| Internet-Draft | IPN dot ARPA | November 2026 |
| Kline | Expires 20 May 2027 | [Page] |
This document requests the creation of ipn.arpa, an infrastructure DNS namespace for the IPN node identifiers of the Bundle Protocol (BP). Each Fully Qualified Node Number has a DNS base name of the form N.A.ipn.arpa, where A is the Allocator Identifier and N is the Node Number. A BP Agent can query beneath that name for the records that a convergence-layer discovery binding specifies, such as SRV or SVCB records. These records identify the convergence-layer endpoints, including relays, that are advertised for reaching a destination.¶
This note is to be removed before publishing as an RFC.¶
The latest revision of this draft can be found at https://ekline.github.io/draft-dtn-ipn-arpa/draft-ek-dtn-ipn-arpa.html. Status information for this document may be found at https://datatracker.ietf.org/doc/draft-ek-dtn-ipn-arpa/.¶
Discussion of this document takes place on the Delay/Disruption Tolerant Networking mailing list (mailto:dtn@ietf.org), which is archived at https://mailarchive.ietf.org/arch/browse/dtn/. Subscribe at https://www.ietf.org/mailman/listinfo/dtn/.¶
Source for this draft and an issue tracker can be found at https://github.com/ekline/draft-dtn-ipn-arpa.¶
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 20 May 2027.¶
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. 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.¶
The Bundle Protocol (BP) [RFC9171] supports communication in networks that are subject to delay and disruption. The ipn URI scheme [RFC9758] identifies BP nodes and services with numbers. An Allocator assigns Node Numbers, and IANA assigns an Allocator Identifier to each Allocator. The Allocator Identifier and the Node Number together form a Fully Qualified Node Number (FQNN).¶
A BP Agent can know an IPN destination without knowing a DNS domain
associated with its allocator. The ipn.arpa namespace provides a
common starting point for discovery and links DNS administration to
allocator assignments. Each FQNN has one DNS base name,
N.A.ipn.arpa. (Section 3.1). A BP Agent can query beneath
that name for the records that a convergence-layer (CL) discovery
binding defines. These records advertise CL endpoints of the
destination node or of a relay that accepts Bundles for onward
forwarding.¶
DNS discovery is optional, and it supplies only candidate next hops (Section 4.1). It is not a BP routing protocol, and it does not do contact planning. The BP Agent and its configured policy select paths and make forwarding decisions.¶
Public provisioning is limited to eligible Allocator Identifiers (Section 7.3). Names for Allocator Identifier zero and Experimental Use Allocator Identifiers are served only by explicitly configured local DNS (Section 7.2, Section 4.3).¶
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.¶
This document uses the following terms from [RFC9758]:¶
A 32-bit unsigned integer that identifies the Allocator that assigned a Node Number (Section 3.2 of [RFC9758]).¶
A 32-bit unsigned integer that identifies a node in the context of an Allocator (Section 3.3 of [RFC9758]).¶
The combination of an Allocator Identifier and a Node Number, which identifies one node (Section 3.3.1 of [RFC9758]). This document writes an FQNN as (Allocator Identifier, Node Number), for example (978879,2).¶
An unsigned integer that identifies a service on a node (Section 3.5 of [RFC9758]).¶
A URI in the 'ipn' URI scheme. It contains an FQNN and a Service Number (Section 3 of [RFC9758]). An ipn URI used as a BP Endpoint Identifier (EID) is an "ipn EID" (Section 5 of [RFC9758]).¶
This document defines the following terms:¶
The DNS name beneath ipn.arpa that represents an FQNN (Section 3.1).¶
The specification of DNS discovery for one CL. It defines the query names, record types, processing rules, and fallback behavior. This document specifies the TCPCLv4 binding (Section 3.2).¶
The DNS name, relative to a DNS base name, at which a discovery binding queries for records.¶
The entity that does the DNS provisioning and operational functions for the ipn.arpa zone (Section 7.1).¶
A BP Agent populates its Forwarding Information Base (FIB) from configuration and from other sources. It can also create FIB entries from DNS records, as this section describes.¶
Discovery uses the FQNN of the destination node.¶
Before constructing a DNS name, a BP Agent MUST parse and interpret the destination ipn URI or ipn EID according to [RFC9758]. This includes the text syntax of Section 4 of [RFC9758] and the CBOR decoding of Section 6.2 of [RFC9758]. The result is a numeric Allocator Identifier, Node Number, and Service Number. An Allocator Identifier that the text form omits has the value zero (Section 3.2.2 of [RFC9758]); the DNS name still contains the allocator label.¶
For Allocator Identifier A and Node Number N, the DNS base name MUST be:¶
N.A.ipn.arpa.¶
Each of N and A MUST be represented as a single label containing its unsigned decimal value, using ASCII digits without leading zeroes. The value zero is the single digit "0". The labels are not padded and are not split into individual digits.¶
The Service Number is not part of the name, so all ipn URIs with the same FQNN map to the same DNS base name:¶
| ipn URI | DNS base name |
|---|---|
ipn:978879.2.345
|
2.978879.ipn.arpa.
|
ipn:978879.2.0
|
2.978879.ipn.arpa.
|
ipn:16384.345
|
16384.0.ipn.arpa.
|
ipn:0.16384.345
|
16384.0.ipn.arpa.
|
The name comes from the parsed numbers, not from the text of the URI.
Equivalent representations of an FQNN therefore give the same name.
This includes ipn:16384.345 and ipn:0.16384.345, and the
two-element and three-element CBOR encodings of an ipn EID
(Section 6.4 of [RFC9758]).¶
The same rule constructs names for Allocator Identifier zero, including Private Use Node Numbers, and for Experimental Use Allocator Identifiers. These names are not in the public hierarchy (Section 7.2), but local DNS can serve them (Section 4.3).¶
The Null ipn URI (Section 3.1 of [RFC9758]) and LocalNode ipn URIs (Section 3.4.2 of [RFC9758]) do not identify a remote destination. A BP Agent MUST NOT perform DNS discovery for the Null ipn URI or for a LocalNode ipn URI. The BP Agent identifies these cases from the complete FQNN after interpretation: (0,0) is the Null ipn URI, and (0,4294967295) is a LocalNode ipn URI. For example, an ipn URI with a zero Allocator Identifier, a zero Node Number, and a non-zero Service Number is the Null ipn URI (Section 3.4.1 of [RFC9758]). Node Numbers 0 and 4294967295 have no special meaning under other Allocator Identifiers; (978879,0) is an ordinary FQNN. Section 5.4 of [RFC9758] also prohibits the publication of LocalNode ipn EIDs in any node identification directory, such as DNS.¶
This document specifies TCPCLv4 [TCPCL] discovery with SRV records [RFC2782] at the following CL-specific query name:¶
_dtn-bundle._tcp.N.A.ipn.arpa.¶
For example, the query name for the FQNN (978879,2) is
_dtn-bundle._tcp.2.978879.ipn.arpa.. [TCPCL] registers the
"dtn-bundle" service name and TCP port 4556, but it does not define
this discovery procedure.¶
The BP Agent constructs the query name directly. Unlike conventional reverse DNS in in-addr.arpa and ip6.arpa, this procedure needs no preliminary PTR lookup.¶
SRV processing follows [RFC2782], including priority, weight, port, and target processing. The following rules apply to this binding:¶
A new TCPCLv4 discovery result requires a usable SRV record. If the lookup obtains no usable SRV record, discovery supplies no new candidate.¶
The address-record fallback in the usage rules of [RFC2782] does not apply. After an unsuccessful SRV lookup, a BP Agent MUST NOT resolve the DNS base name for address records and connect to TCP port 4556 as part of this binding.¶
An SRV Target of "." indicates that the service is not available at that domain ([RFC2782]).¶
These rules apply only to this binding; other bindings define their own rules (Section 3.5). The rules do not affect independently obtained forwarding information or retained discovery results (Section 4.1, Section 4.2).¶
The examples describe one deployment: a fictional space agency that offers a shared public ingress service. Other deployments can use a separate relay for each node, or publish the endpoints of each node. This document does not require the architecture shown here.¶
The agency administers Allocator Identifier 978879. This identifier, and the identifiers in Section 4.4, are from the range that Section 9.1 of [RFC9758] reserves for examples. The excerpts do not request public provisioning of these identifiers. They omit routine zone infrastructure, such as SOA and NS records, and DNSSEC records.¶
| FQNN | Destination | DNS base name |
|---|---|---|
| (978879,1) | mission operations center |
1.978879.ipn.arpa.
|
| (978879,2) | mars-orbiter |
2.978879.ipn.arpa.
|
| (978879,3) | lunar-lander |
3.978879.ipn.arpa.
|
The destination names are descriptive labels. They do not depend on the Node Numbers.¶
Two relay BP Agents, relay-a.space-agency.example. and
relay-b.space-agency.example., provide the ingress service for all
three destinations. Each relay is a separate BP node with its own
Node ID, and each relay has onward routes to all three destinations.
Onward forwarding uses the cislunar and deep-space ground facilities,
routes, and contact arrangements of the agency. These are not shown.¶
Example A and Example B are alternative configurations of the allocator's zone. The agency publishes only one of them, because a CNAME and an SRV RRset cannot coexist at the same owner name (Section 3.4). Both configurations use these address records:¶
$ORIGIN space-agency.example. $TTL 3600 relay-a IN A 192.0.2.10 relay-a IN AAAA 2001:db8::10 relay-b IN A 192.0.2.11 relay-b IN AAAA 2001:db8::11
The allocator's zone contains the SRV records for each destination:¶
$ORIGIN 978879.ipn.arpa. $TTL 3600 _dtn-bundle._tcp.1 IN SRV 10 1 4556 relay-a.space-agency.example. _dtn-bundle._tcp.1 IN SRV 10 1 4556 relay-b.space-agency.example. _dtn-bundle._tcp.2 IN SRV 10 1 4556 relay-a.space-agency.example. _dtn-bundle._tcp.2 IN SRV 10 1 4556 relay-b.space-agency.example. _dtn-bundle._tcp.3 IN SRV 10 1 4556 relay-a.space-agency.example. _dtn-bundle._tcp.3 IN SRV 10 1 4556 relay-b.space-agency.example.¶
A BP Agent that uses TCPCLv4 to reach ipn:978879.2.345:¶
constructs the query name _dtn-bundle._tcp.2.978879.ipn.arpa.;¶
receives two SRV records, with the targets
relay-a.space-agency.example. and
relay-b.space-agency.example., each with priority 10, weight 1,
and port 4556;¶
orders the candidates as specified in [RFC2782]; because the priorities and weights are equal, either relay can be first;¶
resolves the selected target; for relay-a.space-agency.example.,
it receives 192.0.2.10 and 2001:db8::10; and¶
attempts a TCPCLv4 connection to [2001:db8::10]:4556 or
192.0.2.10:4556. If the attempt fails, it can try
relay-b.space-agency.example. ([2001:db8::11]:4556 or
192.0.2.11:4556). If it establishes a session with a peer that
its policy accepts (Section 5.1), it transfers the
Bundle.¶
Instead of SRV records, the allocator's zone contains a CNAME for each destination. Each CNAME points to one shared name in the agency's zone:¶
$ORIGIN 978879.ipn.arpa.
$TTL 3600
_dtn-bundle._tcp.1 IN CNAME (
_dtn-bundle._tcp.ingress.space-agency.example. )
_dtn-bundle._tcp.2 IN CNAME (
_dtn-bundle._tcp.ingress.space-agency.example. )
_dtn-bundle._tcp.3 IN CNAME (
_dtn-bundle._tcp.ingress.space-agency.example. )
¶
The agency publishes the shared SRV RRset in its own zone:¶
$ORIGIN space-agency.example.
$TTL 3600
_dtn-bundle._tcp.ingress IN SRV (
10 1 4556 relay-a.space-agency.example. )
_dtn-bundle._tcp.ingress IN SRV (
10 1 4556 relay-b.space-agency.example. )
¶
The BP Agent sends the same query as in Example A. Ordinary DNS resolution follows the CNAME and returns the shared SRV RRset, which contains the same two relays. The remaining steps are the same as in Example A.¶
The agency maintains the relay selection in one SRV RRset. It can
change the relays in space-agency.example. without changes to the
allocator's zone.¶
The following DNS rules apply to the configurations above:¶
A CNAME aliases only its own owner name
(Section 3.6.2 of [RFC1034]). To redirect a service lookup, the
CNAME is at the complete service owner name, such as
_dtn-bundle._tcp.2.978879.ipn.arpa.. A CNAME at
2.978879.ipn.arpa. does not redirect the lookup.¶
A CNAME record and an SRV RRset cannot coexist at the same owner name (Section 3.6.2 of [RFC1034]).¶
An SRV Target other than "." must have address records and must not be an alias ([RFC2782]; Section 5.1 of [RFC6672]).¶
The discovery binding for another CL defines its query names, record types, processing rules, and fallback behavior relative to the DNS base name. If a binding uses SVCB records, [RFC9460] and the binding govern AliasMode, ServiceMode, parameter compatibility, endpoint selection, and fallback. SRV rules do not apply to other record types. For example, a ServiceMode SVCB TargetName of "." means the owner name of the record (Section 2.5.2 of [RFC9460]), not that the service is unavailable.¶
This document does not require a BP Agent to query for every CL that it supports, and it does not specify an order in which to try CLs.¶
If DNS does not meet the reliability or performance requirements of a deployment, particularly during disruption, the deployment can obtain forwarding information by other means.¶
Discovery succeeds when a BP Agent obtains usable connection information under the applicable discovery binding (Section 3.2, Section 3.5). A syntactically valid DNS answer does not always contain usable connection information. The binding and the record-type specification define candidate selection and compatibility checks.¶
A discovery result identifies candidate CL endpoints. A candidate can be the destination node or a relay that accepts Bundles for onward forwarding. A result does not show that the destination is reachable now, that a candidate will accept a session or a Bundle, or that the Bundle will be delivered.¶
DNS terminology [RFC2308] distinguishes these unsuccessful outcomes:¶
The name at which resolution stopped, possibly the target of an alias, does not exist.¶
The name exists, but it has no records of the requested type.¶
The lookup did not complete. This does not show that the service is absent.¶
These outcomes apply only to the DNS discovery attempt. They do not show that the FQNN is invalid or that other means cannot reach it. The absence of discovery records for one CL does not show the absence of service for another CL. Section 3.2 specifies the TCPCLv4 result when no usable SRV record exists.¶
An unsuccessful discovery attempt does not invalidate forwarding information from other sources, such as configured routes or other routing mechanisms. It does not require deletion of a Bundle. The existing policies of the BP Agent govern retries and route precedence.¶
Positive and negative DNS caching follow the existing DNS specifications, including [RFC2308].¶
A retained discovery result depends on the remaining TTLs of the records that produced it: aliases, SRV or SVCB records, and address records. These TTLs can be different. Copying a result into forwarding information should not silently make it permanent configuration. A BP Agent does not need to repeat the lookup before each Bundle transfer. TTL expiry does not require it to close an established CL session or to delete a Bundle; refreshed information affects later connection attempts.¶
If refresh is not possible, explicit operational policy governs any continued use of expired discovery information. That policy considers possible changes to endpoints, aliases, or administrative control. [RFC8767] describes how resolvers can serve stale DNS data; it does not specify how a BP Agent retains forwarding information.¶
TTL expiry is different from signature expiry. Validation checks the RRSIG validity period (Section 5.3.1 of [RFC4035]), and the signature expiration time limits the TTL of authenticated data (Section 5.3.3 of [RFC4035]). An earlier successful validation therefore does not stay valid indefinitely. Section 5.2.1 describes the handling of validation failures.¶
A deployment that uses a local instance of 0.ipn.arpa or of an Experimental Use allocator subtree (Section 7.2) configures it explicitly with ordinary DNS administration. The administrator loads the zones, selects authoritative servers, and configures resolvers to forward queries for those names. This document does not request automatic resolver behavior or additions to the IANA "Locally-Served DNS Zones" registry [RFC6303].¶
The signed public ipn.arpa namespace does not give a chain of trust to independently administered local data. If the public namespace does not contain a name, normal DNSSEC processing can authenticate the nonexistence of that name. A validator that uses only the public chain therefore cannot authenticate a conflicting local answer. A validator can authenticate a signed local zone with a configured local trust anchor (Section 7 of [RFC6303]). Section 5.10 of [RFC6840] describes how a validator selects among nested trust anchors. The trust anchor is necessary wherever validation occurs: in a trusted validating resolver (Section 5.2.1) or in a BP Agent that validates independently.¶
A deployment can also accept unsigned local data under explicit local policy. This acceptance is a policy decision, not a successful DNSSEC validation. Local resolution and trust settings are best limited to the relevant subtrees, without disabling validation for all of ipn.arpa.¶
Administrators should configure resolution so that queries for these names stay within the intended administrative or experimental scope, also when local resolution fails. Public resolution cannot supply the local data, and queries that reach the public DNS can disclose the queried identifiers (Section 5.4). [RFC6303] describes the reasons to contain such queries. This guidance does not require resolvers in general to recognize these namespaces.¶
An Allocator can hold a range of Allocator Identifiers
(Section 3.2.1 of [RFC9758]). A numeric range does not map to one
DNS subtree: each Allocator Identifier A has its own name,
A.ipn.arpa., directly beneath ipn.arpa. An operator that administers
several identifiers can delegate a zone for each identifier.
Alternatively, it can use DNAME records [RFC6672] to redirect each
allocator subtree into one zone that it manages.¶
In the following example, the agency of Section 3.3 also administers Allocator Identifiers 974848 and 974849. The relay address records are those in Figure 1.¶
$ORIGIN ipn.arpa. $TTL 3600 974848 IN DNAME 974848.allocators.space-agency.example. 974849 IN DNAME 974849.allocators.space-agency.example.¶
$ORIGIN 974848.allocators.space-agency.example. $TTL 3600 _dtn-bundle._tcp.2 IN SRV 10 1 4556 relay-a.space-agency.example. _dtn-bundle._tcp.2 IN SRV 10 1 4556 relay-b.space-agency.example. $ORIGIN 974849.allocators.space-agency.example. _dtn-bundle._tcp.2 IN SRV 10 1 4556 relay-a.space-agency.example. _dtn-bundle._tcp.2 IN SRV 10 1 4556 relay-b.space-agency.example.¶
A BP Agent that discovers TCPCLv4 service for ipn:974848.2.345
constructs the usual query name (Section 3.2):¶
_dtn-bundle._tcp.2.974848.ipn.arpa.¶
The DNAME at 974848.ipn.arpa. redirects the query to:¶
_dtn-bundle._tcp.2.974848.allocators.space-agency.example.¶
There, the BP Agent finds the same relay SRV records as in Section 3.3. Ordinary DNS processing does the redirection (Section 3 of [RFC6672]). The BP Agent does not construct names in the agency's zone.¶
Each target subtree keeps its Allocator Identifier label. Thus (974848,2) and (974849,2) stay distinct, although their SRV records select the same relays. Both FQNNs are also distinct from (978879,2), the mars-orbiter. If both DNAMEs pointed to the same target, queries for (974848,2) and (974849,2) would reach the same records. Distinct target subtrees are therefore necessary when the allocator namespaces must stay distinct. The decimal allocator label in the target is a convenient convention, not a DNS requirement.¶
One zone can contain all of these subtrees. Each redirected Allocator Identifier still needs its own DNAME in ipn.arpa. The technique reduces the number of child zones to administer, not the number of allocator entries in the parent.¶
The following DNAME constraints apply:¶
A DNAME redirects the descendants of its owner name, not the owner name itself (Section 2.3 of [RFC6672]).¶
The source zone cannot contain records beneath the DNAME owner
(Section 2.4 of [RFC6672]). The data for names beneath
974848.ipn.arpa. is in the target subtree.¶
A DNAME rewrites names beneath its owner, but an NS delegation creates a zone cut. A parent-side DNAME cannot coexist with a delegation NS RRset at the same owner name (Section 2.3 of [RFC6672]). A delegated allocator zone can instead publish a DNAME at its own apex, with its SOA and NS records. This arrangement keeps separate child-zone administration.¶
Section 5.3 of [RFC6672] describes DNSSEC considerations for DNAME.
Locally served instances (Section 4.3) can use DNAMEs in the
same way, for example at N.0.ipn.arpa., if the administrator of that
namespace authorizes them.¶
Implementation-specific tools, such as the $GENERATE directive of
BIND [BIND-ARM], can produce repetitive records. For example, this
BIND input produces the two DNAME records above:¶
$ORIGIN ipn.arpa. $TTL 3600 $GENERATE 974848-974849 $ DNAME $.allocators.space-agency.example.¶
The directive expands into ordinary records. It adds no numeric-range matching to DNS, and resolvers need no support for it. The use of such tools is optional.¶
The use of DNS-derived forwarding information involves two separate authentication questions. DNSSEC authenticates the DNS data (Section 5.2). The CL authenticates the session peer (Section 5.3). The authentication of one does not authenticate the other.¶
The configured forwarding and security policies of the BP Agent govern the use of DNS-derived forwarding information. These policies can distinguish sources of forwarding information, for example configured routes and discovery results, or authenticated and unauthenticated results. A BP Agent's configured forwarding policy may permit automatic use of authenticated discovery mappings, subject to the security requirements of the applicable CL.¶
Forged discovery information can redirect connection attempts, and thus bundle transfers, to endpoints that an attacker selects. DNSSEC [RFC9364] gives data origin authentication and integrity for DNS data. It does not show that the advertised service is available; ENUM makes a similar observation (Section 7.1 of [RFC6116]).¶
Operators publishing discovery data for this mechanism in the public DNS SHOULD provide DNSSEC authentication of that data, including the necessary chain of trust. This recommendation applies to all DNS records relied upon in discovery, including records published outside ipn.arpa:¶
aliases and redirections, such as CNAME, DNAME, and SVCB AliasMode records;¶
service records and other records that carry connection parameters, such as SRV and SVCB ServiceMode records; and¶
address records for the resulting targets.¶
The ipn.arpa operator signs ipn.arpa (Section 7.1), and DS records can link delegated allocator zones into that chain of trust (Section 7.3.3). The signature on ipn.arpa does not authenticate records in delegated zones or in external target namespaces. Each record has its own security status.¶
Validation follows standard DNSSEC processing. For example, a CNAME synthesized from a validated DNAME needs no signature of its own (Section 3.1 of [RFC6672]; Section 5.3.1 of [RFC6672]).¶
BP Agents using public DNS discovery SHOULD use DNSSEC validation, either locally or through a trusted validating resolver. When a BP Agent relies on validation by another resolver, the resolver and the channel to it need to satisfy the trust requirements of Section 4.9.3 of [RFC4035].¶
Section 4.3 of [RFC4035] defines the security states secure, insecure, bogus, and indeterminate. An insecure result means that the validator has established that no chain of trust exists from a configured trust anchor to the RRset. For example, normal DNSSEC processing proves that the delegation to the target zone is unsigned. An insecure result does not mean that the validator failed to build a chain. Missing signatures where a secure chain requires them can produce a bogus result. A zone can contain signatures and still be insecure if no chain of trust reaches it.¶
An application can also receive a result whose security status is unknown to it, for example from a resolver that does not report validation results. The absence of an authentication indication does not establish that a result is insecure.¶
This document does not require a policy of accepting only secure results. The configured policy of the BP Agent governs the acceptance of insecure results and of results with unknown status.¶
A known validation failure is a failure of the affected discovery path (Section 4.1). It should not lead to automatic acceptance of the same data without validation, for example through a query on a non-validating path. A BP Agent can still retrieve DNSSEC data to validate it independently. Section 4.3 describes validation for locally served namespaces.¶
The applicable CL specification and the configured security policy govern authentication of the CL peer. For TCPCLv4, Section 4.4 of [TCPCL] specifies session security, including certificate-based peer identification, and Section 4.6 of [TCPCL] specifies the handling of the peer Node ID. This document does not change these procedures.¶
A TCPCLv4 discovery result involves three different identifiers. In Example A (Section 3.3):¶
The discovery name, _dtn-bundle._tcp.2.978879.ipn.arpa., comes
from the destination FQNN.¶
The selected SRV Target, relay-a.space-agency.example., is the DNS
name that the BP Agent resolves to make the TCP connection. If the
configured policy invokes network-level authentication
(Section 4.4.4.2 of [TCPCL]), this name is the DNS-ID reference
identifier. It is also the HostName in the TLS server_name
extension (Section 4.4.3 of [TCPCL]).¶
The peer Node ID is the Node ID of the relay, which the relay sends in its SESS_INIT message (Section 4.6 of [TCPCL]).¶
Network-level authentication of relay-a.space-agency.example.
authenticates the host name only. It does not authenticate the Node
ID of the relay. The destination EID of the Bundle stays
ipn:978879.2.345. Discovery does not supply an expected peer Node
ID. Section 4.6 of [TCPCL] describes the handling of a peer Node ID
that is different from the intended Node ID or that is not
authenticated.¶
Public discovery records can reveal destination identifiers, their association with services and relays, and connection endpoints. Queries for these records can reveal to resolvers and other observers which destinations a BP Agent tries to reach, and when. DNSSEC does not provide confidentiality. [RFC9076] describes DNS privacy threats and mitigations, such as resolver selection and encrypted DNS transports. Operators can limit what published records disclose (Section 3.3.3), and they can keep queries for locally served names within their intended scope (Section 4.3).¶
Bundles addressed to many distinct FQNNs, whether malicious or not, can cause many DNS queries, much retained discovery state, and many connection attempts. Implementations SHOULD bound outstanding discovery work, retained discovery state, and connection attempts, using existing mechanisms such as resolver limits, queueing, and connection management.¶
In accordance with Section 3 of [RFC3172], the IAB is requested to approve the zone "ipn.arpa" as an infrastructure namespace derived from IPN identifiers. Names in this zone are constructed from FQNNs (Section 3.1). BP Agents use them to discover CL endpoints for reaching IPN nodes.¶
This follows the precedent of e164.arpa [RFC6116], in which names are derived from E.164 telephone numbers and queried directly for service records (NAPTR).¶
Following approval by the IAB in accordance with Section 3 of [RFC3172], IANA is requested to add "ipn.arpa" to the ".arpa" domain. Section 3.1 specifies the names in ipn.arpa.¶
IANA is requested to arrange the ongoing administration of ipn.arpa, consistent with the arrangements that the IAB approves under [RFC3172]. Administration includes the processing of the requests in Section 7.3. Administration is separate from the operation of the authoritative DNS servers for ipn.arpa, and this document does not require IANA itself to operate those servers. The ipn.arpa operator does the DNS provisioning and operational functions under these arrangements.¶
The ipn.arpa operator MUST sign the public ipn.arpa zone using DNSSEC and maintain a chain of trust through the arpa zone. This requirement applies only to the ipn.arpa zone. Section 5.2 gives recommendations for the operators of delegated allocator zones and other namespaces.¶
This document restricts the publication of discovery data for two sets of names in the public ipn.arpa hierarchy. These restrictions come from this document, not from [RFC9758]. They apply only to DNS discovery, and they do not change any allocation or its Bundle Protocol semantics.¶
The restrictions apply to every means of placing discovery data for the affected FQNNs in the public ipn.arpa hierarchy:¶
discovery records, such as SRV, SVCB, or CNAME records, at or beneath the restricted names;¶
NS delegations of the restricted names, or of names beneath them, to servers that publish such discovery data; and¶
CNAME or DNAME records at or beneath the restricted names that redirect discovery queries to data published elsewhere in the global DNS.¶
Names in these subtrees are absent from the public zone. They receive ordinary negative answers with normal DNSSEC processing. This document requests no special public delegation for them. For comparison, home.arpa has an insecure public delegation to accommodate validators without local trust configuration (Section 6.2 of [RFC8375]; Section 7 of [RFC8375]). This document does not request that arrangement. Section 4.3 gives operational guidance for local instances.¶
"0.ipn.arpa" corresponds to Allocator Identifier zero, the Default Allocator (Section 3.2.2 of [RFC9758]). Public DNS discovery for FQNNs with Allocator Identifier zero is intentionally outside the scope of this document. A populated "0.ipn.arpa" zone MUST NOT be published in the public ipn.arpa hierarchy specified by this document. A "0.ipn.arpa" zone MAY exist in local DNS resolvers or servers that serve it authoritatively as a locally served zone [RFC6303], much as the IN-ADDR.ARPA zones for the private address space of [RFC1918] are served locally (Section 4.1 of [RFC6303]).¶
The restriction covers every Node Number under Allocator Identifier zero. This includes the Private Use Node Numbers 1 through 16383 (Section 3.4.3 of [RFC9758]) and the Node Numbers 2097152 through 268435455 reserved for Private or Experimental Use (Section 3.2.1 of [RFC7116]; Section 9.2 of [RFC9758]).¶
The restriction does not mean that Allocator Identifier zero has only local significance. The registry of the Default Allocator (Section 9.2 of [RFC9758]) also contains assigned Node Numbers. Public support would need administration and delegation procedures for these allocations, and a treatment of the private ranges and of locally served DNS data.¶
Note to the Working Group (to be removed before publication as an RFC): The working group can reconsider public DNS support for Allocator Identifier zero if it chooses to take on that scope.¶
Section 9.1 of [RFC9758] designates Allocator Identifiers 1073741824 through 2147483647 (0x40000000 through 0x7FFFFFFF) as Experimental Use. Experimental Use values are not assigned to any party (Section 4.2 of [RFC8126]), so independent experiments can use the same identifier. A globally authoritative mapping to discovery data is therefore not appropriate.¶
For each Experimental Use Allocator Identifier A, a populated subtree beneath "A.ipn.arpa" MUST NOT be published in the public ipn.arpa hierarchy. Names beneath "A.ipn.arpa" MAY be served by local DNS resolvers or servers within the scope of an experiment.¶
IANA and the ipn.arpa operator are requested to maintain a procedure that links requests for DNS provisioning at "A.ipn.arpa" to the entry for Allocator Identifier A in the "'ipn' Scheme URI Allocator Identifiers" registry (Section 9.1 of [RFC9758]). IANA and the ipn.arpa operator determine the storage format, the request interface, and the presentation of provisioning information.¶
DNS provisioning is optional. It can be requested when an Allocator Identifier is assigned or at any later time. A DNS entry is not a condition for the validity or use of an allocator assignment.¶
Eligible Allocator Identifiers are those assigned for operational use in the registry, subject to the exclusions in this document. This document does not request public provisioning for:¶
Allocator Identifier zero (Section 7.2.1);¶
Experimental Use Allocator Identifiers (Section 7.2.2);¶
unassigned or reserved Allocator Identifiers; or¶
Allocator Identifiers 974848 through 978943, which are reserved for documentation examples (Section 9.1 of [RFC9758]).¶
The authorized DNS administrators of the allocator manage the records beneath the allocator name, such as node-level discovery records and further delegations. These procedures do not require IANA or the ipn.arpa operator to provision individual Node Numbers.¶
The parent-zone configuration at "A.ipn.arpa" is one of the following:¶
The requester supplies the domain names of the authoritative nameservers, any addresses needed for glue, and any DS records (Section 7.3.3).¶
The requester supplies the fully qualified target domain name (Section 4.4).¶
A DNAME at the allocator name does not coexist with an NS delegation, or with the DS records of a delegation, at that name (Section 2.3 of [RFC6672]).¶
One request can cover several Allocator Identifiers within the allocation of the requester. The request identifies each Allocator Identifier and its configuration. Each allocator name still has its own parent-zone records; DNS has no numeric-range syntax.¶
The ipn.arpa operator MUST support authenticated requests to add, update, and remove DS records for delegated allocator zones. Section 7.3.1 applies to these requests. The operator of a delegated allocator zone coordinates changes to its signing keys with updates to the DS records in ipn.arpa.¶
This support enables secure delegations. It does not require that every delegated allocator zone be signed.¶
DS records apply only to NS delegations. A DNAME in ipn.arpa is not a delegation. The ipn.arpa operator signs it as ordinary data in the ipn.arpa zone, and the target namespace has its own DNSSEC chain of trust.¶
The allocation procedures of Section 9.1 of [RFC9758], including Expert Review, continue to apply to numeric assignments of Allocator Identifiers (Section 7.3.1).¶
The ipn.arpa operator does the applicable DNS technical checks on provisioning requests under its published operational procedures. The technical requirements of IANA for delegations in the zones that it manages [IANA-NS-REQS] are relevant background. DNAME provisioning is subject to the requirements of [RFC6672].¶