanism: a secondary performed a
   periodic query for the SOA (start of zone authority) record (based on
   the refresh timer) to determine if an AXFR was required.

   [RFC1995] and [RFC1996] introduced the concepts of IXFR and NOTIFY,
   respectively, to provide for prompt propagation of zone updates.
   This has largely replaced AXFR where possible, particularly for
   dynamically updated zones.

   [RFC5936] subsequently redefined the specification of AXFR to improve
   performance and interoperability.

   In this document, the term 'XFR mechanism' is used to describe the
   entire set of message exchanges between a secondary and a primary
   that concludes with a successful AXFR or IXFR request/response.  This
   set may or may not include:

   *  NOTIFY messages

   *  SOA queries

   *  Fallback from IXFR to AXFR

   *  Fallback from IXFR over UDP to IXFR over TCP

   The term is used to encompass the range of permutations that are
   possible and is useful to distinguish the 'XFR mechanism' from a
   single XFR request/response exchange.

5.1.  AXFR Mechanism



   The figure below provides an outline of an AXFR mechanism including
   NOTIFYs.

      Secondary                            Primary

          |              NOTIFY               |
          | <-------------------------------- |  UDP
          | --------------------------------> |
          |          NOTIFY Response          |
          |                                   |
          |                                   |
          |            SOA Request            |
          | --------------------------------> |  UDP (or part of
          | <-------------------------------- |  a TCP session)
          |           SOA Response            |
          |                                   |
          |                                   |
          |                                   |
          |            AXFR Request           | ---
          | --------------------------------> |   |
          | <-------------------------------- |   |
          |          AXFR Response 1          |   |
          |             (Zone data)           |   |
          |                                   |   |
          | <-------------------------------- |   | TCP
          |          AXFR Response 2          |   | Session
          |             (Zone data)           |   |
          |                                   |   |
          | <-------------------------------- |   |
          |          AXFR Response 3          |   |
          |             (Zone data)           | ---
          |                                   |

                          Figure 1: AXFR Mechanism

   1.  An AXFR is often (but not always) preceded by a NOTIFY (over UDP)
       from the primary to the secondary.  A secondary may also initiate
       an AXFR based on a refresh timer or scheduled/triggered zone
       maintenance.

   2.  The secondary will normally (but not always) make an SOA query to
       the primary to obtain the serial number of the zone held by the
       primary.

   3.  If the primary serial is higher than the secondary's serial
       (using Serial Number Arithmetic [RFC1982]), the secondary makes
       an AXFR request (over TCP) to the primary, after which the AXFR
       data flows in one or more AXFR responses on the TCP connection.
       [RFC5936] defines this specific step as an 'AXFR session', i.e.,
       as an AXFR query message and the sequence of AXFR response
       messages returned for it.

   [RFC5936] re-specified AXFR, providing additional guidance beyond
   that provided in [RFC1034] and [RFC1035] and importantly specified
   that AXFR must use TCP as the transport protocol.

   Additionally, Sections 4.1, 4.1.1, and 4.1.2 of [RFC5936] provide
   improved guidance for AXFR clients and servers with regard to reuse
   of TCP connections for multiple AXFRs and AXFRs of different zones.
   However, [RFC5936] was constrained by having to be backwards
   compatible with some very early basic implementations of AXFR.  For
   example, it outlines that the SOA query can also happen on this
   connection.  However, this can cause interoperability problems with
   older implementations that support only the trivial case of one AXFR
   per connection.

5.2.  IXFR Mechanism



   The figure below provides an outline of the IXFR mechanism including
   NOTIFYs.

      Secondary                            Primary

          |              NOTIFY               |
          | <-------------------------------- |  UDP
          | --------------------------------> |
          |          NOTIFY Response          |
          |                                   |
          |                                   |
          |            SOA Request            |
          | --------------------------------> |  UDP or TCP
          | <-------------------------------- |
          |           SOA Response            |
          |                                   |
          |                                   |
          |                                   |
          |            IXFR Request           |
          | --------------------------------> |  UDP or TCP
          | <-------------------------------- |
          |            IXFR Response          |
          |             (Zone data)           |
          |                                   |
          |                                   | ---
          |            IXFR Request           |    |
          | --------------------------------> |    | Retry over
          | <-------------------------------- |    | TCP if
          |            IXFR Response          |    | required
          |             (Zone data)           | ---

                          Figure 2: IXFR Mechanism

   1.  An IXFR is normally (but not always) preceded by a NOTIFY (over
       UDP) from the primary to the secondary.  A secondary may also
       initiate an IXFR based on a refresh timer or scheduled/triggered
       zone maintenance.

   2.  The secondary will normally (but not always) make an SOA query to
       the primary to obtain the serial number of the zone held by the
       primary.

   3.  If the primary serial is higher than the secondary's serial
       (using Serial Number Arithmetic [RFC1982]), the secondary makes
       an IXFR request to the primary, after which the primary sends an
       IXFR response.

   [RFC1995] specifies that IXFR may use UDP if the entire IXFR response
   can be contained in a single DNS packet, otherwise, TCP is used.  In
   fact, it says:

   |  Thus, a client should first make an IXFR query using UDP.

   So there may be a fourth step above where the client falls back to
   IXFR over TCP.  There may also be an additional step where the
   secondary must fall back to AXFR because, e.g., the primary does not
   support IXFR.

   However, it is noted that most of the widely used open-source
   implementations of authoritative name servers (including both [BIND]
   and [NSD]) do IXFR using TCP by default in their latest releases.
   For BIND, TCP connections are sometimes used for SOA queries, but, in
   general, they are not used persistently and are closed after an IXFR
   is completed.

5.3.  Data Leakage of NOTIFY and SOA Message Exchanges



   This section presents a rationale for considering the encryption of
   the other messages in the XFR mechanism.

   Since the SOA of the published zone can be trivially discovered by
   simply querying the publicly available authoritative servers, leakage
   of this resource record (RR) via such a direct query is not discussed
   in the following sections.

5.3.1.  NOTIFY



   Unencrypted NOTIFY messages identify configured secondaries on the
   primary.

   [RFC1996] also states:

   |  If ANCOUNT>0, then the answer section represents an unsecure hint
   |  at the new RRset for this <QNAME,QCLASS,QTYPE>.

   But since the only query type (QTYPE) for NOTIFY defined at the time
   of this writing is SOA, this does not pose a potential leak.

5.3.2.  SOA



   For hidden XFR servers (either primaries or secondaries), an SOA
   response directly from that server only additionally leaks the degree
   of SOA serial number lag of any downstream secondary of that server.

6.  Updates to Existing Specifications



   For convenience, the term 'XFR over TCP' is used in this document to
   mean both IXFR over TCP and AXFR over TCP; therefore, statements that
   use that term update both [RFC1995] and [RFC5936] and implicitly also
   apply to XoT.  Differences in behavior specific to XoT are discussed
   in Section 7.

   Both [RFC1995] and [RFC5936] were published sometime before TCP
   became a widely supported transport for DNS.  [RFC1995], in fact,
   says nothing with respect to optimizing IXFRs over TCP or reusing
   already open TCP connections to perform IXFRs or other queries.
   Therefore, there arguably is an implicit assumption that a TCP
   connection is used for one and only one IXFR request.  Indeed, many
   major open-source implementations take this approach (at the time of
   this writing).  And whilst [RFC5936] gives guidance on connection
   reuse for AXFR, it predates more recent specifications describing
   persistent TCP connections (e.g., [RFC7766], [RFC7828]), and AXFR
   implementations again often make less-than-optimal use of open
   connections.

   Given this, new implementations of XoT will clearly benefit from
   specific guidance on TCP/TLS connection usage for XFR, because this
   will:

   *  result in more consistent XoT implementations with better
      interoperability and

   *  remove any need for XoT implementations to support legacy behavior
      for XoT connections that XFR-over-TCP implementations have
      historically often supported.

   Therefore, this document updates both the previous specifications for
   XFR over TCP ([RFC1995] and [RFC5936]) to clarify that:

   *  Implementations MUST use [RFC7766] ("DNS Transport over TCP -
      Implementation Requirements") to optimize the use of TCP
      connections.

   *  Whilst [RFC7766] states that "DNS clients SHOULD pipeline their
      queries" on TCP connections, it did not distinguish between XFRs
      and other queries for this behavior.  It is now recognized that
      XFRs are not as latency sensitive as other queries and can be
      significantly more complex for clients to handle, both because of
      the large amount of state that must be kept and because there may
      be multiple messages in the responses.  For these reasons, it is
      clarified here that a valid reason for not pipelining queries is
      when they are all XFR queries, i.e., clients sending multiple XFRs
      MAY choose not to pipeline those queries.  Clients that do not
      pipeline XFR queries therefore have no additional requirements to
      handle out-of-order or intermingled responses (as described
      later), since they will never receive them.

   *  Implementations SHOULD use the edns-tcp-keepalive EDNS(0) option
      [RFC7828] to manage persistent connections.  This is more flexible
      than the alternative of simply using fixed timeouts.

   The following sections include detailed clarifications on the updates
   to XFR behavior implied in [RFC7766] and how the use of [RFC7828]
   applies specifically to XFR exchanges.  They also discuss how IXFR
   and AXFR can reuse the same TCP connection.

   For completeness, the recent specification of extended DNS error
   (EDE) codes [RFC8914] is also mentioned here.  For zone transfers,
   when returning REFUSED to a zone transfer request from an
   'unauthorized' client (e.g., where the client is not listed in an ACL
   for zone transfers or does not sign the request with a valid TSIG
   key), the extended DNS error code 18 - Prohibited can also be sent.

6.1.  Update to RFC 1995 for IXFR over TCP



   For clarity, an IXFR-over-TCP server compliant with this
   specification MUST be able to handle multiple concurrent IXoT
   requests on a single TCP connection (for the same and different
   zones) and SHOULD send the responses as soon as they are available,
   which might be out of order compared to the requests.

6.2.  Update to RFC 5936 for AXFR over TCP



   For clarity, an AXFR-over-TCP server compliant with this
   specification MUST be able to handle multiple concurrent AXoT
   sessions on a single TCP connection (for the same and different
   zones).  The response streams for concurrent AXFRs MAY be
   intermingled, and AXFR-over-TCP clients compliant with this
   specification, which pipeline AXFR requests, MUST be able to handle
   this.

6.3.  Updates to RFCs 1995 and 5936 for XFR over TCP



6.3.1.  Connection Reuse



   As specified, XFR-over-TCP clients SHOULD reuse any existing open TCP
   connection when starting any new XFR request to the same primary, and
   for issuing SOA queries, instead of opening a new connection.  The
   number of TCP connections between a secondary and primary SHOULD be
   minimized (also see Section 6.4).

   Valid reasons for not reusing existing connections might include:

   *  As already noted in [RFC7766], separate connections for different
      zones might be preferred for operational reasons.  In this case,
      the number of concurrent connections for zone transfers SHOULD be
      limited to the total number of zones transferred between the
      client and server.

   *  A configured limit for the number of outstanding queries or XFR
      requests allowed on a single TCP connection has been reached.

   *  The message ID pool has already been exhausted on an open
      connection.

   *  A large number of timeouts or slow responses have occurred on an
      open connection.

   *  An edns-tcp-keepalive EDNS(0) option with a timeout of 0 has been
      received from the server, and the client is in the process of
      closing the connection (see Section 6.3.4).

   If no TCP connections are currently open, XFR clients MAY send SOA
   queries over UDP or a new TCP connection.

6.3.2.  AXFRs and IXFRs on the Same Connection



   Neither [RFC1995] nor [RFC5936] explicitly discuss the use of a
   single TCP connection for both IXFR and AXFR requests.  [RFC5936]
   does make the general statement:

   |  Non-AXFR session traffic can also use an open connection.

   In this document, the above is clarified to indicate that
   implementations capable of both AXFR and IXFR and compliant with this
   specification SHOULD:

   *  use the same TCP connection for both AXFR and IXFR requests to the
      same primary,

   *  pipeline such requests (if they pipeline XFR requests in general)
      and MAY intermingle them, and

   *  send the response(s) for each request as soon as they are
      available, i.e., responses MAY be sent intermingled.

   For some current implementations, adding all the above functionality
   would introduce significant code complexity.  In such a case, there
   will need to be an assessment of the trade-off between that and the
   performance benefits of the above for XFR.

6.3.3.  XFR Limits



   The server MAY limit the number of concurrent IXFRs, AXFRs, or total
   XFR transfers in progress (or from a given secondary) to protect
   server resources.  Servers SHOULD return SERVFAIL if this limit is
   hit, since it is a transient error and a retry at a later time might
   succeed (there is no previous specification for this behavior).

6.3.4.  The edns-tcp-keepalive EDNS(0) Option



   XFR clients that send the edns-tcp-keepalive EDNS(0) option on every
   XFR request provide the server with maximum opportunity to update the
   edns-tcp-keepalive timeout.  The XFR server may use the frequency of
   recent XFRs to calculate an average update rate as input to the
   decision of what edns-tcp-keepalive timeout to use.  If the server
   does not support edns-tcp-keepalive, the client MAY keep the
   connection open for a few seconds ([RFC7766] recommends that servers
   use timeouts of at least a few seconds).

   Whilst the specification for EDNS(0) [RFC6891] does not specifically
   mention AXFRs, it does say:

   |  If an OPT record is present in a received request, compliant
   |  responders MUST include an OPT record in their respective
   |  responses.

   In this document, the above is clarified to indicate that if an OPT
   record is present in a received AXFR request, compliant responders
   MUST include an OPT record in each of the subsequent AXFR responses.
   Note that this requirement, combined with the use of edns-tcp-
   keepalive, enables AXFR servers to signal the desire to close a
   connection (when existing transactions have competed) due to low
   resources by sending an edns-tcp-keepalive EDNS(0) option with a
   timeout of 0 on any AXFR response.  This does not signal that the
   AXFR is aborted, just that the server wishes to close the connection
   as soon as possible.

6.3.5.  Backwards Compatibility



   Certain legacy behaviors were noted in [RFC5936], with provisions
   that implementations may want to offer options to fallback to legacy
   behavior when interoperating with servers known to not support
   [RFC5936].  For purposes of interoperability, IXFR and AXFR
   implementations may want to continue offering such configuration
   options, as well as supporting some behaviors that were
   underspecified prior to this work (e.g., performing IXFR and AXFRs on
   separate connections).  However, XoT connections should have no need
   to do so.

6.4.  Update to RFC 7766



   [RFC7766] made general implementation recommendations with regard to
   TCP/TLS connection handling:

   |  To mitigate the risk of unintentional server overload, DNS clients
   |  MUST take care to minimize the number of concurrent TCP
   |  connections made to any individual server.  It is RECOMMENDED that
   |  for any given client/server interaction there SHOULD be no more
   |  than one connection for regular queries, one for zone transfers,
   |  and one for each protocol that is being used on top of TCP (for
   |  example, if the resolver was using TLS).  However, it is noted
   |  that certain primary/ secondary configurations with many busy
   |  zones might need to use more than one TCP connection for zone
   |  transfers for operational reasons (for example, to support
   |  concurrent transfers of multiple zones).

   Whilst this recommends a particular behavior for the clients using
   TCP, it does not relax the requirement for servers to handle 'mixed'
   traffic (regular queries and zone transfers) on any open TCP/TLS
   connection.  It also overlooks the potential that other transports
   might want to take the same approach with regard to using separate
   connections for different purposes.

   This specification updates the above general guidance in [RFC7766] to
   provide the same separation of connection purpose (regular queries
   and zone transfers) for all transports being used on top of TCP.

   Therefore, it is RECOMMENDED that for each protocol used on top of
   TCP in any given client/server interaction there SHOULD be no more
   than one connection for regular queries and one for zone transfers.

   As an illustration, it could be imagined that in the future such an
   interaction could hypothetically include one or all of the following:

   *  one TCP connection for regular queries

   *  one TCP connection for zone transfers

   *  one TLS connection for regular queries

   *  one TLS connection for zone transfers

   *  one DoH connection for regular queries

   *  one DoH connection for zone transfers

   Section 6.3.1 provides specific details of the reasons why more than
   one connection for a given transport might be required for zone
   transfers from a particular client.

7.  XoT Specification



7.1.  Connection Establishment



   During connection establishment, the Application-Layer Protocol
   Negotiation (ALPN) token "dot" [DoT-ALPN] MUST be selected in the TLS
   handshake.

7.2.  TLS Versions



   All implementations of this specification MUST use only TLS 1.3
   [RFC8446] or later.

7.3.  Port Selection



   The connection for XoT SHOULD be established using port 853, as
   specified in [RFC7858], unless there is mutual agreement between the
   primary and secondary to use a port other than port 853 for XoT.
   There MAY be agreement to use different ports for AXoT and IXoT or
   for different zones.

7.4.  High-Level XoT Descriptions



   It is useful to note that in XoT it is the secondary that initiates
   the TLS connection to the primary for an XFR request so that, in
   terms of connectivity, the secondary is the TLS client and the
   primary is the TLS server.

   The figure below provides an outline of the AXoT mechanism including
   NOTIFYs.

      Secondary                            Primary

          |              NOTIFY               |
          | <-------------------------------- |  UDP
          | --------------------------------> |
          |          NOTIFY Response          |
          |                                   |
          |                                   |
          |            SOA Request            |
          | --------------------------------> |  UDP (or part of
          | <-------------------------------- |  a TCP/TLS session)
          |           SOA Response            |
          |                                   |
          |                                   |
          |                                   |
          |            AXFR Request           | ---
          | --------------------------------> |   |
          | <-------------------------------- |   |
          |          AXFR Response 1          |   |
          |             (Zone data)           |   |
          |                                   |   |
          | <-------------------------------- |   | TLS
          |          AXFR Response 2          |   | Session
          |             (Zone data)           |   |
          |                                   |   |
          | <-------------------------------- |   |
          |          AXFR Response 3          |   |
          |             (Zone data)           | ---
          |                                   |

                          Figure 3: AXoT Mechanism

   The figure below provides an outline of the IXoT mechanism including
   NOTIFYs.

      Secondary                            Primary

          |              NOTIFY               |
          | <-------------------------------- |  UDP
          | --------------------------------> |
          |          NOTIFY Response          |
          |                                   |
          |                                   |
          |            SOA Request            |
          | --------------------------------> |  UDP (or part of
          | <-------------------------------- |  a TCP/TLS session)
          |           SOA Response            |
          |                                   |
          |                                   |
          |                                   |
          |            IXFR Request           | ---
          | --------------------------------> |    |
          | <-------------------------------- |    |
          |            IXFR Response          |    |
          |             (Zone data)           |    |
          |                                   |    | TLS
          |                                   |    | session
          |            IXFR Request           |    |
          | --------------------------------> |    |
          | <-------------------------------- |    |
          |            IXFR Response          |    |
          |             (Zone data)           | ---

                          Figure 4: IXoT Mechanism

7.5.  XoT Transfers



   For a zone transfer between two endpoints to be considered protected
   with XoT, all XFR requests and responses for that zone MUST be sent
   over TLS connections, where at a minimum:

   *  The client MUST authenticate the server by use of an
      authentication domain name using a Strict Privacy profile, as
      described in [RFC8310].

   *  The server MUST validate the client is authorized to request or
      proxy a zone transfer by using one or both of the following
      methods:

      -  mutual TLS (mTLS)

      -  an IP-based ACL (which can be either per message or per
         connection) combined with a valid TSIG/SIG(0) signature on the
         XFR request

   If only one method is selected, then mTLS is preferred because it
   provides strong cryptographic protection at both endpoints.

   Authentication mechanisms are discussed in full in Section 9, and the
   rationale for the above requirement is discussed in Section 10.
   Transfer group policies are discussed in Section 11.

7.6.  XoT Connections



   The details in Section 6 about, e.g., persistent connections and XFR
   message handling, are fully applicable to XoT connections as well.
   However, any behavior specified here takes precedence for XoT.

   If no TLS connections are currently open, XoT clients MAY send SOA
   queries over UDP, TCP, or TLS.

7.7.  XoT vs. ADoT



   As noted earlier, there is currently no specification for encryption
   of connections from recursive resolvers to authoritative servers.
   Some authoritative servers are experimenting with ADoT, and
   opportunistic encryption has also been raised as a possibility;
   therefore, it is highly likely that use of encryption by
   authoritative servers will evolve in the coming years.

   This raises questions in the short term with regard to TLS connection
   and message handling for authoritative servers.  In particular, there
   is likely to be a class of authoritative servers that wish to use XoT
   in the near future with a small number of configured secondaries but
   that do not wish to support DoT for regular queries from recursives
   in that same time frame.  These servers have to potentially cope with
   probing and direct queries from recursives and from test servers and
   also potential attacks that might wish to make use of TLS to overload
   the server.

   [RFC5936] clearly states that non-AXFR session traffic can use an
   open connection; however, this requirement needs to be reevaluated
   when considering the application of the same model to XoT.  Proposing
   that a server should also start responding to all queries received
   over TLS just because it has enabled XoT would be equivalent to
   defining a form of authoritative DoT.  This specification does not
   propose that, but it also does not prohibit servers from answering
   queries unrelated to XFR exchanges over TLS.  Rather, this
   specification simply outlines in later sections:

   *  the utilization of EDE codes by XoT servers in response to queries
      on TLS connections that they are not willing to answer (see
      Section 7.8)

   *  the operational and policy options that an operator of a XoT
      server has with regard to managing TLS connections and messages
      (see Appendix A)

7.8.  Response RCODES



   XoT clients and servers MUST implement EDE codes.  If a XoT server
   receives non-XoT traffic it is not willing to answer on a TLS
   connection, it SHOULD respond with REFUSED and the extended DNS error
   code 21 - Not Supported [RFC8914].  XoT clients should not send any
   further queries of this type to the server for a reasonable period of
   time (for example, one hour), i.e., long enough that the server
   configuration or policy might be updated.

   Historically, servers have used the REFUSED RCODE for many
   situations; therefore, clients often had no detailed information on
   which to base an error or fallback path when queries were refused.
   As a result, the client behavior could vary significantly.  XoT
   servers that refuse queries must cater to the fact that client
   behavior might vary from continually retrying queries regardless of
   receiving REFUSED to every query or, at the other extreme, clients
   may decide to stop using the server over any transport.  This might
   be because those clients are either non-XoT clients or do not
   implement EDE codes.

7.9.  AXoT Specifics



7.9.1.  Padding AXoT Responses



   The goal of padding AXoT responses is two fold:

   *  to obfuscate the actual size of the transferred zone to minimize
      information leakage about the entire contents of the zone

   *  to obfuscate the incremental changes to the zone between SOA
      updates to minimize information leakage about zone update activity
      and growth

   Note that the reuse of XoT connections for transfers of multiple
   different zones slightly complicates any attempt to analyze the
   traffic size and timing to extract information.  Also, effective
   padding may require the state to be kept because zones may grow and/
   or shrink over time.

   It is noted here that, depending on the padding policies eventually
   developed for XoT, the requirement to obfuscate the total zone size
   might require a server to create 'empty' AXoT responses, that is,
   AXoT responses that contain no RRs apart from an OPT RR containing
   the EDNS(0) option for padding.  For example, without this
   capability, the maximum size that a tiny zone could be padded to
   would theoretically be limited if there had to be a minimum of 1 RR
   per packet.

   However, as with existing AXFR, the last AXoT response message sent
   MUST contain the same SOA that was in the first message of the AXoT
   response series in order to signal the conclusion of the zone
   transfer.

   [RFC5936] says:

   |  Each AXFR response message SHOULD contain a sufficient number of
   |  RRs to reasonably amortize the per-message overhead, up to the
   |  largest number that will fit within a DNS message (taking the
   |  required content of the other sections into account, as described
   |  below).

   'Empty' AXoT responses generated in order to meet a padding
   requirement will be exceptions to the above statement.  For
   flexibility, for future proofing, and in order to guarantee support
   for future padding policies, it is stated here that secondary
   implementations MUST be resilient to receiving padded AXoT responses,
   including 'empty' AXoT responses that contain only an OPT RR
   containing the EDNS(0) option for padding.

   Recommendations of specific policies for padding AXoT responses are
   out of scope for this specification.  Detailed considerations of such
   policies and the trade-offs involved are expected to be the subject
   of future work.

7.10.  IXoT Specifics



7.10.1.  Condensation of Responses



   [RFC1995] says that condensation of responses is optional and MAY be
   done.  Whilst it does add complexity to generating responses, it can
   significantly reduce the size of responses.  However, any such
   reduction might be offset by increased message size due to padding.
   This specification does not update the optionality of condensation
   for XoT responses.

7.10.2.  Fallback to AXFR



   Fallback to AXFR can happen, for example, if the server is not able
   to provide an IXFR for the requested SOA.  Implementations differ in
   how long they store zone deltas and how many may be stored at any one
   time.

   Just as with IXFR over TCP, after a failed IXFR, an IXoT client
   SHOULD request the AXFR on the already open XoT connection.

7.10.3.  Padding of IXoT Responses



   The goal of padding IXoT responses is to obfuscate the incremental
   changes to the zone between SOA updates to minimize information
   leakage about zone update activity and growth.  Both the size and
   timing of the IXoT responses could reveal information.

   IXFR responses can vary greatly in size from the order of 100 bytes
   for one or two record updates to tens of thousands of bytes for
   large, dynamic DNSSEC-signed zones.  The frequency of IXFR responses
   can also depend greatly on if and how the zone is DNSSEC signed.

   In order to guarantee support for future padding policies, it is
   stated here that secondary implementations MUST be resilient to
   receiving padded IXoT responses.

   Recommendation of specific policies for padding IXoT responses are
   out of scope for this specification.  Detailed considerations of such
   padding policies, the use of traffic obfuscation techniques (such as
   generating fake XFR traffic), and the trade-offs involved are
   expected to be the subject of future work.

7.11.  Name Compression and Maximum Payload Sizes



   It is noted here that name compression [RFC1035] can be used in XFR
   responses to reduce the size of the payload; however, the maximum
   value of the offset that can be used in the name compression pointer
   structure is 16384.  For some DNS implementations, this limits the
   size of an individual XFR response used in practice to something
   around the order of 16 KB.  In principle, larger payload sizes can be
   supported for some responses with more sophisticated approaches
   (e.g., by precalculating the maximum offset required).

   Implementations may wish to offer options to disable name compression
   for XoT responses to enable larger payloads.  This might be
   particularly helpful when padding is used, since minimizing the
   payload size is not necessarily a useful optimization in this case
   and disabling name compression will reduce the resources required to
   construct the payload.

8.  Multi-primary Configurations



   This model can provide flexibility and redundancy, particularly for
   IXFR.  A secondary will receive one or more NOTIFY messages and can
   send an SOA to all of the configured primaries.  It can then choose
   to send an XFR request to the primary with the highest SOA (or based
   on other criteria, e.g., RTT).

   When using persistent connections, the secondary may have a XoT
   connection already open to one or more primaries.  Should a secondary
   preferentially request an XFR from a primary to which it already has
   an open XoT connection or the one with the highest SOA (assuming it
   doesn't have a connection open to it already)?

   Two extremes can be envisaged here.  The first one can be considered
   a 'preferred primary connection' model.  In this case, the secondary
   continues to use one persistent connection to a single primary until
   it has reason not to.  Reasons not to might include the primary
   repeatedly closing the connection, long query/response RTTs on
   transfers, or the SOA of the primary being an unacceptable lag behind
   the SOA of an alternative primary.

   The other extreme can be considered a 'parallel primary connection'
   model.  Here, a secondary could keep multiple persistent connections
   open to all available primaries and only request XFRs from the
   primary with the highest serial number.  Since normally the number of
   secondaries and primaries in direct contact in a transfer group is
   reasonably low, this might be feasible if latency is the most
   significant concern.

   Recommendation of a particular scheme is out of scope of this
   document, but implementations are encouraged to provide configuration
   options that allow operators to make choices about this behavior.

9.  Authentication Mechanisms



   To provide context to the requirements in Section 7.5, this section
   provides a brief summary of some of the existing authentication and
   validation mechanisms (both transport independent and TLS specific)
   that are available when performing zone transfers.  Section 10 then
   discusses in more detail specifically how a combination of TLS
   authentication, TSIG, and IP-based ACLs interact for XoT.

   In this document, the mechanisms are classified based on the
   following properties:

   Data Origin Authentication (DO):
      Authentication 1) of the fact that the DNS message originated from
      the party with whom credentials were shared and 2) of the data
      integrity of the message contents (the originating party may or
      may not be the party operating the far end of a TCP/TLS connection
      in a 'proxy' scenario).

   Channel Confidentiality (CC):
      Confidentiality of the communication channel between the client
      and server (i.e., the two endpoints of a TCP/TLS connection) from
      passive surveillance.

   Channel Authentication (CA):
      Authentication of the identity of the party to whom a TCP/TLS
      connection is made (this might not be a direct connection between
      the primary and secondary in a proxy scenario).

9.1.  TSIG



   TSIG [RFC8945] provides a mechanism for two or more parties to use
   shared secret keys that can then be used to create a message digest
   to protect individual DNS messages.  This allows each party to
   authenticate that a request or response (and the data in it) came
   from the other party, even if it was transmitted over an unsecured
   channel or via a proxy.

   Properties:  Data origin authentication.

9.2.  SIG(0)



   SIG(0) [RFC2931] similarly provides a mechanism to digitally sign a
   DNS message but uses public key authentication, where the public keys
   are stored in DNS as KEY RRs and a private key is stored at the
   signer.

   Properties:  Data origin authentication.

9.3.  TLS



9.3.1.  Opportunistic TLS



   Opportunistic TLS for DoT is defined in [RFC8310] and can provide a
   defense against passive surveillance, providing on-the-wire
   confidentiality.  Essentially:

   *  if clients know authentication information for a server, they
      SHOULD try to authenticate the server,

   *  if this fails or clients do not know the information, they MAY
      fallback to using TLS without authentication, or

   *  clients MAY fallback to using cleartext if TLS is not available.

   As such, it does not offer a defense against active attacks (e.g., an
   on-path active attacker on the connection from client to server) and
   is not considered as useful for XoT.

   Properties:  None guaranteed.

9.3.2.  Strict TLS



   Strict TLS for DoT [RFC8310] requires that a client is configured
   with an authentication domain name (and/or Subject Public Key Info
   (SPKI) pin set) that MUST be used to authenticate the TLS handshake
   with the server.  If authentication of the server fails, the client
   will not proceed with the connection.  This provides a defense for
   the client against active surveillance, providing client-to-server
   authentication and end-to-end channel confidentiality.

   Properties:  Channel confidentiality and channel authentication (of
      the server).

9.3.3.  Mutual TLS



   This is an extension to Strict TLS [RFC8310] that requires that a
   client is configured with an authentication domain name (and/or SPKI
   pin set) and a client certificate.  The client offers the certificate
   for authentication by the server, and the client can authenticate the
   server the same way as in Strict TLS.  This provides a defense for
   both parties against active surveillance, providing bidirectional
   authentication and end-to-end channel confidentiality.

   Properties:  Channel confidentiality and mutual channel
      authentication.

9.4.  IP-Based ACL on the Primary



   Most DNS server implementations offer an option to configure an IP-
   based ACL, which is often used in combination with TSIG-based ACLs to
   restrict access to zone transfers on primary servers on a per-query
   basis.

   This is also possible with XoT, but it must be noted that, as with
   TCP, the implementation of such an ACL cannot be enforced on the
   primary until an XFR request is received on an established
   connection.

   As discussed in Appendix A, an IP-based per-connection ACL could also
   be implemented where only TLS connections from recognized secondaries
   are accepted.

   Properties:  Channel authentication of the client.

9.5.  ZONEMD



   For completeness, ZONEMD [RFC8976] ("Message Digest for DNS Zones")
   is described here.  The ZONEMD message digest is a mechanism that can
   be used to verify the content of a standalone zone.  It is designed
   to be independent of the transmission channel or mechanism, allowing
   a general consumer of a zone to do origin authentication of the
   entire zone contents.  Note that the current version of [RFC8976]
   states:

   |  As specified herein, ZONEMD is impractical for large, dynamic
   |  zones due to the time and resources required for digest
   |  calculation.  However, the ZONEMD record is extensible so that new
   |  digest schemes may be added in the future to support large,
   |  dynamic zones.

   It is complementary but orthogonal to the above mechanisms and can be
   used in conjunction with XoT but is not considered further here.

10.  XoT Authentication



   It is noted that zone transfer scenarios can vary from a simple
   single primary/secondary relationship where both servers are under
   the control of a single operator to a complex hierarchical structure
   that includes proxies and multiple operators.  Each deployment
   scenario will require specific analysis to determine which
   combination of authentication methods are best suited to the
   deployment model in question.

   The XoT authentication requirement specified in Section 7.5 addresses
   the issue of ensuring that the transfers are encrypted between the
   two endpoints directly involved in the current transfers.  The
   following table summarizes the properties of a selection of the
   mechanisms discussed in Section 9.  The two-letter abbreviations for
   the properties are used below: (S) indicates the secondary and (P)
   indicates the primary.

    +================+=======+=======+=======+=======+=======+=======+
    | Method         | DO(S) | CC(S) | CA(S) | DO(P) | CC(P) | CA(P) |
    +================+=======+=======+=======+=======+=======+=======+
    | Strict TLS     |       |   Y   |   Y   |       |   Y   |       |
    +----------------+-------+-------+-------+-------+-------+-------+
    | Mutual TLS     |       |   Y   |   Y   |       |   Y   |   Y   |
    +----------------+-------+-------+-------+-------+-------+-------+
    | ACL on primary |       |       |       |       |       |   Y   |
    +----------------+-------+-------+-------+-------+-------+-------+
    | TSIG           |   Y   |       |       |   Y   |       |       |
    +----------------+-------+-------+-------+-------+-------+-------+

          Table 1: Properties of Authentication Methods for XoT

   Based on this analysis, it can be seen that:

   *  Using just mutual TLS can be considered a standalone solution
      since both endpoints are cryptographically authenticated.

   *  Using secondary-side Strict TLS with a primary-side IP-based ACL
      and TSIG/SIG(0) combination provides sufficient protection to be
      acceptable.

   Using just an IP-based ACL could be susceptible to attacks that can
   spoof TCP IP addresses; using TSIG/SIG(0) alone could be susceptible
   to attacks that were able to capture such messages should they be
   accidentally sent in cleartext by any server with the key.

11.  Policies for Both AXoT and IXoT



   Whilst the protection of the zone contents in a transfer between two
   endpoints can be provided by the XoT protocol, the protection of all
   the transfers of a given zone requires operational administration and
   policy management.

   The entire group of servers involved in XFR for a particular set of
   zones (all the primaries and all the secondaries) is called the
   'transfer group'.

   In order to assure the confidentiality of the zone information, the
   entire transfer group MUST have a consistent policy of using XoT.  If
   any do not, this is a weak link for attackers to exploit.  For
   clarification, this means that within any transfer group both AXFRs
   and IXFRs for a zone MUST all use XoT.

   An individual zone transfer is not considered protected by XoT unless
   both the client and server are configured to use only XoT, and the
   overall zone transfer is not considered protected until all members
   of the transfer group are configured to use only XoT with all other
   transfers servers (see Section 12).

   A XoT policy MUST specify if:

   *  mutual TLS is used and/or

   *  an IP-based ACL and TSIG/SIG(0) combination is used.

   Since this may require configuration of a number of servers who may
   be under the control of different operators, the desired consistency
   could be hard to enforce and audit in practice.

   Certain aspects of the policies can be relatively easy to test
   independently, e.g., by requesting zone transfers without TSIG, from
   unauthorized IP addresses or over cleartext DNS.  Other aspects, such
   as if a secondary will accept data without a TSIG digest or if
   secondaries are using Strict as opposed to Opportunistic TLS, are
   more challenging.

   The mechanics of coordinating or enforcing such policies are out of
   the scope of this document but may be the subject of future
   operational guidance.

12.  Implementation Considerations



   Server implementations may want to also offer options that allow ACLs
   on a zone to specify that a specific client can use either XoT or
   TCP.  This would allow for flexibility while clients are migrating to
   XoT.

   Client implementations may similarly want to offer options to cater
   to the multi-primary case where the primaries are migrating to XoT.

13.  Operational Considerations



   If the options described in Section 12 are available, such
   configuration options MUST only be used in a 'migration mode' and
   therefore should be used with great care.

   It is noted that use of a TLS proxy in front of the primary server is
   a simple deployment solution that can enable server-side XoT.

14.  IANA Considerations



   This document has no IANA actions.

15.  Security Considerations



   This document specifies a security measure against a DNS risk: the
   risk that an attacker collects entire DNS zones through eavesdropping
   on cleartext DNS zone transfers.

   This does not mitigate:

   *  the risk that some level of zone activity might be inferred by
      observing zone transfer sizes and timing on encrypted connections
      (even with padding applied), in combination with obtaining SOA
      records by directly querying authoritative servers,

   *  the risk that hidden primaries might be inferred or identified via
      observation of encrypted connections, or

   *  the risk of zone contents being obtained via zone enumeration
      techniques.

   Security concerns of DoT are outlined in [RFC7858] and [RFC8310].

16.  References



16.1.  Normative References



   [DoT-ALPN] IANA, "TLS Application-Layer Protocol Negotiation (ALPN)
              Protocol IDs", <https://www.iana.org/assignments/tls-
              extensiontype-values/>.

   [RFC1034]  Mockapetris, P., "Domain names - concepts and facilities",
              STD 13, RFC 1034, DOI 10.17487/RFC1034, November 1987,
              <https://www.rfc-editor.org/info/rfc1034>.

   [RFC1035]  Mockapetris, P., "Domain names - implementation and
              specification", STD 13, RFC 1035, DOI 10.17487/RFC1035,
              November 1987, <https://www.rfc-editor.org/info/rfc1035>.

   [RFC1995]  Ohta, M., "Incremental Zone Transfer in DNS", RFC 1995,
              DOI 10.17487/RFC1995, August 1996,
              <https://www.rfc-editor.org/info/rfc1995>.

   [RFC1996]  Vixie, P., "A Mechanism for Prompt Notification of Zone
              Changes (DNS NOTIFY)", RFC 1996, DOI 10.17487/RFC1996,
              August 1996, <https://www.rfc-editor.org/info/rfc1996>.

   [RFC2119]  Bradner, S., "Key words for use in RFCs to Indicate
              Requirement Levels", BCP 14, RFC 2119,
              DOI 10.17487/RFC2119, March 1997,
              <https://www.rfc-editor.org/info/rfc2119>.

   [RFC2931]  Eastlake 3rd, D., "DNS Request and Transaction Signatures
              ( SIG(0)s )", RFC 2931, DOI 10.17487/RFC2931, September
              2000, <https://www.rfc-editor.org/info/rfc2931>.

   [RFC5936]  Lewis, E. and A. Hoenes, Ed., "DNS Zone Transfer Protocol
              (AXFR)", RFC 5936, DOI 10.17487/RFC5936, June 2010,
              <https://www.rfc-editor.org/info/rfc5936>.

   [RFC6973]  Cooper, A., Tschofenig, H., Aboba, B., Peterson, J.,
              Morris, J., Hansen, M., and R. Smith, "Privacy
              Considerations for Internet Protocols", RFC 6973,
              DOI 10.17487/RFC6973, July 2013,
              <https://www.rfc-editor.org/info/rfc6973>.

   [RFC7766]  Dickinson, J., Dickinson, S., Bellis, R., Mankin, A., and
              D. Wessels, "DNS Transport over TCP - Implementation
              Requirements", RFC 7766, DOI 10.17487/RFC7766, March 2016,
              <https://www.rfc-editor.org/info/rfc7766>.

   [RFC7828]  Wouters, P., Abley, J., Dickinson, S., and R. Bellis, "The
              edns-tcp-keepalive EDNS0 Option", RFC 7828,
              DOI 10.17487/RFC7828, April 2016,
              <https://www.rfc-editor.org/info/rfc7828>.

   [RFC7858]  Hu, Z., Zhu, L., Heidemann, J., Mankin, A., Wessels, D.,
              and P. Hoffman, "Specification for DNS over Transport
              Layer Security (TLS)", RFC 7858, DOI 10.17487/RFC7858, May
              2016, <https://www.rfc-editor.org/info/rfc7858>.

   [RFC8174]  Leiba, B., "Ambiguity of Uppercase vs Lowercase in RFC
              2119 Key Words", BCP 14, RFC 8174, DOI 10.17487/RFC8174,
              May 2017, <https://www.rfc-editor.org/info/rfc8174>.

   [RFC8310]  Dickinson, S., Gillmor, D., and T. Reddy, "Usage Profiles
              for DNS over TLS and DNS over DTLS", RFC 8310,
              DOI 10.17487/RFC8310, March 2018,
              <https://www.rfc-editor.org/info/rfc8310>.

   [RFC8446]  Rescorla, E., "The Transport Layer Security (TLS) Protocol
              Version 1.3", RFC 8446, DOI 10.17487/RFC8446, August 2018,
              <https://www.rfc-editor.org/info/rfc8446>.

   [RFC8499]  Hoffman, P., Sullivan, A., and K. Fujiwara, "DNS
              Terminology", BCP 219, RFC 8499, DOI 10.17487/RFC8499,
              January 2019, <https://www.rfc-editor.org/info/rfc8499>.

   [RFC8914]  Kumari, W., Hunt, E., Arends, R., Hardaker, W., and D.
              Lawrence, "Extended DNS Errors", RFC 8914,
              DOI 10.17487/RFC8914, October 2020,
              <https://www.rfc-editor.org/info/rfc8914>.

   [RFC8945]  Dupont, F., Morris, S., Vixie, P., Eastlake 3rd, D.,
              Gudmundsson, O., and B. Wellington, "Secret Key
              Transaction Authentication for DNS (TSIG)", STD 93,
              RFC 8945, DOI 10.17487/RFC8945, November 2020,
              <https://www.rfc-editor.org/info/rfc8945>.

16.2.  Informative References



   [BIND]     ISC, "BIND 9.16.16", <https://www.isc.org/bind/>.

   [DPRIVE-DNSOQUIC]
              Huitema, C., Dickinson, S., and A. Mankin, "Specification
              of DNS over Dedicated QUIC Connections", Work in Progress,
              Internet-Draft, draft-ietf-dprive-dnsoquic-03, 12 July
              2021, <https://datatracker.ietf.org/doc/html/draft-ietf-
              dprive-dnsoquic-03>.

   [NIST-GUIDE]
              Chandramouli, R. and S. Rose, "Secure Domain Name System
              (DNS) Deployment Guide", September 2013,
              <https://nvlpubs.nist.gov/nistpubs/SpecialPublications/
              NIST.SP.800-81-2.pdf>.

   [NSD]      NLnet Labs, "NSD 4.3.6",
              <https://www.nlnetlabs.nl/projects/nsd/about/>.

   [NSEC3-attacks]
              Goldberg, S., Naor, N., Papadopoulos, D., Reyzin, L.,
              Vasant, S., and A. Ziv, "Stretching NSEC3 to the Limit:
              Efficient Zone Enumeration Attacks on NSEC3 Variants",
              February 2015,
              <https://www.cs.bu.edu/~goldbe/papers/nsec3attacks.pdf>.

   [NSEC5]    Vcelak, J., Goldberg, S., Papadopoulos, D., Huque, S., and
              D. Lawrence, "NSEC5, DNSSEC Authenticated Denial of
              Existence", Work in Progress, Internet-Draft, draft-
              vcelak-nsec5-08, 29 December 2018,
              <https://datatracker.ietf.org/doc/html/draft-vcelak-
              nsec5-08>.

   [RFC1982]  Elz, R. and R. Bush, "Serial Number Arithmetic", RFC 1982,
              DOI 10.17487/RFC1982, August 1996,
              <https://www.rfc-editor.org/info/rfc1982>.

   [RFC5155]  Laurie, B., Sisson, G., Arends, R., and D. Blacka, "DNS
              Security (DNSSEC) Hashed Authenticated Denial of
              Existence", RFC 5155, DOI 10.17487/RFC5155, March 2008,
              <https://www.rfc-editor.org/info/rfc5155>.

   [RFC6891]  Damas, J., Graff, M., and P. Vixie, "Extension Mechanisms
              for DNS (EDNS(0))", STD 75, RFC 6891,
              DOI 10.17487/RFC6891, April 2013,
              <https://www.rfc-editor.org/info/rfc6891>.

   [RFC8484]  Hoffman, P. and P. McManus, "DNS Queries over HTTPS
              (DoH)", RFC 8484, DOI 10.17487/RFC8484, October 2018,
              <https://www.rfc-editor.org/info/rfc8484>.

   [RFC8976]  Wessels, D., Barber, P., Weinberg, M., Kumari, W., and W.
              Hardaker, "Message Digest for DNS Zones", RFC 8976,
              DOI 10.17487/RFC8976, February 2021,
              <https://www.rfc-editor.org/info/rfc8976>.

   [RFC9076]  Wicinski, T., Ed., "DNS Privacy Considerations", RFC 9076,
              DOI 10.17487/RFC9076, July 2021,
              <https://www.rfc-editor.org/info/rfc9076>.

   [TLS-ESNI] Rescorla, E., Oku, K., Sullivan, N., and C. A. Wood, "TLS
              Encrypted Client Hello", Work in Progress, Internet-Draft,
              draft-ietf-tls-esni-13, 12 August 2021,
              <https://datatracker.ietf.org/doc/html/draft-ietf-tls-
              esni-13>.

Appendix A.  XoT Server Connection Handling



   This appendix provides a non-normative outline of the pros and cons
   of XoT server connection-handling options.

   For completeness, it is noted that an earlier draft version of this
   document suggested using a XoT-specific ALPN to negotiate TLS
   connections that supported only a limited set of queries (SOA, XFRs);
   however, this did not gain support.  Reasons given included
   additional code complexity and the fact that XoT and ADoT are both
   DNS wire format and so should share the "dot" ALPN.

A.1.  Listening Only on a Specific IP Address for TLS



   Obviously, a name server that hosts a zone and services queries for
   the zone on an IP address published in an NS record may wish to use a
   separate IP address for XoT to listen for TLS, only publishing that
   address to its secondaries.

   Pros:  Probing of the public IP address will show no support for TLS.
      ACLs will prevent zone transfer on all transports on a per-query
      basis.

   Cons:  Attackers passively observing traffic will still be able to
      observe TLS connections to the separate address.

A.2.  Client-Specific TLS Acceptance



   Primaries that include IP-based ACLs and/or mutual TLS in their
   authentication models have the option of only accepting TLS
   connections from authorized clients.  This could be implemented
   either using a proxy or directly in the DNS implementation.

   Pros:  Connection management happens at setup time.  The maximum
      number of TLS connections a server will have to support can be
      easily assessed.  Once the connection is accepted, the server
      might well be willing to answer any query on that connection since
      it is coming from a configured secondary, and a specific response
      policy on the connection may not be needed (see below).

   Cons:  Currently, none of the major open-source implementations of a
      DNS authoritative server support such an option.

A.3.  SNI-Based TLS Acceptance



   Primaries could also choose to only accept TLS connections based on a
   Server Name Indication (SNI) that was published only to their
   secondaries.

   Pros:  Reduces the number of accepted connections.

   Cons:  As above.  Also, this is not a recommended use of SNI.  For
      SNIs sent in the clear, this would still allow attackers passively
      observing traffic to potentially abuse this mechanism.  The use of
      Encrypted Client Hello [TLS-ESNI] may be of use here.

A.4.  Transport-Specific Response Policies



   Some primaries might rely on TSIG/SIG(0) combined with per-query, IP-
   based ACLs to authenticate secondaries.  In this case, the primary
   must accept all incoming TLS/TCP connections and then apply a
   transport-specific response policy on a per-query basis.

   As an aside, whilst [RFC7766] makes a general purpose distinction in
   the advice to clients about their usage of connections (between
   regular queries and zone transfers), this is not strict, and nothing
   in the DNS protocol prevents using the same connection for both types
   of traffic.  Hence, a server cannot know the intention of any client
   that connects to it; it can only inspect the messages it receives on
   such a connection and make per-query decisions about whether or not
   to answer those queries.

   Example policies a XoT server might implement are:

   strict:     REFUSE all queries on TLS connections, except SOA and
               authorized XFR requests

   moderate:   REFUSE all queries on TLS connections until one is
               received that is signed by a recognized TSIG/SIG(0) key,
               then answer all queries on the connection after that

   complex:    apply a heuristic to determine which queries on a TLS
               connections to REFUSE

   relaxed:    answer all non-XoT queries on all TLS connections with
               the same policy applied to TCP queries

   Pros:  Allows for flexible behavior by the server that could be
      changed over time.

   Cons:  The server must handle the burden of accepting all TLS
      connections just to perform XFRs with a small number of
      secondaries.  Client behavior to a REFUSED response is not clearly
      defined (see Section 7.8).  Currently, none of the major open-
      source implementations of a DNS authoritative server offer an
      option for different response policies in different transports
      (but such functionality could potentially be implemented using a
      proxy).

A.4.1.  SNI-Based Response Policies



   In a similar fashion, XoT servers might use the presence of an SNI in
   the Client Hello to determine which response policy to initially
   apply to the TLS connections.

   Pros:  This has the potential to allow a clean distinction between a
      XoT service and any future DoT-based service for answering
      recursive queries.

   Cons:  As above.

Acknowledgements



   The authors thank Tony Finch, Benno Overeinder, Shumon Huque, Tim
   Wicinski, and many other members of DPRIVE for review and
   discussions.

   The authors particularly thank Peter van Dijk, Ondrej Sury, Brian
   Dickson, and several other open-source DNS implementers for valuable
   discussion and clarification on the issue associated with pipelining
   XFR queries and handling out-of-order/intermingled responses.

Contributors

   Significant contributions to the document were made by:

   Han Zhang
   Salesforce
   San Francisco, CA
   United States of America

   Email: hzhang@salesforce.com


Authors' Addresses



   Willem Toorop
   NLnet Labs
   Science Park 400
   1098 XH Amsterdam
   Netherlands

   Email: willem@nlnetlabs.nl


   Sara Dickinson
   Sinodun IT
   Magdalen Centre
   Oxford Science Park
   Oxford
   OX4 4GA
   United Kingdom

   Email: sara@sinodun.com


   Shivan Sahib
   Brave Software
   Vancouver BC
   Canada

   Email: shivankaulsahib@gmail.com


   Pallavi Aras
   Salesforce
   Herndon, VA
   United States of America

   Email: paras@salesforce.com


   Allison Mankin
   Salesforce
   Herndon, VA
   United States of America

   Email: allison.mankin@gmail.com