Skip to main content

QUIC Profile for Deep Space
draft-ietf-tiptop-quic-profile-00

Document Type Active Internet-Draft (tiptop WG)
Authors Marc Blanchet , Wesley Eddy
Last updated 2026-09-10
Replaces draft-many-tiptop-quic-profile
RFC stream Internet Engineering Task Force (IETF)
Intended RFC status (None)
Formats
Additional resources GitHub Repository
Mailing list discussion
Stream WG state WG Document
Document shepherd (None)
IESG IESG state I-D Exists
Consensus boilerplate Unknown
Telechat date (None)
Responsible AD (None)
Send notices to (None)
draft-ietf-tiptop-quic-profile-00
Internet Engineering Task Force                              M. Blanchet
Internet-Draft                                                  Viagenie
Intended status: Informational                                W. M. Eddy
Expires: 14 March 2027                              Aalyria Technologies
                                                       10 September 2026

                      QUIC Profile for Deep Space
                   draft-ietf-tiptop-quic-profile-00

Abstract

   Deep space communications involve long delays (e.g., the Earth to
   Mars one-way delay is ~4-20 minutes) and often intermittent
   communications.  In this context, the default transport parameters of
   QUIC stacks, tuned for the terrestrial Internet, are not suitable for
   deep space.  This document defines a QUIC profile for deep space.  It
   provides guidance on how to estimate and set transport parameters,
   advice to space mission operators and application developers on how
   to configure QUIC for the deep space use case, and guidance to QUIC
   stack developers on properly exposing the required transport
   parameters in their API.

Status of This Memo

   This Internet-Draft is submitted in full conformance with the
   provisions of BCP 78 and BCP 79.

   Internet-Drafts are working documents of the Internet Engineering
   Task Force (IETF).  Note that other groups may also distribute
   working documents as Internet-Drafts.  The list of current Internet-
   Drafts is at https://datatracker.ietf.org/drafts/current/.

   Internet-Drafts are draft documents valid for a maximum of six months
   and may be updated, replaced, or obsoleted by other documents at any
   time.  It is inappropriate to use Internet-Drafts as reference
   material or to cite them other than as "work in progress."

   This Internet-Draft will expire on 14 March 2027.

Copyright Notice

   Copyright (c) 2026 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.

Blanchet & Eddy           Expires 14 March 2027                 [Page 1]
Internet-Draft         QUIC Profile for Deep Space        September 2026

   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  . . . . . . . . . . . . . . . . . . . . . . . .   3
     1.1.  Requirements Language . . . . . . . . . . . . . . . . . .   3
   2.  Path Characteristics Calculations . . . . . . . . . . . . . .   4
     2.1.  RTT Calculation . . . . . . . . . . . . . . . . . . . . .   4
     2.2.  Bandwidth-Delay Product (BDP) Calculation . . . . . . . .   4
   3.  Connection Establishment  . . . . . . . . . . . . . . . . . .   5
     3.1.  Initial RTT . . . . . . . . . . . . . . . . . . . . . . .   5
     3.2.  Connection Establishment with a Known Peer  . . . . . . .   6
     3.3.  Connection Migration and 0RTT . . . . . . . . . . . . . .   7
   4.  Connection Maintenance  . . . . . . . . . . . . . . . . . . .   7
     4.1.  Idle Timeout  . . . . . . . . . . . . . . . . . . . . . .   7
     4.2.  Keep Alive and Middleboxes  . . . . . . . . . . . . . . .   8
     4.3.  Application Considerations  . . . . . . . . . . . . . . .   8
   5.  Bandwidth Management  . . . . . . . . . . . . . . . . . . . .   8
     5.1.  Congestion Control  . . . . . . . . . . . . . . . . . . .   9
       5.1.1.  Window Size . . . . . . . . . . . . . . . . . . . . .  11
     5.2.  Flow Control  . . . . . . . . . . . . . . . . . . . . . .  12
       5.2.1.  Max Data  . . . . . . . . . . . . . . . . . . . . . .  12
     5.3.  Packet Size . . . . . . . . . . . . . . . . . . . . . . .  12
     5.4.  Sending Pace  . . . . . . . . . . . . . . . . . . . . . .  12
     5.5.  Padding . . . . . . . . . . . . . . . . . . . . . . . . .  13
     5.6.  Path MTU Discovery  . . . . . . . . . . . . . . . . . . .  13
   6.  Reliable Delivery under Loss and Intermittence  . . . . . . .  13
     6.1.  Acknowledgement Frequency . . . . . . . . . . . . . . . .  14
     6.2.  Forward Error Correction (FEC)  . . . . . . . . . . . . .  15
     6.3.  Relay Buffering and Store-and-Forward Durability  . . . .  15
     6.4.  Intermittence Awareness . . . . . . . . . . . . . . . . .  16
   7.  TLS . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  16
   8.  Applicability of the QUIC Profile for Some Example
           Scenarios . . . . . . . . . . . . . . . . . . . . . . . .  18
     8.1.  Earth to Cruising Spacecraft (CS) . . . . . . . . . . . .  18
     8.2.  Earth-Moon with Full Connectivity (LF)  . . . . . . . . .  18
     8.3.  Earth-Moon with Intermittent Connectivity (LI)  . . . . .  19
     8.4.  Earth-Mars with Full Connectivity (MF)  . . . . . . . . .  19
     8.5.  Earth-Mars with Intermittent Connectivity (MI)  . . . . .  20
     8.6.  Solar Conjunction (SC)  . . . . . . . . . . . . . . . . .  20
   9.  IANA Considerations . . . . . . . . . . . . . . . . . . . . .  21
   10. Security Considerations . . . . . . . . . . . . . . . . . . .  21
   11. References  . . . . . . . . . . . . . . . . . . . . . . . . .  22
     11.1.  Normative References . . . . . . . . . . . . . . . . . .  22

Blanchet & Eddy           Expires 14 March 2027                 [Page 2]
Internet-Draft         QUIC Profile for Deep Space        September 2026

     11.2.  Informative References . . . . . . . . . . . . . . . . .  22
   Acknowledgements  . . . . . . . . . . . . . . . . . . . . . . . .  25
   Authors' Addresses  . . . . . . . . . . . . . . . . . . . . . . .  25

1.  Introduction

   Deep space communications involve long delays (e.g., ~4-20 minutes
   one way between Earth and Mars) and intermittent communications
   caused by orbital dynamics, such as when an orbiter passes over a
   rover every 6 hours for a duration of 15 minutes.

   The default transport parameters of typical QUIC stacks, tuned for
   the terrestrial Internet, assume low latency, such as 100-200 ms, and
   relatively continuous connectivity.  Therefore, parameters such as
   initial_rtt and maximum_idle_timeout have default values that are
   typically not suitable for deep space.

   Space missions are scheduled in advance, and parameters such as the
   maximum round-trip time or bandwidth are known and determined in
   advance.  Given the relatively low bandwidth in space and the
   intermittent communications, bandwidth is very precious, and
   therefore any unneeded communication should be minimized as much as
   possible.  Use cases and requirements for deep space IP are further
   discussed in [I-D.ietf-tiptop-usecase].

   As described in [I-D.ietf-tiptop-ip-architecture], packets may be
   stored at either layer 2 or layer 3 by orbiters during the periods
   where connectivity to the next hop is not possible.

   To better illustrate the implications for various transport
   parameters, this document refers to the connectivity scenarios
   described in [I-D.ietf-tiptop-usecase], using their two-letter
   acronyms: Earth to cruising spacecraft (CS), Earth-Moon with full
   connectivity (LF), Earth-Moon with intermittent connectivity (LI),
   Earth-Mars with full connectivity (MF), Earth-Mars with intermittent
   connectivity (MI), and solar conjunction (SC).

   This document discusses how to configure a number of QUIC transport
   parameters.  The parameter names are from the IANA registry
   [iana_quic_transport].

1.1.  Requirements Language

   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.

Blanchet & Eddy           Expires 14 March 2027                 [Page 3]
Internet-Draft         QUIC Profile for Deep Space        September 2026

2.  Path Characteristics Calculations

   This section describes how to calculate the path characteristics of a
   mission, which are then used to set the transport parameters
   discussed in the following sections.

2.1.  RTT Calculation

   A QUIC stack continuously estimates the round-trip time (RTT) between
   the two peers over the lifetime of the connection.  This estimate
   drives loss detection [RFC9002]: both the time threshold after which
   a sent packet is declared lost and its data retransmitted, and the
   probe timeout (PTO) after which a probe is sent when no
   acknowledgement arrives, are derived from the RTT estimate.  Using
   the Earth-Mars with intermittent connectivity (MI) scenario described
   in [I-D.ietf-tiptop-usecase], it is necessary to prime the QUIC stack
   with the right initial values, to avoid, for example, the probe
   timeout firing and sending probes about one second after a packet is
   sent, based on the default initial RTT of 333 ms, while the expected
   RTT is 2 hours.

   A space application designer should calculate the maximum RTT for its
   mission.  Using the MI scenario, the maximum RTT due to the maximum
   two-way delay is 48 minutes and the maximum RTT due to the overpass
   frequency is 6 hours; therefore, the maximum RTT is 6 hours 48
   minutes.

   A space application designer should calculate the minimum RTT for its
   mission.  Using the MI scenario, the minimum RTT due to the minimum
   two-way delay is 8 minutes and the minimum RTT due to the overpass
   frequency is 0, assuming direct line of sight for the whole path;
   therefore, the minimum RTT is 8 minutes.

2.2.  Bandwidth-Delay Product (BDP) Calculation

   A QUIC stack, like any transport stack, manages the pacing of packets
   sent from the source to avoid overloading the network, creating
   congestion, or overwhelming the other peer.

   A space application designer should calculate the bandwidth-delay
   product (BDP) of the whole path for its mission.  The minimum BDP
   should be calculated with the minimum RTT and the minimum bandwidth
   used during those times.  The maximum BDP should be calculated with
   the maximum RTT and the maximum bandwidth.

Blanchet & Eddy           Expires 14 March 2027                 [Page 4]
Internet-Draft         QUIC Profile for Deep Space        September 2026

3.  Connection Establishment

   This section discusses the parameters and mechanisms related to
   establishing a QUIC connection in deep space.

3.1.  Initial RTT

   To prime the QUIC stack with the expected RTT of the mission, an
   application should set the Initial RTT on connection establishment to
   the maximum RTT as calculated in Section 2.1.

   If the configured initial RTT is too low, then retransmissions of the
   initial packet will be sent before the acknowledgement is received,
   and therefore extra bandwidth will be used.  If the actual one-way
   delay is long enough, such as when communicating with Mars, then not
   only will retransmissions occur, but the connection will be closed by
   the client before the acknowledgement is received; therefore, no QUIC
   connection is possible.  In long one-way delay scenarios, setting the
   initial_rtt properly is mandatory.  Simulations
   [deepspace_simulations][johannes_frisch] show that with a one-way
   delay of 15 seconds or more and a default initial RTT of 333 ms, the
   initial RTT MUST be set for the connection not to time out during the
   establishment period.  This 15-second threshold comes from the
   max_idle_timeout, which is set to 30 seconds by default in many QUIC
   stacks: with a one-way delay of 15 seconds or more, the RTT reaches
   or exceeds 30 seconds, so the connection is torn down by the idle
   timeout before the peer's first response can arrive (see
   Section 4.1).

   These simulations [deepspace_simulations] further show that an Earth
   to Mars direct connection with a 4-minute one-way delay (8-minute
   RTT) and default parameters never completes its handshake: the client
   retransmits the initial packet and then closes the connection long
   before the first response arrives.  Once the initial_rtt is set above
   the real RTT, the handshake completes in a single round-trip, and the
   same configuration scales without modification up to an Earth to
   Voyager link with an approximately two-day RTT.  For intermittent
   relayed paths, the initial_rtt should be primed with the maximum RTT
   that includes the store-and-forward contact gap, as computed in
   Section 2.1, and not merely the propagation delay.

Blanchet & Eddy           Expires 14 March 2027                 [Page 5]
Internet-Draft         QUIC Profile for Deep Space        September 2026

   If the configured initial RTT is too high, then when a packet is
   lost, the retransmission will be started later than the optimal time;
   therefore, the total time to transmit all the data, including
   recovery of the losses, will be longer than if it were set properly,
   but the QUIC stack will still converge and deliver reliable data.  As
   the initial RTT is only used to prime the RTT at the beginning of the
   connection, the real measured RTT will replace the too-high initial
   value.

   However, this self-correction does not apply during connection
   establishment, which is the case where an over-set initial_rtt is
   most costly.  Before the first acknowledgement is received there is
   no measured RTT sample, so QUIC's probe timeout (PTO, [RFC9002]) is
   derived entirely from initial_rtt and evaluates to three times
   initial_rtt.  Loss of a handshake packet therefore delays the entire
   connection by one such PTO, a cost that is linear in initial_rtt and
   cannot be replaced by a measured sample because none exists yet.
   Simulations [deepspace_simulations] of a direct Earth to Mars
   connection at the longest one-way delay (22 minutes, a 44-minute RTT)
   in which the very first Initial packet is lost show this directly:
   with initial_rtt set to ten times the one-way delay (about five times
   the real RTT), the lost Initial is recovered only after an 11-hour
   PTO, so the handshake that otherwise completes in one 44-minute round
   trip instead completes in 11 hours 44 minutes; the connection still
   establishes and delivers reliably, but the single lost packet costs
   three times the over-provisioned initial_rtt.  Had initial_rtt been
   primed at the real RTT, the same loss would have cost roughly 2.2
   hours.  Operators should therefore set initial_rtt high enough to
   establish the connection (per the requirement above) but close to the
   maximum expected RTT, and avoid padding it by a large multiple, so
   that recovery from any establishment-phase loss is not needlessly
   inflated.

   The initial_rtt transport parameter is specified in [RFC9002].

3.2.  Connection Establishment with a Known Peer

   If a new connection is set up with a known peer, the other peer can
   provide to the connecting peer some configuration parameters based on
   the previous connections, using the careful resume mechanism
   [RFC9959].  Simulations have shown its usefulness for Moon scenarios
   [natia_piotrowski].

Blanchet & Eddy           Expires 14 March 2027                 [Page 6]
Internet-Draft         QUIC Profile for Deep Space        September 2026

3.3.  Connection Migration and 0RTT

   QUIC connection migration enables a client to restore the connection
   after a change in its IP address or port, as seen by the server.  The
   migration requires 1 RTT to complete.  However, given previously
   exchanged security credentials, data can be sent in the initial
   packet that re-establishes the connection, a process known as 0RTT.
   However, 0RTT is subject to replay attacks [RFC9001], which means
   that if it is used, the application should be careful about what is
   sent in the 0RTT initial packet, following the guidance provided in
   section 5.6 of [RFC9001].  A maximum time is also set for using 0RTT,
   as discussed in Section 7.  An application or mission may also decide
   to disable the use of 0RTT given the security risks.

4.  Connection Maintenance

   This section discusses keeping a QUIC connection alive over long
   delays and intermittence.

4.1.  Idle Timeout

   To prevent the QUIC stack from terminating a connection due to a lack
   of activity from the other peer, an application should set the Idle
   Timeout on connection establishment to at least the maximum RTT as
   calculated in Section 2.1.  This is a minimum: as discussed below, a
   larger value may be needed, since the application may not be sending
   data for periods longer than the maximum RTT.

   If the configured idle timeout is too low, then the other peer may
   terminate the connection before all the data is received.  In this
   case, the QUIC stack on the sender side will need to reestablish the
   connection, possibly using the 0RTT mechanism, and resend the data
   that was not previously acknowledged.  The application shall still
   recover and provide full data reliability, but at the expense of more
   total time and extra bandwidth.  If the configured idle timeout is
   too high, then the other peer will close its side of the connection
   later than needed in the event of a lost connection.  In this case,
   the resources used by keeping the connection open, such as memory,
   will not be released as quickly as they could be if the idle timeout
   were properly set.

   The effective idle timeout is the minimum of the max_idle_timeout of
   both peers; therefore, careful configuration is needed on both peers.

   Simulations [deepspace_simulations] illustrate this boundary
   precisely: with the default 30-second idle timeout, a connection
   whose RTT is 29.8 seconds completes, whereas an otherwise identical
   connection at a 30-second RTT is torn down before each response

Blanchet & Eddy           Expires 14 March 2027                 [Page 7]
Internet-Draft         QUIC Profile for Deep Space        September 2026

   arrives.  In intermittent scenarios, the gap that matters is not the
   RTT but the entire blackout between contacts; therefore the
   max_idle_timeout MUST exceed the longest expected contact gap, or
   keep alives MUST be sent within it.

   The max_idle_timeout transport parameter is specified in section 8.2
   of [RFC9000].

4.2.  Keep Alive and Middleboxes

   As discussed in section 10.1 of [RFC9000], a peer can send PING
   frames as connection keep-alives, timed, accounting for delay and
   intermittence, to arrive before the max_idle_timeout is triggered.
   For example, some QUIC stacks [quinn_keepalive] can be configured
   with a keep-alive interval.

   On the terrestrial Internet, keep-alives are particularly useful to
   prevent middleboxes from timing out the UDP mapping too early,
   typically after 30 seconds, as discussed in section 3.2 of [RFC9308].
   However, in deep space, given long delays and disruptions, such small
   middlebox timeouts will effectively prevent connections from
   succeeding.  Therefore, middleboxes are not recommended in deep space
   or, if used, should be properly configured with much larger timeouts.

4.3.  Application Considerations

   On the terrestrial Internet, the cost, in time or bandwidth, of re-
   establishing a connection to the same peer is very low, since it is
   re-established very quickly and without much use of bandwidth.
   Therefore, applications are often designed in a pattern of
   establishing a connection, sending a query, getting the response,
   closing the connection, and redoing the whole process for the next
   query.  In deep space, the cost, especially the delay, of re-
   establishing a connection is very high in the presence of long RTTs
   or intermittence.  Therefore, applications should be carefully
   designed not to close connections after a single query-response
   exchange if new queries are possible in the future.

5.  Bandwidth Management

   Given the scarcity of bandwidth in deep space, this section discusses
   the parameters and mechanisms that govern the use of the available
   bandwidth.

Blanchet & Eddy           Expires 14 March 2027                 [Page 8]
Internet-Draft         QUIC Profile for Deep Space        September 2026

5.1.  Congestion Control

   In the Internet stack, congestion control (CC) in transport protocols
   is crucial to sharing network resources among concurrent traffic
   flows.  Minimizing congestion is necessary in order to prevent
   unnecessary queueing latency and packet drops.  Problems with typical
   Internet congestion control algorithms in deep space settings can
   include: (1) assumptions of timely in-band feedback/information, (2)
   interpretation of large delays as congestion/loss, and (3) lack of
   knowledge about overall network state and management decisions.

   Internet transport protocols use estimates of loss events, round-trip
   latency, and Explicit Congestion Notification (ECN)
   [RFC3168][RFC8087][RFC8311] to sense congestion.  All of these
   methods operate in a "closed loop" fashion on the scale of RTTs, in
   order to react responsively in normal Internet conditions.  When the
   RTTs are many minutes or more in some deep space scenarios, this is
   not effective in quickly detecting and reacting to congestion.
   Instead of relying on in-band closed loop detection and response to
   congestion, deep space congestion control could be assisted by
   management/control plane information, as deep space access,
   networking, and mission operations are substantially coordinated, in
   contrast to normal Internet access.

   [I-D.ietf-tiptop-ip-architecture] discusses that, given intermittent
   connectivity in deep space, nodes may need to temporarily store
   either L2 frames or L3 packets when links are down, until the link is
   up again.  This behavior will be interpreted by various CC algorithms
   as congestion.  However, this interpretation is not necessarily
   correct, since the delays generated are due to waiting for a lower-
   layer transmission opportunity, not necessarily to a lack of capacity
   when it becomes available to transmit.

   Deep space networking differs from traditional Internet use, because
   it is highly coordinated.  Antennas must be physically pointed and
   tracked, and operations centers must work together across agencies,
   companies, etc. in order to manage resources (including spectrum
   usage, physical systems, compute and storage, etc.) and orchestrate
   configurations and activities over time.  Planning and scheduling are
   significant activities and result in the ability to understand
   traffic flow needs and allocate capacity for application flows,
   unlike in the traditional Internet.

Blanchet & Eddy           Expires 14 March 2027                 [Page 9]
Internet-Draft         QUIC Profile for Deep Space        September 2026

   Other space-oriented transport protocols, such as SCPS-TP, have
   included rate-based open loop congestion control algorithms.  These
   algorithms send at pre-determined rates, without necessarily
   requiring in-band feedback, with the rate information over time
   provided by variables managed through an external scheduling system
   or other means.

   Therefore, QUIC stacks for deep space should be configurable to use
   rate-based open-loop congestion control, which allows relevant
   configuration variables such as the window, pacing rate, expected
   RTT, and others to be governed by schedule-driven management/control
   plane cues, rather than closed-loop in-band probing/estimation.

   Simulations [deepspace_simulations] quantify the cost of running a
   conventional loss-based controller on such links by transferring the
   same lossy multi-packet workload twice, once with CUBIC and once with
   congestion control disabled, across six regimes spanning Earth to
   Moon and Earth to Mars at both the closest and farthest one-way
   delays, over both continuous and intermittent relayed paths, with a 5
   percent loss applied to the downlink the controller governs.  The
   choice of controller does not affect reliability: every run delivered
   100 percent of its requests, because end-to-end delivery is a
   property of QUIC's own loss recovery, not of the congestion
   controller.  The choice affects only completion time, and CUBIC is
   always slower, never faster, because it misreads the channel loss as
   congestion and multiplicatively reduces its window on every loss,
   then re-grows it over the minutes-long round trips.

   On a continuous lossy link the penalty is a steady factor of roughly
   1.5 (from +46 percent at the Moon to +60 percent at the closest Mars
   distance) independent of distance: each multi-packet response that
   the open-loop arm ships in about one round trip is instead slow-
   started by CUBIC over several round trips, while each loss cuts the
   window it had just grown.  Under intermittent store-and-forward the
   penalty is far larger and is counted in missed contact passes rather
   than round trips, because the open-loop sender arrives at pass-open
   with a full window and drains the queue within the pass, whereas
   CUBIC arrives with a small, loss-shrunk window, cannot fill the pass,
   and the remainder waits for the next orbit.  The cost reached a
   factor of about 3.6 at Mars and a factor of about 12 at the Moon,
   where the round trip is tiny relative to the orbit and the open-loop
   arm finishes all of its work inside the first pass; the harm
   therefore scales with how bursty the contact is, not with the loss
   rate, and is worst precisely where open-loop control is most
   efficient.  CUBIC also raises the total packet count (about +30
   percent in one Mars run) through window probing and extra recovery
   round trips, spending more of the scarce return channel for a slower
   transfer, and its conservative initial window can further delay

Blanchet & Eddy           Expires 14 March 2027                [Page 10]
Internet-Draft         QUIC Profile for Deep Space        September 2026

   connection establishment, which on the farthest intermittent path
   waited additional contact windows before reaching the connected
   state.  Operators whose stack cannot disable congestion control
   should budget for at least this 1.5-fold time penalty on continuous
   lossy links and several missed orbits per transfer on intermittent
   ones, and size idle timeouts accordingly.

5.1.1.  Window Size

   A QUIC stack manages the pacing of the source by the window size.  A
   typical value used for the Internet is 2 times the BDP.  In space,
   careful consideration is required.  Too low a BDP means that the
   source node may not be sending enough packets to completely use the
   network and the available bandwidth of the links, which is suboptimal
   given the scarcity of communications in space.  Therefore, an
   application should not use a BDP lower than the minimum BDP as
   calculated in Section 2.2.  Too large a BDP may use too much of the
   bandwidth of the links.

   Since packets may be stored at either layer 2 or layer 3 by
   intermediate nodes, the maximum storage of in-flight packets in these
   intermediary nodes is to be considered.  Therefore, space operations
   should properly identify the best window size based on the minimum
   and maximum BDP and the storage size of the intermediary nodes for
   the mission/application.  As those parameters are known in advance
   for a mission, they can be set appropriately by the application on
   connection establishment.

   When a relayed path includes a downstream link slower than an
   upstream one, for example a fast Earth to orbiter trunk feeding a
   slower orbiter to surface proximity link, the source outbound rate,
   whether governed by the pacing rate, the window, or an explicit
   outbound rate limit, SHOULD be set to the rate of the slowest
   downstream link.  Simulations [deepspace_simulations] show that a
   rate set above the slowest link does not increase the end-to-end
   throughput, which is bounded by that link, but instead forces the
   intermediary node to buffer the excess: a rate limit at twice the
   slow-link rate caused the orbiter to buffer approximately half of
   each transfer, whereas matching the rate to the slow link reduced the
   peak buffer to a single packet, with an identical total transfer
   time.  Sizing the outbound rate to the bottleneck thus bounds the in-
   network storage required at store-and-forward nodes, which is a
   scarce resource on spacecraft.

Blanchet & Eddy           Expires 14 March 2027                [Page 11]
Internet-Draft         QUIC Profile for Deep Space        September 2026

5.2.  Flow Control

   A receiver can tell the source how much data it is willing to accept
   by sending the Max Data parameters.  As discussed above for window
   size, too low a value would not fully use the bandwidth of the
   network.

5.2.1.  Max Data

   initial_max_data is the maximum number of bytes that can be sent on a
   connection [RFC9000]. initial_max_stream_data is similar but per
   stream.  Given the BDP of a typical deep space connection,
   applications should set these parameters to sufficiently large values
   so that the source is capable of sending data while the bandwidth is
   available.

   The various initial_max_data transport parameters are specified in
   section 8.2 of [RFC9000].

5.3.  Packet Size

   There are pros and cons to sending large packets.  Sending smaller
   packets means using more bandwidth because of multiple headers,
   unless header compression is used; however, when a packet is lost,
   the cost of that loss, in the amount of data to retransmit, is
   smaller than with larger packets, an important consideration in space
   where retransmissions are costly.  Sending larger packets means more
   efficient use of bandwidth, but in the presence of significant packet
   loss, it may in fact result in using more bandwidth than smaller
   packets, since the larger lost packets have to be retransmitted.

5.4.  Sending Pace

   In the Internet context, QUIC stacks may elect not to wait long to
   add more frames to a single packet, in order to deliver data faster
   to the receiving endpoint.  By contrast, for deep space applications,
   where transmission times are many orders of magnitude longer than on
   the Internet, a QUIC stack may be configured to wait "a bit more" to
   add more frames to a single packet.  For example, before sending a
   packet, a QUIC stack may wait until it has processed all incoming
   packets, in case that processing adds more frames to the response
   packet.  See Section 13 of [RFC9000] for more information.

Blanchet & Eddy           Expires 14 March 2027                [Page 12]
Internet-Draft         QUIC Profile for Deep Space        September 2026

5.5.  Padding

   When the QUIC frames do not fully use the packet Maximum Transmission
   Unit (MTU), padding to fill the MTU can be used to make traffic
   analysis more difficult, as discussed in section 8.2 of [RFC9002].
   However, this consumes more bandwidth, which is an important
   consideration in deep space.  Therefore, a mission or application
   should weigh the advantages and disadvantages of padding.

5.6.  Path MTU Discovery

   To find the optimum MTU, some QUIC stacks implement Path MTU
   discovery [RFC8899] by sending increasingly larger packets until the
   maximum MTU is found, which involves packet loss when the limit is
   reached.  Given that in deep space the MTUs of all links on the paths
   are typically known in advance, and that probing is inefficient and
   not timely in this context, the application developer may elect to
   disable the path MTU discovery mechanism and set the real path MTU at
   connection establishment, by setting the max_udp_payload_size
   transport parameter, specified in section 18.2 of [RFC9000], to the
   known path MTU minus the IP and UDP header sizes.

   For instance, a 1280 byte path MTU might be assumed, based on
   knowledge of the network design and assurance from network providers
   that this would be supported without fragmentation, after accounting
   for any encapsulation present on the scheduled paths.

   Given coordination between the small number of network operators
   involved in an end-to-end path, support for large MTUs might be
   coordinated and known in advance.  This could be desirable for
   efficiency, but is out of scope of this document.

   Since scheduled paths may change over time, to enable the use of a
   known optimal value at any time, the path MTU can also be exposed by
   an implementation as a dynamically managed parameter, set according
   to a schedule.

6.  Reliable Delivery under Loss and Intermittence

   This section discusses the mechanisms involved in delivering data
   reliably over lossy and intermittent deep space paths:
   acknowledgements, loss recovery, and in-network buffering.

Blanchet & Eddy           Expires 14 March 2027                [Page 13]
Internet-Draft         QUIC Profile for Deep Space        September 2026

6.1.  Acknowledgement Frequency

   QUIC stacks have various mechanisms to trigger acknowledgements
   (ACKs), as described in [RFC9000] and [I-D.ietf-quic-ack-frequency].
   There are advantages to sending "frequent" acknowledgements, such as
   enabling the source to free the memory used by acknowledged frames
   earlier.  However, this uses more capacity for sending ACKs, and
   could be problematic in cases of very low link bandwidth or network
   path capacity asymmetry.  The latter case was tested and is discussed
   in [johannes_frisch].

   Adjusted ACK frequency information is provided to a QUIC peer through
   an in-band signal in a QUIC ACK_FREQUENCY frame extension to the QUIC
   base specification.  Usage needs to be negotiated via transport
   parameters on connection establishment, and desired values need to be
   provided (e.g. directly by an application or supplied to the stack
   via a management system).  The max_ack_delay and ack_delay_exponent
   transport parameters are specified in section 8.2 of [RFC9000], and
   requested max_ack_delay changes are provided in ACK_FREQUENCY frames.

   The optimal ACK frequency is related to the bandwidth asymmetry
   between the directions of data and acknowledgement flow.  This may
   vary over time, but also can be known in advance by a route
   orchestrator or other management system.  In order to optimize
   performance, ACK frequency may be exposed as a dynamically controlled
   or time-varying parameter.

   Simulations [deepspace_simulations] sweeping the ACK frequency
   extension parameters (the ack-eliciting threshold and the maximum
   acknowledgement delay) across direct and intermittent Earth-Moon and
   Earth-Mars paths, with multi-packet bulk responses, show that
   batching acknowledgements removes 80 to 99 percent of the return-path
   ACK packets in every regime, and also reduces the store-and-forward
   backlog held at a relay.  A safe setting across all regimes is an
   ack-eliciting threshold of approximately 10 packets combined with a
   max_ack_delay of approximately one tenth of the RTT: the threshold is
   then reached within each multi-packet response, so acknowledgements
   still fire promptly and the reduction comes at no cost in completion
   time.  Raising the threshold beyond the length of a response makes
   the receiver fall back on the max_ack_delay timer; this is harmless
   on short-RTT links, but where the RTT is comparable to the contact
   pass a held acknowledgement can miss its pass and cost a whole orbit,
   or trip the sender's probe timeout.  An over-aggressive max_ack_delay
   of several RTTs always increases completion time (from +16 to +143
   percent in these runs); the max_ack_delay must remain well below one
   RTT.

Blanchet & Eddy           Expires 14 March 2027                [Page 14]
Internet-Draft         QUIC Profile for Deep Space        September 2026

6.2.  Forward Error Correction (FEC)

   CCSDS deep space links use Forward Error Correction (FEC) at layer 2,
   as specified in the TM and TC Synchronization and Channel Coding
   Recommended Standards [CCSDS_TM_CC][CCSDS_TC_CC], using a pipeline of
   codecs, enabling a low frame error rate in the presence of a higher
   signal bit error rate.  While FEC for QUIC has been defined
   [I-D.michel-quic-fec], it remains to be seen whether it is really
   needed for deep space.

   Simulations [deepspace_simulations] quantify the cost of end-to-end
   recovery and inform this trade-off.  Because QUIC loss recovery is
   clocked by the RTT, a single lost packet is recovered in
   approximately one to two RTTs regardless of distance: a fast
   retransmit when a following acknowledgement exposes the gap, or a
   probe timeout otherwise.  Its wall-clock cost therefore grows
   linearly with the RTT, from under two seconds at the Moon to
   approximately sixteen minutes (two RTTs) at Mars for a single lost
   response.  At interplanetary distances, link-layer or transport-layer
   FEC is therefore very valuable, as it greatly reduces the packet
   losses that the transport would otherwise have to recover through
   costly end-to-end retransmissions.

6.3.  Relay Buffering and Store-and-Forward Durability

   Intermittent relay scenarios rely on IP-layer store-and-forward at
   intermediate nodes: a relay (for example an orbiter) buffers packets
   that arrive while a downstream link is down and forwards them when
   contact resumes.  This buffer is operational state that a relay may
   lose while it still holds in-flight data, for example because of a
   reboot, a power cycle, a radiation-induced single-event upset, memory
   corruption, or an on-board memory reset.  The QUIC connection runs
   end-to-end and does not itself observe such a loss; its loss
   detection and retransmission mechanisms will recover the lost data,
   but the cost of that recovery, as discussed below, ranges from
   negligible to several orbital periods depending on the path
   characteristics.

   Simulations [deepspace_simulations] stress this case by clearing the
   relay buffer at random times while it is non-empty, across the same
   low-orbit relay topologies used elsewhere in this document.  The
   finding is that the survivability of a relay buffer loss is governed
   by the ratio of the end-to-end RTT to the blackout (inter-contact)
   duration:

   *  When the RTT is much smaller than the blackout (for example an
      Earth-Moon relay with an approximately 2.6-second RTT against a
      blackout on the order of an hour), the sender's loss probes (PTO)

Blanchet & Eddy           Expires 14 March 2027                [Page 15]
Internet-Draft         QUIC Profile for Deep Space        September 2026

      repeatedly refresh the relay's buffer during the blackout, so a
      retransmission that entered the relay after the loss is already
      buffered downstream when the next contact opens and carries the
      data through on the same pass.  The loss is masked at negligible
      cost in completion time.

   *  When the RTT approaches the contact cadence (for example an Earth-
      Mars relay near closest approach with an approximately 8-minute
      RTT), probes are sparse, so a lost buffered request usually has no
      surviving retransmission waiting when the pass opens and misses
      it, waiting a full orbit.  Delivery remains reliable, but each
      buffer loss costs on the order of one orbital period, roughly
      halving throughput.

   *  At the longest Earth-Mars delay (an approximately 44-minute RTT, a
      large fraction of the orbital period), recovery from a single loss
      takes on the order of several orbits, and recurring losses can
      prevent the connection from ever being established, because the
      handshake itself is buffered across a blackout and is discarded
      again before it can complete.

6.4.  Intermittence Awareness

   Another way to solve the generic problem is to make the transport
   aware of the intermittence periods, so that when there is a direct
   end-to-end path without any intermittence, the normal QUIC behavior,
   such as congestion control, may be used with proper RTT
   configuration, and a different behavior and configuration may be used
   in the context of intermittence.  Such dynamic configuration changes
   might be driven by Time-Variant Routing (TVR) schedules, for example
   using the YANG data model of [I-D.ietf-tvr-schedule-yang]; the
   details are out of scope of this document.

7.  TLS

   As discussed in section 4.4 of [RFC9001], the certificate chain may
   consume a large number of bytes, which may be limited at the time of
   authentication.  In deep space, this consideration becomes even more
   important, given the delays and low bandwidth.  Therefore, the
   various methods discussed in section 4.4 of [RFC9001] should be
   considered.

   As discussed in section 4.6 of [RFC9001], TLS imposes a limit of 7
   days between the original connection and any attempt to use 0-RTT.
   Given that a spacecraft may travel for a long time before
   reconnecting, missions should consider this possible limitation on
   the use of 0RTT.

Blanchet & Eddy           Expires 14 March 2027                [Page 16]
Internet-Draft         QUIC Profile for Deep Space        September 2026

   As discussed in section 6.5 of [RFC9001], a key update initiation
   should wait 3 times the PTO.  Given that the PTO is calculated based
   on the RTT, and that the RTT may vary significantly in deep space
   because of intermittence, key updates should be carefully planned.

   TLS keys shall be updated depending on the security policy of the
   mission or the application.  QUIC TLS [RFC9001] enables the update of
   keys from each peer.  [I-D.ietf-quic-extended-key-update] enables
   "frequent key updates without requiring a full handshake
   renegotiation.  This approach allows applications to refresh their
   encryption keys more often using ephemeral keys, improving forward
   secrecy and reducing the risk of key compromise over long-lived
   connections".  Therefore, this mechanism is recommended for deep
   space applications.

   Simulations [deepspace_simulations] of the Extended Key Update
   mechanism over deep space paths confirm that it operates correctly
   under deep space delay and intermittent connectivity.  On a
   continuous Earth-Mars path (8 minutes round-trip time), a long-lived
   connection of one thousand requests performed nineteen forward-secret
   key updates with no disruption; on an intermittent Mars orbiter
   store-and-forward relay, the same mechanism performed thirteen
   updates and completed in the same time as the equivalent run without
   key updates, adding only about three packets per update.

   Unlike the RFC 9001 key update, which ratchets the existing secret
   using only locally available material, an Extended Key Update
   requires a full round-trip exchange of ExtendedKeyUpdate messages (a
   request and a response carrying fresh key shares) before the Key
   Phase bit is changed.  The cost of one update is therefore
   approximately two round-trip times.  In deep space, the frequency of
   key updates SHOULD be expressed in units of round-trip time or
   contact cadence (for example, every N requests, or once per contact
   window) rather than as a fixed wall-clock period, so that an update
   can complete within the time and connectivity available.  The header
   protection keys are not changed by a key update.

Blanchet & Eddy           Expires 14 March 2027                [Page 17]
Internet-Draft         QUIC Profile for Deep Space        September 2026

   Across intermittent connectivity, an Extended Key Update initiated
   during a loss of contact is carried by the relay's store-and-forward
   buffer in the same way as application data, and completes once
   contact is restored.  No special handling is required across
   blackouts: the same idle timeout and congestion control configuration
   (see Section 4.1 and Section 5.1) that allow data to traverse
   intermittent contact also allow the key update exchange to traverse
   it.  When both peers schedule key updates and their requests cross in
   flight, the clash is resolved by the tie-break defined in
   [I-D.ietf-quic-extended-key-update] so that the keys still advance by
   a single generation; scheduling updates from a single peer avoids
   redundant updates.

8.  Applicability of the QUIC Profile for Some Example Scenarios

   This section provides guidance for a possible QUIC profile for the
   connectivity scenarios described in [I-D.ietf-tiptop-usecase],
   identified by their two-letter acronyms.  This list is not intended
   to be exhaustive or normative.  Moreover, not all parameters are
   discussed, just the most relevant ones for each scenario.  The reader
   should refer to the sections above for a complete discussion and
   references.

8.1.  Earth to Cruising Spacecraft (CS)

   As discussed in Section 3.1, if the one-way delay is 15 seconds or
   more and a new QUIC connection is started, the establishment will not
   succeed without setting the initial_rtt based on the delay.  If the
   one-way delay is lower than 15 seconds, a non-profiled QUIC
   connection should be established, but at the cost of multiple
   retransmissions of the initial packet.  The max_idle_timeout should
   be set depending on the one-way delay and the application traffic
   pattern.

8.2.  Earth-Moon with Full Connectivity (LF)

   Given the relatively low one-way delay of around 1.5 seconds, a QUIC
   connection with no specific configuration will be successful, but
   there will be unnecessary retransmissions of initial packets by the
   source, as discussed in Section 3.1.  Given full connectivity and the
   relatively low one-way delay, if a connection is closed, the cost, in
   time or bandwidth, of re-establishing it is quite low.  Therefore,
   complicated configurations to avoid connection timeouts may be
   avoided in favor of simplicity.  Moreover, 0RTT may be disabled at
   little cost, for better security.  As discussed in [johannes_frisch]
   and [natia_piotrowski], the Bottleneck Bandwidth and Round-trip
   propagation time (BBR) congestion controller [I-D.ietf-ccwg-bbr] may
   be used and will perform better than others.

Blanchet & Eddy           Expires 14 March 2027                [Page 18]
Internet-Draft         QUIC Profile for Deep Space        September 2026

8.3.  Earth-Moon with Intermittent Connectivity (LI)

   If the periods of reachability are "long enough", then the cost of
   establishing a new QUIC connection at the start of a new reachability
   period may not be too high.  Given that the schedule of intermittence
   is known, the QUIC connection can also be closed gracefully with
   explicit close signaling at the end of the reachability period.  This
   scenario does not require long-term buffering of packets in
   satellites, as no packets shall be sent during the blackout periods.
   During the reachability periods, the scenario is identical to
   Section 8.2.

   The paragraph above assumes that the applications or some application
   orchestrator know the schedule of intermittence.  If the schedule is
   unknown to at least one of the peers, then packets may be sent during
   the blackout periods, which means that the packets will be stored
   somewhere in the network until reachability is back.  This makes the
   RTT highly variable, depending on when those events happen.  As
   discussed in Section 5.1, typical closed-loop congestion controllers
   still deliver reliably in this context, but they perform poorly, as
   they misinterpret the storage delays as congestion.  Therefore, a
   rate-based open-loop congestion control algorithm is preferred.

   At the short Earth-Moon delay, the relay's store-and-forward buffer
   is also resilient to loss: simulations [deepspace_simulations] show
   that even when the orbiter buffer is cleared while it holds data, the
   connection recovers at negligible cost, because the RTT is small
   relative to the blackout and the sender's loss probes quickly refill
   the buffer before the next pass.  This is in contrast to the Earth-
   Mars relay case and is discussed further in Section 6.3.

8.4.  Earth-Mars with Full Connectivity (MF)

   Compared to the Earth-Moon full connectivity scenario, the one-way
   delay is large enough that a QUIC connection without specific
   configuration will not be established, as demonstrated in
   [deepspace_simulations].  Therefore, the very minimum is to set the
   initial_rtt to an appropriate value as discussed in Section 3.1, such
   as the maximum one-way delay times 2 plus some margin.  Moreover, the
   max_idle_timeout should be set to an appropriate value as discussed
   in Section 4.1, such as a multiple of the largest RTT; it may even be
   set to a larger value depending on the application traffic pattern,
   or keep-alives may be used with a period smaller than the configured
   max_idle_timeout.

Blanchet & Eddy           Expires 14 March 2027                [Page 19]
Internet-Draft         QUIC Profile for Deep Space        September 2026

8.5.  Earth-Mars with Intermittent Connectivity (MI)

   As discussed in Section 8.4, the initial_rtt and max_idle_timeout
   must be set large enough, but in this case they must also include the
   maximum RTT based on the expected contact gap plus the maximum one-
   way delay times 2.

   Compared to Section 8.3, the long delay typically makes it infeasible
   to open a new connection at the beginning of the reachability window
   and then close it at the end of that window.  In many cases, the
   distance RTT is larger than the satellite overpass over the Mars
   asset, which means the reachability window is far too small.  In this
   context, the QUIC connection has to be opened once and kept open over
   multiple orbit periods.  Similarly, the congestion controller should
   be of the rate-based open-loop type.

   Packets should be buffered in intermediate nodes of the network, in
   order to minimize retransmissions and to maximize the use of the
   limited bandwidth.

   Simulations [deepspace_simulations] of a low Mars orbit relay, with
   an approximately 118-minute orbital period and a 10-minute surface
   pass, confirm this behavior.  At a 22-minute one-way delay the
   connection effectively delivers a single request/response exchange
   per orbit, because each request's 22-minute uplink reaches the
   orbiter after the current pass has closed and must be stored until
   the next pass; the achievable throughput is then bounded by the
   contact cadence rather than by the transport or the link RTT.
   Notably, even a request issued during a pass can miss that pass when
   the one-way delay is a large fraction of the pass duration.  Across
   these runs, with the intermediate buffers never cleared, no packet
   was lost, which confirms that IP-layer store-and-forward, combined
   with the initial_rtt and max_idle_timeout settings discussed above,
   is sufficient for reliable delivery over scheduled intermittent
   paths.

8.6.  Solar Conjunction (SC)

   In this scenario, described in [I-D.ietf-tiptop-usecase],
   communications between Earth and the assets at a celestial body are
   completely interrupted for a long period: around two weeks
   approximately every 26 months in the case of Mars.  Unlike the
   intermittent connectivity scenarios, the interruption is long, total,
   and precisely predictable years in advance.  As the body approaches
   the Sun in the sky, the link also degrades progressively, with
   increasing error rates, before the complete blackout.

Blanchet & Eddy           Expires 14 March 2027                [Page 20]
Internet-Draft         QUIC Profile for Deep Space        September 2026

   A first approach is to close connections gracefully before the
   conjunction and re-establish them after it ends, following the known
   schedule.  This releases the connection resources on both peers and
   avoids sending any packets, such as retransmissions or keep-alives,
   that cannot traverse the path.  Note that the TLS limit of 7 days
   between the original connection and any use of 0-RTT, discussed in
   Section 7, is shorter than a typical conjunction; therefore, the
   connection re-established after the conjunction cannot benefit from
   0-RTT and requires a full handshake.  The careful resume mechanism
   (see Section 3.2) may still be used to prime the new connection with
   the parameters of the previous one, as the path characteristics are
   mostly unchanged across the conjunction.

   A second approach is to keep the connection open across the
   conjunction.  In this case, the max_idle_timeout of both peers must
   exceed the conjunction duration plus a margin, and all traffic,
   including keep-alives, should be suspended during the blackout using
   the schedule-driven, rate-based open-loop congestion control
   discussed in Section 5.1, so that no bandwidth is wasted on packets
   that cannot be delivered.  Key updates should not be scheduled close
   to or during the conjunction, as discussed in Section 7.

   Communications local to the celestial body, such as between surface
   assets and orbiters, are not affected by the conjunction and can
   continue to operate normally, whether on separate connections or
   through local proxies; only the connections crossing near the Sun are
   interrupted.

9.  IANA Considerations

   This memo includes no request to IANA.

10.  Security Considerations

   The use of 0-RTT is subject to replay attacks [RFC9001] and therefore
   may be disabled depending on the security policy of the mission.

   Certificates and keys need to be renewed before their expiration,
   taking into account the delay to send, receive, and confirm.
   Protocols such as OCSP [RFC6960], which provide online real-time
   validation and revocation checking, will likely not work given the
   long delays; therefore, certificates need to be validated using local
   trust anchors.

   The use of long-term keys, such as ones set prior to launch, may
   create exposure; therefore, keys should be renewed at an appropriate
   frequency.

Blanchet & Eddy           Expires 14 March 2027                [Page 21]
Internet-Draft         QUIC Profile for Deep Space        September 2026

11.  References

11.1.  Normative References

   [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>.

   [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>.

11.2.  Informative References

   [RFC3168]  Ramakrishnan, K., Floyd, S., and D. Black, "The Addition
              of Explicit Congestion Notification (ECN) to IP",
              RFC 3168, DOI 10.17487/RFC3168, September 2001,
              <https://www.rfc-editor.org/info/rfc3168>.

   [RFC6960]  Santesson, S., Myers, M., Ankney, R., Malpani, A.,
              Galperin, S., and C. Adams, "X.509 Internet Public Key
              Infrastructure Online Certificate Status Protocol - OCSP",
              RFC 6960, DOI 10.17487/RFC6960, June 2013,
              <https://www.rfc-editor.org/info/rfc6960>.

   [RFC8087]  Fairhurst, G. and M. Welzl, "The Benefits of Using
              Explicit Congestion Notification (ECN)", RFC 8087,
              DOI 10.17487/RFC8087, March 2017,
              <https://www.rfc-editor.org/info/rfc8087>.

   [RFC8311]  Black, D., "Relaxing Restrictions on Explicit Congestion
              Notification (ECN) Experimentation", RFC 8311,
              DOI 10.17487/RFC8311, January 2018,
              <https://www.rfc-editor.org/info/rfc8311>.

   [RFC8899]  Fairhurst, G., Jones, T., Tüxen, M., Rüngeler, I., and T.
              Völker, "Packetization Layer Path MTU Discovery for
              Datagram Transports", RFC 8899, DOI 10.17487/RFC8899,
              September 2020, <https://www.rfc-editor.org/info/rfc8899>.

   [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>.

Blanchet & Eddy           Expires 14 March 2027                [Page 22]
Internet-Draft         QUIC Profile for Deep Space        September 2026

   [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>.

   [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>.

   [RFC9308]  Kühlewind, M. and B. Trammell, "Applicability of the QUIC
              Transport Protocol", RFC 9308, DOI 10.17487/RFC9308,
              September 2022, <https://www.rfc-editor.org/info/rfc9308>.

   [RFC9959]  Kuhn, N., Stephan, E., Fairhurst, G., Secchi, R., and C.
              Huitema, "Careful Resume: Convergence of Congestion
              Control from Retained State", RFC 9959,
              DOI 10.17487/RFC9959, May 2026,
              <https://www.rfc-editor.org/info/rfc9959>.

   [I-D.michel-quic-fec]
              Michel, F. and O. Bonaventure, "Forward Erasure Correction
              for QUIC loss recovery", Work in Progress, Internet-Draft,
              draft-michel-quic-fec-01, 23 October 2023,
              <https://datatracker.ietf.org/doc/html/draft-michel-quic-
              fec-01>.

   [I-D.ietf-quic-ack-frequency]
              Iyengar, J., Swett, I., and M. Kühlewind, "QUIC
              Acknowledgment Frequency", Work in Progress, Internet-
              Draft, draft-ietf-quic-ack-frequency-14, 5 February 2026,
              <https://datatracker.ietf.org/doc/html/draft-ietf-quic-
              ack-frequency-14>.

   [I-D.ietf-tiptop-usecase]
              Blanchet, M., Eddy, W., and M. Eubanks, "IP in Deep Space:
              Key Characteristics, Use Cases and Requirements", Work in
              Progress, Internet-Draft, draft-ietf-tiptop-usecase-02, 6
              July 2026, <https://datatracker.ietf.org/doc/html/draft-
              ietf-tiptop-usecase-02>.

   [I-D.ietf-tiptop-ip-architecture]
              Blanchet, M., Eddy, W., and T. Li, "An Architecture for IP
              in Deep Space", Work in Progress, Internet-Draft, draft-
              ietf-tiptop-ip-architecture-01, 5 July 2026,
              <https://datatracker.ietf.org/doc/html/draft-ietf-tiptop-
              ip-architecture-01>.

Blanchet & Eddy           Expires 14 March 2027                [Page 23]
Internet-Draft         QUIC Profile for Deep Space        September 2026

   [I-D.ietf-quic-extended-key-update]
              Rosomakho, Y., Tschofenig, H., and T. Reddy.K, "Extended
              Key Update for QUIC", Work in Progress, Internet-Draft,
              draft-ietf-quic-extended-key-update-03, 6 July 2026,
              <https://datatracker.ietf.org/doc/html/draft-ietf-quic-
              extended-key-update-03>.

   [I-D.ietf-tvr-schedule-yang]
              Qu, Y., Lindem, A., Kinzie, E., Fedyk, D., and M.
              Blanchet, "A YANG Data Model for Scheduled Attributes",
              Work in Progress, Internet-Draft, draft-ietf-tvr-schedule-
              yang-12, 10 June 2026,
              <https://datatracker.ietf.org/doc/html/draft-ietf-tvr-
              schedule-yang-12>.

   [I-D.ietf-ccwg-bbr]
              Cardwell, N., Swett, I., and J. Beshay, "BBR Congestion
              Control", Work in Progress, Internet-Draft, draft-ietf-
              ccwg-bbr-06, 6 July 2026,
              <https://datatracker.ietf.org/doc/html/draft-ietf-ccwg-
              bbr-06>.

   [iana_quic_transport]
              Internet Assigned Numbers Authority(IANA), "QUIC Transport
              Parameters", September 2025,
              <https://www.iana.org/assignments/quic/quic.xhtml#quic-
              transport>.

   [CCSDS_TM_CC]
              Consultative Committee on Space Data Systems (CCSDS), "TM
              Synchronization and Channel Coding, Blue Book 131.0-B-5",
              September 2023, <https://ccsds.org/Pubs/131x0b5.pdf>.

   [CCSDS_TC_CC]
              Consultative Committee on Space Data Systems (CCSDS), "TC
              Synchronization and Channel Coding, Blue Book 231.0-B-4",
              July 2021, <https://ccsds.org/Pubs/231x0b4e1.pdf>.

   [deepspace_simulations]
              Marc Blanchet, "Deepspace Simulations of QUIC Transport
              Parameters", July 2026,
              <https://github.com/deepspaceip/deepspace-ip-simulations>.

   [natia_piotrowski]
              Natia Piotrowski, "QUIC to the Moon Measurements",
              November 2025,
              <https://datatracker.ietf.org/meeting/124/materials/
              slides-124-tiptop-quic-to-the-moon-measurements-00>.

Blanchet & Eddy           Expires 14 March 2027                [Page 24]
Internet-Draft         QUIC Profile for Deep Space        September 2026

   [johannes_frisch]
              Johannes Frisch, "Evaluation of QUIC in Interplanetary
              Networks", April 2026,
              <https://datatracker.ietf.org/meeting/126/materials/
              slides-126-tiptop-TBD>.

   [quinn_keepalive]
              Quinn, "Struct TransportConfig", 2026,
              <https://docs.rs/quinn/latest/quinn/
              struct.TransportConfig.html#method.keep_alive_interval>.

Acknowledgements

   This document and its underlying work have been reviewed and
   discussed by many, who have provided valuable feedback and comments,
   including disagreements, making the document more solid overall.
   These people are, in no specific order: Lars Eggert, Christian
   Huitema, Adolfo Ochagavia, Mirja Kuehlewind, Michael Richardson.

   The Quinn QUIC stack was used for testing.  We would like to
   acknowledge the help of Benjamin Saunders and Adolfo Ochagavia in
   using Quinn.

Authors' Addresses

   Marc Blanchet
   Viagenie
   Canada
   Email: marc.blanchet@viagenie.ca

   Wesley M. Eddy
   Aalyria Technologies
   United States of America
   Email: wes@aalyria.com

Blanchet & Eddy           Expires 14 March 2027                [Page 25]