RFC 9250




Internet Engineering Task Force (IETF)                        C. Huitema
Request for Comments: 9250                          Private Octopus Inc.
Category: Standards Track                                   S. Dickinson
ISSN: 2070-1721                                               Sinodun IT
                                                               A. Mankin
                                                              Salesforce
                                                                May 2022


                  DNS over Dedicated QUIC Connections

Abstract



   This document describes the use of QUIC to provide transport
   confidentiality for DNS.  The encryption provided by QUIC has similar
   properties to those provided by TLS, while QUIC transport eliminates
   the head-of-line blocking issues inherent with TCP and provides more
   efficient packet-loss recovery than UDP.  DNS over QUIC (DoQ) has
   privacy properties similar to DNS over TLS (DoT) specified in RFC
   7858, and latency characteristics similar to classic DNS over UDP.
   This specification describes the use of DoQ as a general-purpose
   transport for DNS and includes the use of DoQ for stub to recursive,
   recursive to authoritative, and zone transfer scenarios.

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/rfc9250.

Copyright Notice



   Copyright (c) 2022 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 Revised BSD License text as described in Section 4.e of the
   Trust Legal Provisions and are provided without warranty as described
   in the Revised BSD License.

Table of Contents



   1.  Introduction
   2.  Key Words
   3.  Design Considerations
     3.1.  Provide DNS Privacy
     3.2.  Design for Minimum Latency
     3.3.  Middlebox Considerations
     3.4.  No Server-Initiated Transactions
   4.  Specifications
     4.1.  Connection Establishment
       4.1.1.  Port Selection
     4.2.  Stream Mapping and Usage
       4.2.1.  DNS Message IDs
     4.3.  DoQ Error Codes
       4.3.1.  Transaction Cancellation
       4.3.2.  Transaction Errors
       4.3.3.  Protocol Errors
       4.3.4.  Alternative Error Codes
     4.4.  Connection Management
     4.5.  Session Resumption and 0-RTT
     4.6.  Message Sizes
   5.  Implementation Requirements
     5.1.  Authentication
     5.2.  Fallback to Other Protocols on Connection Failure
     5.3.  Address Validation
     5.4.  Padding
     5.5.  Connection Handling
       5.5.1.  Connection Reuse
       5.5.2.  Resource Management
       5.5.3.  Using 0-RTT and Session Resumption
       5.5.4.  Controlling Connection Migration for Privacy
     5.6.  Processing Queries in Parallel
     5.7.  Zone Transfer
     5.8.  Flow Control Mechanisms
   6.  Security Considerations
   7.  Privacy Considerations
     7.1.  Privacy Issues with 0-RTT data
     7.2.  Privacy Issues with Session Resumption
     7.3.  Privacy Issues with Address Validation Tokens
     7.4.  Privacy Issues with Long Duration Sessions
     7.5.  Traffic Analysis
   8.  IANA Considerations
     8.1.  Registration of a DoQ Identification String
     8.2.  Reservation of a Dedicated Port
     8.3.  Reservation of an Extended DNS Error Code: Too Early
     8.4.  DNS-over-QUIC Error Codes Registry
   9.  References
     9.1.  Normative References
     9.2.  Informative References
   Appendix A.  The NOTIFY Service
   Acknowledgements

   Authors' Addresses



1.  Introduction



   Domain Name System (DNS) concepts are specified in "Domain names -
   concepts and facilities" [RFC1034].  The transmission of DNS queries
   and responses over UDP and TCP is specified in "Domain names -
   implementation and specification" [RFC1035].

   This document presents a mapping of the DNS protocol over the QUIC
   transport [RFC9000] [RFC9001].  DNS over QUIC is referred to here as
   DoQ, in line with "DNS Terminology" [DNS-TERMS].

   The goals of the DoQ mapping are:

   1.  Provide the same DNS privacy protection as DoT [RFC7858].  This
       includes an option for the client to authenticate the server by
       means of an authentication domain name as specified in "Usage
       Profiles for DNS over TLS and DNS over DTLS" [RFC8310].

   2.  Provide an improved level of source address validation for DNS
       servers compared to classic DNS over UDP.



   3.  Provide a transport that does not impose path MTU limitations on
       the size of DNS responses it can send.



   In order to achieve these goals, and to support ongoing work on
   encryption of DNS, the scope of this document includes:

   *  the "stub to recursive resolver" scenario (also called the "stub
      to recursive" scenario in this document)

   *  the "recursive resolver to authoritative nameserver" scenario
      (also called the "recursive to authoritative" scenario in this
      document), and

   *  the "nameserver to nameserver" scenario (mainly used for zone
      transfers (XFR) [RFC1995] [RFC5936]).

   In other words, this document specifies QUIC as a general-purpose
   transport for DNS.

   The specific non-goals of this document are:

   1.  No attempt is made to evade potential blocking of DoQ traffic by
       middleboxes.

   2.  No attempt to support server-initiated transactions, which are
       used only in DNS Stateful Operations (DSO) [RFC8490].

   Specifying the transmission of an application over QUIC requires
   specifying how the application's messages are mapped to QUIC streams,
   and generally how the application will use QUIC.  This is done for
   HTTP in "Hypertext Transfer Protocol Version 3 (HTTP/3)" [HTTP/3].
   The purpose of this document is to define the way DNS messages can be
   transmitted over QUIC.

   DNS over HTTPS (DoH) [RFC8484] can be used with HTTP/3 to get some of
   the benefits of QUIC.  However, a lightweight direct mapping for DoQ
   can be regarded as a more natural fit for both the recursive to
   authoritative and zone transfer scenarios, which rarely involve
   intermediaries.  In these scenarios, the additional overhead of HTTP
   is not offset by, for example, benefits of HTTP proxying and caching
   behavior.

   In this document, Section 3 presents the reasoning that guided the
   proposed design.  Section 4 specifies the actual mapping of DoQ.
   Section 5 presents guidelines on the implementation, usage, and
   deployment of DoQ.

2.  Key Words

   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.

3.  Design Considerations

   This section and its subsections present the design guidelines that
   were used for DoQ.  While all other sections in this document are
   normative, this section is informative in nature.

3.1.  Provide DNS Privacy



   DoT [RFC7858] defines how to mitigate some of the issues described in
   "DNS Privacy Considerations" [RFC9076] by specifying how to transmit
   DNS messages over TLS.  The "Usage Profiles for DNS over TLS and DNS
   over DTLS" [RFC8310] specify Strict and Opportunistic usage profiles
   for DoT including how stub resolvers can authenticate recursive
   resolvers.

   QUIC connection setup includes the negotiation of security parameters
   using TLS, as specified in "Using TLS to Secure QUIC" [RFC9001],
   enabling encryption of the QUIC transport.  Transmitting DNS messages
   over QUIC will provide essentially the same privacy protections as
   DoT [RFC7858] including Strict and Opportunistic usage profiles
   [RFC8310].  Further discussion on this is provided in Section 7.

3.2.  Design for Minimum Latency



   QUIC is specifically designed to reduce protocol-induced delays, with
   features such as:

   1.  Support for 0-RTT data during session resumption.

   2.  Support for advanced packet-loss recovery procedures as specified
       in "QUIC Loss Detection and Congestion Control" [RFC9002].

   3.  Mitigation of head-of-line blocking by allowing parallel delivery
       of data on multiple streams.

   This mapping of DNS to QUIC will take advantage of these features in
   three ways:

   1.  Optional support for sending 0-RTT data during session resumption
       (the security and privacy implications of this are discussed in
       later sections).

   2.  Long-lived QUIC connections over which multiple DNS transactions
       are performed, generating the sustained traffic required to
       benefit from advanced recovery features.

   3.  Mapping of each DNS Query/Response transaction to a separate
       stream, to mitigate head-of-line blocking.  This enables servers
       to respond to queries "out of order".  It also enables clients to
       process responses as soon as they arrive, without having to wait
       for in-order delivery of responses previously posted by the
       server.

   These considerations are reflected in the mapping of DNS traffic to
   QUIC streams in Section 4.2.

3.3.  Middlebox Considerations



   Using QUIC might allow a protocol to disguise its purpose from
   devices on the network path using encryption and traffic analysis
   resistance techniques like padding, traffic pacing, and traffic
   shaping.  This specification does not include any measures that are
   designed to avoid such classification; the padding mechanisms defined
   in Section 5.4 are intended to obfuscate the specific records
   contained in DNS queries and responses, but not the fact that this is
   DNS traffic.  Consequently, firewalls and other middleboxes might be
   able to distinguish DoQ from other protocols that use QUIC, like
   HTTP, and apply different treatment.

   The lack of measures in this specification to avoid protocol
   classification is not an endorsement of such practices.

3.4.  No Server-Initiated Transactions



   As stated in Section 1, this document does not specify support for
   server-initiated transactions within established DoQ connections.
   That is, only the initiator of the DoQ connection may send queries
   over the connection.

   DSO does support server-initiated transactions within existing
   connections.  However, DoQ as defined here does not meet the criteria
   for an applicable transport for DSO because it does not guarantee in-
   order delivery of messages; see Section 4.2 of [RFC8490].

4.  Specifications



4.1.  Connection Establishment



   DoQ connections are established as described in the QUIC transport
   specification [RFC9000].  During connection establishment, DoQ
   support is indicated by selecting the Application-Layer Protocol
   Negotiation (ALPN) token "doq" in the crypto handshake.

4.1.1.  Port Selection



   By default, a DNS server that supports DoQ MUST listen for and accept
   QUIC connections on the dedicated UDP port 853 (Section 8), unless
   there is a mutual agreement to use another port.

   By default, a DNS client desiring to use DoQ with a particular server
   MUST establish a QUIC connection to UDP port 853 on the server,
   unless there is a mutual agreement to use another port.

   DoQ connections MUST NOT use UDP port 53.  This recommendation
   against use of port 53 for DoQ is to avoid confusion between DoQ and
   the use of DNS over UDP [RFC1035].  The risk of confusion exists even
   if two parties agreed on port 53, as other parties without knowledge
   of that agreement might still try to use that port.

   In the stub to recursive scenario, the use of port 443 as a mutually
   agreed alternative port can be operationally beneficial, since port
   443 is used by many services using QUIC and HTTP-3 and is thus less
   likely to be blocked than other ports.  Several mechanisms for stubs
   to discover recursives offering encrypted transports, including the
   use of custom ports, are the subject of ongoing work.

4.2.  Stream Mapping and Usage



   The mapping of DNS traffic over QUIC streams takes advantage of the
   QUIC stream features detailed in Section 2 of [RFC9000], the QUIC
   transport specification.

   DNS query/response traffic [RFC1034] [RFC1035] follows a simple
   pattern in which the client sends a query, and the server provides
   one or more responses (multiple responses can occur in zone
   transfers).

   The mapping specified here requires that the client select a separate
   QUIC stream for each query.  The server then uses the same stream to
   provide all the response messages for that query.  In order for
   multiple responses to be parsed, a 2-octet length field is used in
   exactly the same way as the 2-octet length field defined for DNS over
   TCP [RFC1035].  The practical result of this is that the content of
   each QUIC stream is exactly the same as the content of a TCP
   connection that would manage exactly one query.

   All DNS messages (queries and responses) sent over DoQ connections
   MUST be encoded as a 2-octet length field followed by the message
   content as specified in [RFC1035].

   The client MUST select the next available client-initiated
   bidirectional stream for each subsequent query on a QUIC connection,
   in conformance with the QUIC transport specification [RFC9000].
   Packet losses and other network events might cause queries to arrive
   in a different order.  Servers SHOULD process queries as they arrive,
   as not doing so would cause unnecessary delays.

   The client MUST send the DNS query over the selected stream and MUST
   indicate through the STREAM FIN mechanism that no further data will
   be sent on that stream.

   The server MUST send the response(s) on the same stream and MUST
   indicate, after the last response, through the STREAM FIN mechanism
   that no further data will be sent on that stream.

   Therefore, a single DNS transaction consumes a single bidirectional
   client-initiated stream.  This means that the client's first query
   occurs on QUIC stream 0, the second on 4, and so on (see Section 2.1
   of [RFC9000]).

   Servers MAY defer processing of a query until the STREAM FIN has been
   indicated on the stream selected by the client.

   Servers and clients MAY monitor the number of "dangling" streams.
   These are open streams where the following events have not occurred
   after implementation-defined timeouts:

   *  the expected queries or responses have not been received or,

   *  the expected queries or responses have been received but not the
      STREAM FIN

   Implementations MAY impose a limit on the number of such dangling
   streams.  If limits are encountered, implementations MAY close the
   connection.

4.2.1.  DNS Message IDs



   When sending queries over a QUIC connection, the DNS Message ID MUST
   be set to 0.  The stream mapping for DoQ allows for unambiguous
   correlation of queries and responses, so the Message ID field is not
   required.

   This has implications for proxying DoQ messages to and from other
   transports.  For example, proxies may have to manage the fact that
   DoQ can support a larger number of outstanding queries on a single
   connection than, for example, DNS over TCP, because DoQ is not
   limited by the Message ID space.  This issue already exists for DoH,
   where a Message ID of 0 is recommended.

   When forwarding a DNS message from DoQ over another transport, a DNS
   Message ID MUST be generated according to the rules of the protocol
   that is in use.  When forwarding a DNS message from another transport
   over DoQ, the Message ID MUST be set to 0.

4.3.  DoQ Error Codes



   The following error codes are defined for use when abruptly
   terminating streams, for use as application protocol error codes when
   aborting reading of streams, or for immediately closing connections:

   DOQ_NO_ERROR (0x0):  No error.  This is used when the connection or
      stream needs to be closed, but there is no error to signal.

   DOQ_INTERNAL_ERROR (0x1):  The DoQ implementation encountered an
      internal error and is incapable of pursuing the transaction or the
      connection.

   DOQ_PROTOCOL_ERROR (0x2):  The DoQ implementation encountered a
      protocol error and is forcibly aborting the connection.

   DOQ_REQUEST_CANCELLED (0x3):  A DoQ client uses this to signal that
      it wants to cancel an outstanding transaction.

   DOQ_EXCESSIVE_LOAD (0x4):  A DoQ implementation uses this to signal
      when closing a connection due to excessive load.

   DOQ_UNSPECIFIED_ERROR (0x5):  A DoQ implementation uses this in the
      absence of a more specific error code.

   DOQ_ERROR_RESERVED (0xd098ea5e):  An alternative error code used for
      tests.

   See Section 8.4 for details on registering new error codes.

4.3.1.  Transaction Cancellation



   In QUIC, sending STOP_SENDING requests that a peer cease transmission
   on a stream.  If a DoQ client wishes to cancel an outstanding
   request, it MUST issue a QUIC STOP_SENDING, and it SHOULD use the
   error code DOQ_REQUEST_CANCELLED.  It MAY use a more specific error
   code registered according to Section 8.4.  The STOP_SENDING request
   may be sent at any time but will have no effect if the server
   response has already been sent, in which case the client will simply
   discard the incoming response.  The corresponding DNS transaction
   MUST be abandoned.

   Servers that receive STOP_SENDING act in accordance with Section 3.5
   of [RFC9000].  Servers SHOULD NOT continue processing a DNS
   transaction if they receive a STOP_SENDING.

   Servers MAY impose implementation limits on the total number or rate
   of cancellation requests.  If limits are encountered, servers MAY
   close the connection.  In this case, servers wanting to help client
   debugging MAY use the error code DOQ_EXCESSIVE_LOAD.  There is always
   a trade-off between helping good faith clients debug issues and
   allowing denial-of-service attackers to test server defenses;
   depending on circumstances servers might very well choose to send
   different error codes.

   Note that this mechanism provides a way for secondaries to cancel a
   single zone transfer occurring on a given stream without having to
   close the QUIC connection.

   Servers MUST NOT continue processing a DNS transaction if they
   receive a RESET_STREAM request from the client before the client
   indicates the STREAM FIN.  The server MUST issue a RESET_STREAM to
   indicate that the transaction is abandoned unless:

   *  it has already done so for another reason or

   *  it has already both sent the response and indicated the STREAM
      FIN.

4.3.2.  Transaction Errors



   Servers normally complete transactions by sending a DNS response (or
   responses) on the transaction's stream, including cases where the DNS
   response indicates a DNS error.  For example, a client SHOULD be
   notified of a Server Failure (SERVFAIL, [RFC1035]) through a response
   with the Response Code set to SERVFAIL.

   If a server is incapable of sending a DNS response due to an internal
   error, it SHOULD issue a QUIC RESET_STREAM frame.  The error code
   SHOULD be set to DOQ_INTERNAL_ERROR.  The corresponding DNS
   transaction MUST be abandoned.  Clients MAY limit the number of
   unsolicited QUIC RESET_STREAM frames received on a connection before
   choosing to close the connection.

   Note that this mechanism provides a way for primaries to abort a
   single zone transfer occurring on a given stream without having to
   close the QUIC connection.

4.3.3.  Protocol Errors



   Other error scenarios can occur due to malformed, incomplete, or
   unexpected messages during a transaction.  These include (but are not
   limited to):

   *  a client or server receives a message with a non-zero Message ID

   *  a client or server receives a STREAM FIN before receiving all the
      bytes for a message indicated in the 2-octet length field

   *  a client receives a STREAM FIN before receiving all the expected
      responses

   *  a server receives more than one query on a stream

   *  a client receives a different number of responses on a stream than
      expected (e.g., multiple responses to a query for an A record)

   *  a client receives a STOP_SENDING request

   *  the client or server does not indicate the expected STREAM FIN
      after sending requests or responses (see Section 4.2)

   *  an implementation receives a message containing the edns-tcp-
      keepalive EDNS(0) Option [RFC7828] (see Section 5.5.2)

   *  a client or a server attempts to open a unidirectional QUIC stream

   *  a server attempts to open a server-initiated bidirectional QUIC
      stream

   *  a server receives a "replayable" transaction in 0-RTT data (for
      servers not willing to handle this case, see Section 4.5)

   If a peer encounters such an error condition, it is considered a
   fatal error.  It SHOULD forcibly abort the connection using QUIC's
   CONNECTION_CLOSE mechanism and SHOULD use the DoQ error code
   DOQ_PROTOCOL_ERROR.  In some cases, it MAY instead silently abandon
   the connection, which uses fewer of the local resources but makes
   debugging at the offending node more difficult.

   It is noted that the restrictions on use of the above EDNS(0) option
   has implications for proxying messages from TCP/DoT/DoH over DoQ.

4.3.4.  Alternative Error Codes



   This specification describes specific error codes in Sections 4.3.1,
   4.3.2, and 4.3.3.  These error codes are meant to facilitate
   investigation of failures and other incidents.  New error codes may
   be defined in future versions of DoQ or registered as specified in
   Section 8.4.

   Because new error codes can be defined without negotiation, use of an
   error code in an unexpected context or receipt of an unknown error
   code MUST be treated as equivalent to DOQ_UNSPECIFIED_ERROR.

   Implementations MAY wish to test the support for the error code
   extension mechanism by using error codes not listed in this document,
   or they MAY use DOQ_ERROR_RESERVED.

4.4.  Connection Management



   Section 10 of [RFC9000], the QUIC transport specification, specifies
   that connections can be closed in three ways:

   *  idle timeout

   *  immediate close

   *  stateless reset

   Clients and servers implementing DoQ SHOULD negotiate use of the idle
   timeout.  Closing on idle timeout is done without any packet
   exchange, which minimizes protocol overhead.  Per Section 10.1 of
   [RFC9000], the QUIC transport specification, the effective value of
   the idle timeout is computed as the minimum of the values advertised
   by the two endpoints.  Practical considerations on setting the idle
   timeout are discussed in Section 5.5.2.

   Clients SHOULD monitor the idle time incurred on their connection to
   the server, defined by the time spent since the last packet from the
   server has been received.  When a client prepares to send a new DNS
   query to the server, it SHOULD check whether the idle time is
   sufficiently lower than the idle timer.  If it is, the client SHOULD
   send the DNS query over the existing connection.  If not, the client
   SHOULD establish a new connection and send the query over that
   connection.

   Clients MAY discard their connections to the server before the idle
   timeout expires.  A client that has outstanding queries SHOULD close
   the connection explicitly using QUIC's CONNECTION_CLOSE mechanism and
   the DoQ error code DOQ_NO_ERROR.

   Clients and servers MAY close the connection for a variety of other
   reasons, indicated using QUIC's CONNECTION_CLOSE.  Client and servers
   that send packets over a connection discarded by their peer might
   receive a stateless reset indication.  If a connection fails, all the
   in-progress transactions on that connection MUST be abandoned.

4.5.  Session Resumption and 0-RTT



   A client MAY take advantage of the session resumption and 0-RTT
   mechanisms supported by QUIC transport [RFC9000] and QUIC TLS
   [RFC9001] if the server supports them.  Clients SHOULD consider
   potential privacy issues associated with session resumption before
   deciding to use this mechanism and specifically evaluate the trade-
   offs presented in the various sections of this document.  The privacy
   issues are detailed in Sections 7.1 and 7.2, and the implementation
   considerations are discussed in Section 5.5.3.

   The 0-RTT mechanism MUST NOT be used to send DNS requests that are
   not "replayable" transactions.  In this specification, only
   transactions that have an OPCODE of QUERY or NOTIFY are considered
   replayable; therefore, other OPCODES MUST NOT be sent in 0-RTT data.
   See Appendix A for a detailed discussion of why NOTIFY is included
   here.

   Servers MAY support session resumption, and MAY do that with or
   without supporting 0-RTT, using the mechanisms described in
   Section 4.6.1 of [RFC9001].  Servers supporting 0-RTT MUST NOT
   immediately process non-replayable transactions received in 0-RTT
   data but instead MUST adopt one of the following behaviors:

   *  Queue the offending transaction and only process it after the QUIC
      handshake has been completed, as defined in Section 4.1.1 of
      [RFC9001].

   *  Reply to the offending transaction with a response code REFUSED
      and an Extended DNS Error Code (EDE) "Too Early" using the
      extended RCODE mechanisms defined in [RFC6891] and the extended
      DNS errors defined in [RFC8914]; see Section 8.3.

   *  Close the connection with the error code DOQ_PROTOCOL_ERROR.

4.6.  Message Sizes



   DoQ queries and responses are sent on QUIC streams, which in theory
   can carry up to 2^62 bytes.  However, DNS messages are restricted in
   practice to a maximum size of 65535 bytes.  This maximum size is
   enforced by the use of a 2-octet message length field in DNS over TCP
   [RFC1035] and DoT [RFC7858], and by the definition of the
   "application/dns-message" for DoH [RFC8484].  DoQ enforces the same
   restriction.

   The Extension Mechanisms for DNS (EDNS(0)) [RFC6891] allow peers to
   specify the UDP message size.  This parameter is ignored by DoQ.  DoQ
   implementations always assume that the maximum message size is 65535
   bytes.

5.  Implementation Requirements



5.1.  Authentication



   For the stub to recursive scenario, the authentication requirements
   are the same as described in DoT [RFC7858] and "Usage Profiles for
   DNS over TLS and DNS over DTLS" [RFC8310].  [RFC8932] states that DNS
   privacy services SHOULD provide credentials that clients can use to
   authenticate the server.  Given this, and to align with the
   authentication model for DoH, DoQ stubs SHOULD use a Strict usage
   profile.  Client authentication for the encrypted stub to recursive
   scenario is not described in any DNS RFC.

   For zone transfer, the authentication requirements are the same as
   described in [RFC9103].

   For the recursive to authoritative scenario, authentication
   requirements are unspecified at the time of writing and are the
   subject of ongoing work in the DPRIVE WG.

5.2.  Fallback to Other Protocols on Connection Failure



   If the establishment of the DoQ connection fails, clients MAY attempt
   to fall back to DoT and then potentially cleartext, as specified in
   DoT [RFC7858] and "Usage Profiles for DNS over TLS and DNS over DTLS"
   [RFC8310], depending on their usage profile.

   DNS clients SHOULD remember server IP addresses that don't support
   DoQ.  Mobile clients might also remember the lack of DoQ support by
   given IP addresses on a per-context basis (e.g., per network or
   provisioning domain).

   Timeouts, connection refusals, and QUIC handshake failures are
   indicators that a server does not support DoQ.  Clients SHOULD NOT
   attempt DoQ queries to a server that does not support DoQ for a
   reasonable period (such as one hour per server).  DNS clients
   following an out-of-band key-pinned usage profile [RFC7858] MAY be
   more aggressive about retrying after DoQ connection failures.

5.3.  Address Validation



   Section 8 of [RFC9000], the QUIC transport specification, defines
   Address Validation procedures to avoid servers being used in address
   amplification attacks.  DoQ implementations MUST conform to this
   specification, which limits the worst-case amplification to a factor
   3.

   DoQ implementations SHOULD consider configuring servers to use the
   Address Validation using Retry Packets procedure defined in
   Section 8.1.2 of [RFC9000], the QUIC transport specification.  This
   procedure imposes a 1-RTT delay for verifying the return routability
   of the source address of a client, similar to the DNS Cookies
   mechanism [RFC7873].

   DoQ implementations that configure Address Validation using Retry
   Packets SHOULD implement the Address Validation for Future
   Connections procedure defined in Section 8.1.3 of [RFC9000], the QUIC
   transport specification.  This defines how servers can send NEW_TOKEN
   frames to clients after the client address is validated in order to
   avoid the 1-RTT penalty during subsequent connections by the client
   from the same address.

5.4.  Padding



   Implementations MUST protect against the traffic analysis attacks
   described in Section 7.5 by the judicious injection of padding.  This
   could be done either by padding individual DNS messages using the
   EDNS(0) Padding Option [RFC7830] or by padding QUIC packets (see
   Section 19.1 of [RFC9000]).

   In theory, padding at the QUIC packet level could result in better
   performance for the equivalent protection, because the amount of
   padding can take into account non-DNS frames such as acknowledgements
   or flow control updates, and also because QUIC packets can carry
   multiple DNS messages.  However, applications can only control the
   amount of padding in QUIC packets if the implementation of QUIC
   exposes adequate APIs.  This leads to the following recommendations:

   *  If the implementation of QUIC exposes APIs to set a padding
      policy, DoQ SHOULD use that API to align the packet length to a
      small set of fixed sizes.

   *  If padding at the QUIC packet level is not available or not used,
      DoQ MUST ensure that all DNS queries and responses are padded to a
      small set of fixed sizes, using the EDNS(0) padding extension as
      specified in [RFC7830].

   Implementations might choose not to use a QUIC API for padding if it
   is significantly simpler to reuse existing DNS message padding logic
   that is applied to other encrypted transports.

   In the absence of a standard policy for padding sizes,
   implementations SHOULD follow the recommendations of the Experimental
   status "Padding Policies for Extension Mechanisms for DNS (EDNS(0))"
   [RFC8467].  While Experimental, these recommendations are referenced
   because they are implemented and deployed for DoT and provide a way
   for implementations to be fully compliant with this specification.

5.5.  Connection Handling



   "DNS Transport over TCP - Implementation Requirements" [RFC7766]
   provides updated guidance on DNS over TCP, some of which is
   applicable to DoQ.  This section provides similar advice on
   connection handling for DoQ.

5.5.1.  Connection Reuse



   Historic implementations of DNS clients are known to open and close
   TCP connections for each DNS query.  To amortize connection setup
   costs, both clients and servers SHOULD support connection reuse by
   sending multiple queries and responses over a single persistent QUIC
   connection.

   In order to achieve performance on par with UDP, DNS clients SHOULD
   send their queries concurrently over the QUIC streams on a QUIC
   connection.  That is, when a DNS client sends multiple queries to a
   server over a QUIC connection, it SHOULD NOT wait for an outstanding
   reply before sending the next query.

5.5.2.  Resource Management



   Proper management of established and idle connections is important to
   the healthy operation of a DNS server.

   An implementation of DoQ SHOULD follow best practices similar to
   those specified for DNS over TCP [RFC7766], in particular with regard
   to:

   *  Concurrent Connections (Section 6.2.2 of [RFC7766], updated by
      Section 6.4 of [RFC9103])

   *  Security Considerations (Section 10 of [RFC7766])

   Failure to do so may lead to resource exhaustion and denial of
   service.

   Clients that want to maintain long duration DoQ connections SHOULD
   use the idle timeout mechanisms defined in Section 10.1 of [RFC9000],
   the QUIC transport specification.  Clients and servers MUST NOT send
   the edns-tcp-keepalive EDNS(0) Option [RFC7828] in any messages sent
   on a DoQ connection (because it is specific to the use of TCP/TLS as
   a transport).

   This document does not make specific recommendations for timeout
   values on idle connections.  Clients and servers should reuse and/or
   close connections depending on the level of available resources.
   Timeouts may be longer during periods of low activity and shorter
   during periods of high activity.

5.5.3.  Using 0-RTT and Session Resumption



   Using 0-RTT for DoQ has many compelling advantages.  Clients can
   establish connections and send queries without incurring a connection
   delay.  Servers can thus negotiate low values of the connection
   timers, which reduces the total number of connections that they need
   to manage.  They can do that because the clients that use 0-RTT will
   not incur latency penalties if new connections are required for a
   query.

   Session resumption and 0-RTT data transmission create privacy risks
   detailed in Sections 7.1 and 7.2.  The following recommendations are
   meant to reduce the privacy risks while enjoying the performance
   benefits of 0-RTT data, subject to the restrictions specified in
   Section 4.5.

   Clients SHOULD use resumption tickets only once, as specified in
   Appendix C.4 of [RFC8446].  By default, clients SHOULD NOT use
   session resumption if the client's connectivity has changed.

   Clients could receive address validation tokens from the server using
   the NEW_TOKEN mechanism; see Section 8 of [RFC9000].  The associated
   tracking risks are mentioned in Section 7.3.  Clients SHOULD only use
   the address validation tokens when they are also using session
   resumption thus avoiding additional tracking risks.

   Servers SHOULD issue session resumption tickets with a sufficiently
   long lifetime (e.g., 6 hours), so that clients are not tempted to
   either keep the connection alive or frequently poll the server to
   renew session resumption tickets.  Servers SHOULD implement the anti-
   replay mechanisms specified in Section 8 of [RFC8446].

5.5.4.  Controlling Connection Migration for Privacy



   DoQ implementations might consider using the connection migration
   features defined in Section 9 of [RFC9000].  These features enable
   connections to continue operating as the client's connectivity
   changes.  As detailed in Section 7.4, these features trade off
   privacy for latency.  By default, clients SHOULD be configured to
   prioritize privacy and start new sessions if their connectivity
   changes.

5.6.  Processing Queries in Parallel



   As specified in Section 7 of [RFC7766] "DNS Transport over TCP -
   Implementation Requirements", resolvers are RECOMMENDED to support
   the preparing of responses in parallel and sending them out of order.
   In DoQ, they do that by sending responses on their specific stream as
   soon as possible, without waiting for availability of responses for
   previously opened streams.

5.7.  Zone Transfer



   [RFC9103] specifies zone transfer over TLS (XoT) and includes updates
   to [RFC1995] (IXFR), [RFC5936] (AXFR), and [RFC7766].  Considerations
   relating to the reuse of XoT connections described there apply
   analogously to zone transfers performed using DoQ connections.  One
   reason for reiterating such specific guidance is the lack of
   effective connection reuse in existing TCP/TLS zone transfer
   implementations today.  The following recommendations apply:

   *  DoQ servers MUST be able to handle multiple concurrent IXFR
      requests on a single QUIC connection.

   *  DoQ servers MUST be able to handle multiple concurrent AXFR
      requests on a single QUIC connection.

   *  DoQ implementations SHOULD

      -  use the same QUIC connection for both AXFR and IXFR requests to
         the same primary

      -  send those requests in parallel as soon as they are queued,
         i.e., do not wait for a response before sending the next query
         on the connection (this is analogous to pipelining requests on
         a TCP/TLS connection)

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

5.8.  Flow Control Mechanisms



   Servers and clients manage flow control using the mechanisms defined
   in Section 4 of [RFC9000].  These mechanisms allow clients and
   servers to specify how many streams can be created, how much data can
   be sent on a stream, and how much data can be sent on the union of
   all streams.  For DoQ, controlling how many streams are created
   allows servers to control how many new requests the client can send
   on a given connection.

   Flow control exists to protect endpoint resources.  For servers,
   global and per-stream flow control limits control how much data can
   be sent by clients.  The same mechanisms allow clients to control how
   much data can be sent by servers.  Values that are too small will
   unnecessarily limit performance.  Values that are too large might
   expose endpoints to overload or memory exhaustion.  Implementations
   or deployments will need to adjust flow control limits to balance
   these concerns.  In particular, zone transfer implementations will
   need to control these limits carefully to ensure both large and
   concurrent zone transfers are well managed.

   Initial values of parameters control how many requests and how much
   data can be sent by clients and servers at the beginning of the
   connection.  These values are specified in transport parameters
   exchanged during the connection handshake.  The parameter values
   received in the initial connection also control how many requests and
   how much data can be sent by clients using 0-RTT data in a resumed
   connection.  Using too small values of these initial parameters would
   restrict the usefulness of allowing 0-RTT data.

6.  Security Considerations



   A Threat Analysis of the Domain Name System is found in [RFC3833].
   This analysis was written before the development of DoT, DoH, and
   DoQ, and probably needs to be updated.

   The security considerations of DoQ should be comparable to those of
   DoT [RFC7858].  DoT as specified in [RFC7858] only addresses the stub
   to recursive scenario, but the considerations about person-in-the-
   middle attacks, middleboxes, and caching of data from cleartext
   connections also apply for DoQ to the resolver to authoritative
   server scenario.  As stated in Section 5.1, the authentication
   requirements for securing zone transfer using DoQ are the same as
   those for zone transfer over DoT; therefore, the general security
   considerations are entirely analogous to those described in
   [RFC9103].

   DoQ relies on QUIC, which itself relies on TLS 1.3 and thus supports
   by default the protections against downgrade attacks described in
   [BCP195].  QUIC-specific issues and their mitigations are described
   in Section 21 of [RFC9000].

7.  Privacy Considerations



   The general considerations of encrypted transports provided in "DNS
   Privacy Considerations" [RFC9076] apply to DoQ.  The specific
   considerations provided there do not differ between DoT and DoQ, and
   they are not discussed further here.  Similarly, "Recommendations for
   DNS Privacy Service Operators" [RFC8932] (which covers operational,
   policy, and security considerations for DNS privacy services) is also
   applicable to DoQ services.

   QUIC incorporates the mechanisms of TLS 1.3 [RFC8446], and this
   enables QUIC transmission of "0-RTT" data.  This can provide
   interesting latency gains, but it raises two concerns:

   1.  Adversaries could replay the 0-RTT data and infer its content
       from the behavior of the receiving server.

   2.  The 0-RTT mechanism relies on TLS session resumption, which can
       provide linkability between successive client sessions.

   These issues are developed in Sections 7.1 and 7.2.

7.1.  Privacy Issues with 0-RTT data



   The 0-RTT data can be replayed by adversaries.  That data may trigger
   queries by a recursive resolver to authoritative resolvers.
   Adversaries may be able to pick a time at which the recursive
   resolver outgoing traffic is observable and thus find out what name
   was queried for in the 0-RTT data.

   This risk is in fact a subset of the general problem of observing the
   behavior of the recursive resolver discussed in "DNS Privacy
   Considerations" [RFC9076].  The attack is partially mitigated by
   reducing the observability of this traffic.  The mandatory replay
   protection mechanisms in TLS 1.3 [RFC8446] limit but do not eliminate
   the risk of replay. 0-RTT packets can only be replayed within a
   narrow window, which is only wide enough to account for variations in
   clock skew and network transmission.

   The recommendation for TLS 1.3 [RFC8446] is that the capability to
   use 0-RTT data should be turned off by default and only enabled if
   the user clearly understands the associated risks.  In the case of
   DoQ, allowing 0-RTT data provides significant performance gains, and
   there is a concern that a recommendation to not use it would simply
   be ignored.  Instead, a set of practical recommendations is provided
   in Sections 4.5 and 5.5.3.

   The specifications in Section 4.5 block the most obvious risks of
   replay attacks, as they only allow for transactions that will not
   change the long-term state of the server.

   The attacks described above apply to the stub resolver to recursive
   resolver scenario, but similar attacks might be envisaged in the
   recursive resolver to authoritative resolver scenario, and the same
   mitigations apply.

7.2.  Privacy Issues with Session Resumption



   The QUIC session resumption mechanism reduces the cost of re-
   establishing sessions and enables 0-RTT data.  There is a linkability
   issue associated with session resumption, if the same resumption
   token is used several times.  Attackers on path between client and
   server could observe repeated usage of the token and use that to
   track the client over time or over multiple locations.

   The session resumption mechanism allows servers to correlate the
   resumed sessions with the initial sessions and thus to track the
   client.  This creates a virtual long duration session.  The series of
   queries in that session can be used by the server to identify the
   client.  Servers can most probably do that already if the client
   address remains constant, but session resumption tickets also enable
   tracking after changes of the client's address.

   The recommendations in Section 5.5.3 are designed to mitigate these
   risks.  Using session tickets only once mitigates the risk of
   tracking by third parties.  Refusing to resume a session if addresses
   change mitigates the incremental risk of tracking by the server (but
   the risk of tracking by IP address remains).

   The privacy trade-offs here may be context specific.  Stub resolvers
   will have a strong motivation to prefer privacy over latency since
   they often change location.  However, recursive resolvers that use a
   small set of static IP addresses are more likely to prefer the
   reduced latency provided by session resumption and may consider this
   a valid reason to use resumption tickets even if the IP address
   changed between sessions.

   Encrypted zone transfer ([RFC9103]) explicitly does not attempt to
   hide the identity of the parties involved in the transfer; at the
   same time, such transfers are not particularly latency sensitive.
   This means that applications supporting zone transfers may decide to
   apply the same protections as stub to recursive applications.

7.3.  Privacy Issues with Address Validation Tokens



   QUIC specifies address validation mechanisms in Section 8 of
   [RFC9000].  Use of an address validation token allows QUIC servers to
   avoid an extra RTT for new connections.  Address validation tokens
   are typically tied to an IP address.  QUIC clients normally only use
   these tokens when setting up a new connection from a previously used
   address.  However, clients are not always aware that they are using a
   new address.  This could be due to NAT, or because the client does
   not have an API available to check if the IP address has changed
   (which can be quite often for IPv6).  There is a linkability risk if
   clients mistakenly use address validation tokens after unknowingly
   moving to a new location.

   The recommendations in Section 5.5.3 mitigates this risk by tying the
   usage of the NEW_TOKEN to that of session resumption, though this
   recommendation does not cover the case where the client is unaware of
   the address change.

7.4.  Privacy Issues with Long Duration Sessions



   A potential alternative to session resumption is the use of long
   duration sessions: if a session remains open for a long time, new
   queries can be sent without incurring connection establishment
   delays.  It is worth pointing out that the two solutions have similar
   privacy characteristics.  Session resumption may allow servers to
   keep track of the IP addresses of clients, but long duration sessions
   have the same effect.

   In particular, a DoQ implementation might take advantage of the
   connection migration features of QUIC to maintain a session even if
   the client's connectivity changes, for example, if the client
   migrates from a Wi-Fi connection to a cellular network connection and
   then to another Wi-Fi connection.  The server would be able to track
   the client location by monitoring the succession of IP addresses used
   by the long duration connection.

   The recommendation in Section 5.5.4 mitigates the privacy concerns
   related to long duration sessions using multiple client addresses.

7.5.  Traffic Analysis



   Even though QUIC packets are encrypted, adversaries can gain
   information from observing packet lengths, in both queries and
   responses, as well as packet timing.  Many DNS requests are emitted
   by web browsers.  Loading a specific web page may require resolving
   dozens of DNS names.  If an application adopts a simple mapping of
   one query or response per packet, or "one QUIC STREAM frame per
   packet", then the succession of packet lengths may provide enough
   information to identify the requested site.

   Implementations SHOULD use the mechanisms defined in Section 5.4 to
   mitigate this attack.

8.  IANA Considerations



8.1.  Registration of a DoQ Identification String



   This document creates a new registration for the identification of
   DoQ in the "TLS Application-Layer Protocol Negotiation (ALPN)
   Protocol IDs" registry [RFC7301].

   The "doq" string identifies DoQ:

   Protocol:  DoQ

   Identification Sequence:  0x64 0x6F 0x71 ("doq")

   Specification:  This document

8.2.  Reservation of a Dedicated Port



   For both TCP and UDP, port 853 is currently reserved for "DNS query-
   response protocol run over TLS/DTLS" [RFC7858].

   However, the specification for DNS over DTLS (DoD) [RFC8094] is
   experimental, limited to stub to resolver, and no implementations or
   deployments currently exist to the authors' knowledge (even though
   several years have passed since the specification was published).

   This specification additionally reserves the use of UDP port 853 for
   DoQ.  QUIC version 1 was designed to be able to coexist with other
   protocols on the same port, including DTLS; see Section 17.2 of
   [RFC9000].  This means that deployments that serve DoD and DoQ (QUIC
   version 1) on the same port will be able to demultiplex the two due
   to the second most significant bit in each UDP payload.  Such
   deployments ought to check the signatures of future versions or
   extensions (e.g., [GREASING-QUIC]) of QUIC and DTLS before deploying
   them to serve DNS on the same port.

   IANA has updated the following value in the "Service Name and
   Transport Protocol Port Number Registry" in the System range.  The
   registry for that range requires IETF Review or IESG Approval
   [RFC6335].

   Service Name:  domain-s

   Port Number:  853

   Transport Protocol(s):  UDP

   Assignee:  IESG

   Contact:  IETF Chair

   Description:  DNS query-response protocol run over DTLS or QUIC

   Reference:  [RFC7858][RFC8094] This document

   Additionally, IANA has updated the Description field for the
   corresponding TCP port 853 allocation to be "DNS query-response
   protocol run over TLS" and removed [RFC8094] from the TCP
   allocation's Reference field for consistency and clarity.

8.3.  Reservation of an Extended DNS Error Code: Too Early



   IANA has registered the following value in the "Extended DNS Error
   Codes" registry [RFC8914]:

   INFO-CODE:  26

   Purpose:  Too Early

   Reference:  This document

8.4.  DNS-over-QUIC Error Codes Registry



   IANA has added a registry for "DNS-over-QUIC Error Codes" on the
   "Domain Name System (DNS) Parameters" web page.

   The "DNS-over-QUIC Error Codes" registry governs a 62-bit space.
   This space is split into three regions that are governed by different
   policies:

   *  Permanent registrations for values between 0x00 and 0x3f (in
      hexadecimal; inclusive), which are assigned using Standards Action
      or IESG Approval as defined in Sections 4.9 and 4.10 of [RFC8126]

   *  Permanent registrations for values larger than 0x3f, which are
      assigned using the Specification Required policy ([RFC8126])

   *  Provisional registrations for values larger than 0x3f, which
      require Expert Review, as defined in Section 4.5 of [RFC8126].

   Provisional reservations share the range of values larger than 0x3f
   with some permanent registrations.  This is by design to enable
   conversion of provisional registrations into permanent registrations
   without requiring changes in deployed systems.  (This design is
   aligned with the principles set in Section 22 of [RFC9000].)

   Registrations in this registry MUST include the following fields:

   Value:  The assigned codepoint

   Status:  "Permanent" or "Provisional"

   Contact:  Contact details for the registrant

   In addition, permanent registrations MUST include:

   Error:  A short mnemonic for the parameter

   Specification:  A reference to a publicly available specification for
      the value (optional for provisional registrations)

   Description:  A brief description of the error code semantics, which
      MAY be a summary if a specification reference is provided

   Provisional registrations of codepoints are intended to allow for
   private use and experimentation with extensions to DoQ.  However,
   provisional registrations could be reclaimed and reassigned for other
   purposes.  In addition to the parameters listed above, provisional
   registrations MUST include:

   Date:  The date of last update to the registration

   A request to update the date on any provisional registration can be
   made without review from the designated expert(s).

   The initial content of this registry is shown in Table 1 and all
   entries share the following fields:

   Status:  Permanent

   Contact:  DPRIVE WG

   Specification:  Section 4.3

   +============+=======================+=============================+
   | Value      | Error                 | Description                 |
   +============+=======================+=============================+
   | 0x0        | DOQ_NO_ERROR          | No error                    |
   +------------+-----------------------+-----------------------------+
   | 0x1        | DOQ_INTERNAL_ERROR    | Implementation error        |
   +------------+-----------------------+-----------------------------+
   | 0x2        | DOQ_PROTOCOL_ERROR    | Generic protocol violation  |
   +------------+-----------------------+-----------------------------+
   | 0x3        | DOQ_REQUEST_CANCELLED | Request cancelled by client |
   +------------+-----------------------+-----------------------------+
   | 0x4        | DOQ_EXCESSIVE_LOAD    | Closing a connection for    |
   |            |                       | excessive load              |
   +------------+-----------------------+-----------------------------+
   | 0x5        | DOQ_UNSPECIFIED_ERROR | No error reason specified   |
   +------------+-----------------------+-----------------------------+
   | 0xd098ea5e | DOQ_ERROR_RESERVED    | Alternative error code used |
   |            |                       | for tests                   |
   +------------+-----------------------+-----------------------------+

            Table 1: Initial DNS-over-QUIC Error Codes Entries

9.  References



9.1.  Normative References



   [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>.

   [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>.

   [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>.

   [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>.

   [RFC7301]  Friedl, S., Popov, A., Langley, A., and E. Stephan,
              "Transport Layer Security (TLS) Application-Layer Protocol
              Negotiation Extension", RFC 7301, DOI 10.17487/RFC7301,
              July 2014, <https://www.rfc-editor.org/info/rfc7301>.

   [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>.

   [RFC7830]  Mayrhofer, A., "The EDNS(0) Padding Option", RFC 7830,
              DOI 10.17487/RFC7830, May 2016,
              <https://www.rfc-editor.org/info/rfc7830>.

   [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>.

   [RFC8126]  Cotton, M., Leiba, B., and T. Narten, "Guidelines for
              Writing an IANA Considerations Section in RFCs", BCP 26,
              RFC 8126, DOI 10.17487/RFC8126, June 2017,
              <https://www.rfc-editor.org/info/rfc8126>.

   [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>.

   [RFC8467]  Mayrhofer, A., "Padding Policies for Extension Mechanisms
              for DNS (EDNS(0))", RFC 8467, DOI 10.17487/RFC8467,
              October 2018, <https://www.rfc-editor.org/info/rfc8467>.

   [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>.

   [RFC9000]  Iyengar, J., Ed. and M. Thomson, Ed., "QUIC: A UDP-Based
              Multiplexed and Secure Transport", RFC 9000,
              DOI 10.17487/RFC9000, May 2021,
              <https://www.rfc-editor.org/info/rfc9000>.

   [RFC9001]  Thomson, M., Ed. and S. Turner, Ed., "Using TLS to Secure
              QUIC", RFC 9001, DOI 10.17487/RFC9001, May 2021,
              <https://www.rfc-editor.org/info/rfc9001>.

   [RFC9103]  Toorop, W., Dickinson, S., Sahib, S., Aras, P., and A.
              Mankin, "DNS Zone Transfer over TLS", RFC 9103,
              DOI 10.17487/RFC9103, August 2021,
              <https://www.rfc-editor.org/info/rfc9103>.

9.2.  Informative References



   [BCP195]   Sheffer, Y., Holz, R., and P. Saint-Andre,
              "Recommendations for Secure Use of Transport Layer
              Security (TLS) and Datagram Transport Layer Security
              (DTLS)", BCP 195, RFC 7525, May 2015.

              Moriarty, K. and S. Farrell, "Deprecating TLS 1.0 and TLS
              1.1", BCP 195, RFC 8996, March 2021.

              <https://www.rfc-editor.org/info/bcp195>

   [DNS-TERMS]
              Hoffman, P. and K. Fujiwara, "DNS Terminology", Work in
              Progress, Internet-Draft, draft-ietf-dnsop-rfc8499bis-03,
              28 September 2021, <https://datatracker.ietf.org/doc/html/
              draft-ietf-dnsop-rfc8499bis-03>.

   [DNS0RTT]  Kahn Gillmor, D., "DNS + 0-RTT", Message to DNS-Privacy WG
              mailing list, 6 April 2016, <https://www.ietf.org/mail-
              archive/web/dns-privacy/current/msg01276.html>.

   [GREASING-QUIC]
              Thomson, M., "Greasing the QUIC Bit", Work in Progress,
              Internet-Draft, draft-ietf-quic-bit-grease-02, 10 November
              2021, <https://datatracker.ietf.org/doc/html/draft-ietf-
              quic-bit-grease-02>.

   [HTTP/3]   Bishop, M., Ed., "Hypertext Transfer Protocol Version 3
              (HTTP/3)", Work in Progress, Internet-Draft, draft-ietf-
              quic-http-34, 2 February 2021,
              <https://datatracker.ietf.org/doc/html/draft-ietf-quic-
              http-34>.

   [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>.

   [RFC3833]  Atkins, D. and R. Austein, "Threat Analysis of the Domain
              Name System (DNS)", RFC 3833, DOI 10.17487/RFC3833, August
              2004, <https://www.rfc-editor.org/info/rfc3833>.

   [RFC6335]  Cotton, M., Eggert, L., Touch, J., Westerlund, M., and S.
              Cheshire, "Internet Assigned Numbers Authority (IANA)
              Procedures for the Management of the Service Name and
              Transport Protocol Port Number Registry", BCP 165,
              RFC 6335, DOI 10.17487/RFC6335, August 2011,
              <https://www.rfc-editor.org/info/rfc6335>.

   [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>.

   [RFC7873]  Eastlake 3rd, D. and M. Andrews, "Domain Name System (DNS)
              Cookies", RFC 7873, DOI 10.17487/RFC7873, May 2016,
              <https://www.rfc-editor.org/info/rfc7873>.

   [RFC8094]  Reddy, T., Wing, D., and P. Patil, "DNS over Datagram
              Transport Layer Security (DTLS)", RFC 8094,
              DOI 10.17487/RFC8094, February 2017,
              <https://www.rfc-editor.org/info/rfc8094>.

   [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>.

   [RFC8490]  Bellis, R., Cheshire, S., Dickinson, J., Dickinson, S.,
              Lemon, T., and T. Pusateri, "DNS Stateful Operations",
              RFC 8490, DOI 10.17487/RFC8490, March 2019,
              <https://www.rfc-editor.org/info/rfc8490>.

   [RFC8932]  Dickinson, S., Overeinder, B., van Rijswijk-Deij, R., and
              A. Mankin, "Recommendations for DNS Privacy Service
              Operators", BCP 232, RFC 8932, DOI 10.17487/RFC8932,
              October 2020, <https://www.rfc-editor.org/info/rfc8932>.

   [RFC9002]  Iyengar, J., Ed. and I. Swett, Ed., "QUIC Loss Detection
              and Congestion Control", RFC 9002, DOI 10.17487/RFC9002,
              May 2021, <https://www.rfc-editor.org/info/rfc9002>.

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

Appendix A.  The NOTIFY Service



   This appendix discusses why it is considered acceptable to send
   NOTIFY (see [RFC1996]) in 0-RTT data.

   Section 4.5 says "The 0-RTT mechanism MUST NOT be used to send DNS
   requests that are not "replayable" transactions".  This specification
   supports sending a NOTIFY in 0-RTT data because although a NOTIFY
   technically changes the state of the receiving server, the effect of
   replaying NOTIFYs has negligible impact in practice.

   NOTIFY messages prompt a secondary to either send an SOA query or an
   XFR request to the primary on the basis that a newer version of the
   zone is available.  It has long been recognized that NOTIFYs can be
   forged and, in theory, used to cause a secondary to send repeated
   unnecessary requests to the primary.  For this reason, most
   implementations have some form of throttling of the SOA/XFR queries
   triggered by the receipt of one or more NOTIFYs.

   [RFC9103] describes the privacy risks associated with both NOTIFY and
   SOA queries and does not include addressing those risks within the
   scope of encrypting zone transfers.  Given this, the privacy benefit
   of using DoQ for NOTIFY is not clear, but for the same reason,
   sending NOTIFY as 0-RTT data has no privacy risk above that of
   sending it using cleartext DNS.

Acknowledgements



   This document liberally borrows text from the HTTP/3 specification
   [HTTP/3] edited by Mike Bishop and from the DoT specification
   [RFC7858] authored by Zi Hu, Liang Zhu, John Heidemann, Allison
   Mankin, Duane Wessels, and Paul Hoffman.

   The privacy issue with 0-RTT data and session resumption was analyzed
   by Daniel Kahn Gillmor (DKG) in a message to the IETF DPRIVE Working
   Group [DNS0RTT].

   Thanks to Tony Finch for an extensive review of the initial draft
   version of this document, and to Robert Evans for the discussion of
   0-RTT privacy issues.  Early reviews by Paul Hoffman and Martin
   Thomson and interoperability tests conducted by Stephane Bortzmeyer
   helped improve the definition of the protocol.

   Thanks also to Martin Thomson and Martin Duke for their later reviews
   focusing on the low-level QUIC details, which helped clarify several
   aspects of DoQ.  Thanks to Andrey Meshkov, Loganaden Velvindron,
   Lucas Pardue, Matt Joras, Mirja Kuelewind, Brian Trammell, and
   Phillip Hallam-Baker for their reviews and contributions.

Authors' Addresses



   Christian Huitema
   Private Octopus Inc.
   427 Golfcourse Rd
   Friday Harbor,  WA 98250
   United States of America
   Email: huitema@huitema.net


   Sara Dickinson
   Sinodun IT
   Oxford Science Park
   Oxford
   OX4 4GA
   United Kingdom
   Email: sara@sinodun.com


   Allison Mankin
   Salesforce
   Email: allison.mankin@gmail.com