IETF Drafts & RFCs

Explicit Congestion Notification (ECN) applies in principle to all transport protocols, but UDP saw little ECN deployment until QUIC became widely adopted, so documentation of UDP socket ECN APIs across platforms is sparse. This document records experiments to get ECN working over UDP in Chromium on Apple, Linux, and Windows platforms. It is a snapshot of platform support at the time of writing, and readers should consult current platform documentation for updates.

Defines YANG data models for managing the RPKI-to-Router (RtR) protocol specified in RFC 6810 and RFC 8210. The models give operators a standard, structured way to configure and monitor the sessions over which relying parties feed validated prefix-origin data to BGP speakers, improving automation and consistency of RPKI-based route validation deployments.

Describes an RDAP extension supporting an extensible set of versioning types. A server can advertise which extension versions it supports, a response can indicate which versions it includes, and a client can request specific versions in a query. It also lets servers publish versioning and deprecation information so transitions between extension versions are coordinated and breaking changes disrupt deployed clients less.

Delivering network services over a Layer 3 tunnel requires proper setup on the links between customer termination points and the provider network, called an attachment circuit (AC), where the underlying link carrying the services is the bearer. Here the bearer is a Layer 3 UDP tunnel. The document extends the YANG service data model for ACs so a UDP tunnel can serve as the Layer 3 bearer.

By its name, this LISP working group draft addresses NAT traversal for the Locator/ID Separation Protocol, covering how LISP xTRs behind network address translators can establish mappings and forward encapsulated traffic. The official abstract could not be retrieved on this run; this summary is approximate and will be grounded from the document's abstract on the next run.

Defines four Post-Quantum/Traditional (PQ/T) hybrid key exchange methods for the Secure Shell (SSH) protocol. They combine the Scloud+ key-encapsulation mechanism, whose security relies on the unstructured Learning With Errors (LWE) problem, with Elliptic Curve Diffie-Hellman over the NIST P-384 and P-521 curves. The aim is to give SSH key exchange post-quantum protection while retaining the assurance of established elliptic-curve Diffie-Hellman.

By its name, this is a revision (bis) of RFC 8111, the LISP Delegated Database Tree (DDT) / Map-Server interface specification, updating how LISP mapping-system queries and referrals are handled. The official abstract could not be retrieved on this run, so this description is approximate and will be grounded from the abstract on the next run.

Updates RFC 7518 to deprecate two algorithms with known security weaknesses: the JWS "none" algorithm and the JWE "RSA1_5" (RSAES-PKCS1-v1_5) algorithm. It also revises the Designated Experts' review instructions so future JOSE algorithm registrations must meet baseline security requirements, reducing the risk that insecure algorithms remain available in deployed JOSE implementations.

Describes use cases for agentic AI communication systems and derives protocol requirements from them. The requirements are intended to guide IETF standardization of protocols for agent-to-agent and agent-to-tool communication, with focus on dialog management, transport, multimodal communication, communication security, and agent identity and authentication.

Examines network-service scenarios where fully homomorphic encryption (FHE) can meet privacy needs existing non-HE solutions do not meet under the stated trust conditions. It covers AI inference, integrated sensing and communication (ISAC), and private information retrieval (PIR), describing each scenario, example flows, post-conditions, existing solutions, and potential requirements. It then summarizes common gaps and evaluation criteria but specifies no protocol or cryptographic scheme.

Defines a YANG data model for Ethernet networks used as a client-layer network of an underlay transport network such as an Optical Transport Network (OTN), and also for cases where Ethernet itself is the transport network. The model supports traffic-engineered topology representation for Ethernet client services in multi-layer transport deployments.

An informational DAWN draft describing a framework for discovering the minimum public information about AI agents before interacting with them, on a local network, within an organization, or between cooperating parties over the Internet. It proposes a two-layer federated architecture: a Local Discovery Plane for zero-configuration advertisement within a site, and a Federation Plane where gateways exchange integrity-protected discovery records across domains. An Export Policy Engine controls what leaves each site; the preferred method publishes DNS records verified with DNSSEC.

Defines nine TLS 1.3 key exchange algorithms based on HQC-KEM, a code-based KEM NIST selected for standardization in forthcoming FIPS 207: three standalone HQC-KEM variants, two PQ/T hybrids pairing HQC-KEM with X25519 or X448, two PQ/PQ hybrids with ML-KEM, and two PQ/PQ/T hybrids combining HQC-KEM and ML-KEM with X25519 or X448. HQC's different hardness assumptions give algorithmic diversity against lattice cryptanalysis. The algorithms target TLS 1.3, DTLS 1.3, and QUIC.

Enhanced VPNs provide VPN services with guaranteed characteristics such as resources, latency, and jitter. A Network Resource Partition (NRP) is a subset of network resources and policies that can be the underlay for one or more Enhanced VPN services. Segment Routing segments can be tied to specific resources, creating resource-aware SIDs; a group of these builds SR-based NRPs with the required topology and resource attributes. The document describes how to build such SR-based NRPs.

By its name, this draft defines a YANG data model for a multi-interface IGMP/MLD proxy, extending IGMP/MLD proxy management across several interfaces for multicast forwarding. The official abstract could not be retrieved on this run, so this description is approximate and will be grounded from the abstract on the next run.

Conformance tests often assume a published test vector is correct, so a flawed vector can wrongly penalize a correct implementation or approve a faulty one. It introduces Verification Defect Quarantine (VDQ), a model tracking the tested subject's status separately from the verifier's fault-detection status. VDQ records reproducible counterexamples, quarantines affected vectors from promoting assurance, keeps unresolved cases INDETERMINATE, and supports append-only corrections. It does not change RFCs or replace the errata process.

By its name, this PIM working group draft defines a YANG data model for EVPN multicast, giving a standard structure for configuring and monitoring multicast forwarding in Ethernet VPN deployments. The official abstract could not be retrieved on this run, so this description is approximate and will be grounded from the abstract on the next run.

By its name, this PIM working group draft defines a YANG data model extension for IGMP and MLD snooping in L2VPN contexts, extending snooping management to Layer 2 VPN services. The official abstract could not be retrieved on this run, so this description is approximate and will be grounded from the abstract on the next run.

Offers guidance on adding post-quantum cryptography to resource-constrained devices such as IoT nodes and dedicated key-storage hardware, which have tight limits on processing power, RAM, and flash and may run on batteries. It stresses hardware security as the foundation, and covers storing seeds rather than full private keys, handling short-lived ephemeral keys efficiently, offloading cryptographic work, and how PQC affects firmware-update mechanisms.

Introduces Verification Promotion Boundaries (VPB), a protocol-neutral profile for government and critical-infrastructure relying parties, to stop a verification result from one assurance stage being treated as proof of a stronger claim. It defines six ordered stages where passing one does not satisfy the next, states that no stage implies endorsement, separates negative findings about the artifact from positive findings that the verifier caught a defect, and makes arithmetic and protocol checks mandatory before any promotion.

Specifies RTP/SDP signaling for Opus multistream (multichannel) audio, letting applications negotiate layouts such as 5.1 and 7.1 in real-time communications. It defines SDP format parameters describing these configurations and Offer/Answer procedures for interoperable negotiation. It defines no new Opus codec behavior, continuing to use the RTP payload format from RFC 7587 and channel mapping from RFC 7845, and aims to standardize an alternative to deployment-specific "multiopus" approaches.

Proposes the Agent Decision Commitment (ADC), a way to record what information an AI payment agent relied on when it chose a payment, since current protocols show who authorized a payment but not the basis for the choice. ADC stores a salted hash of a decision record covering model inputs, model calls, and the resulting answer and action, attaches the commitment to the payment mandate, and by default registers it with a transparency service before authorization. The full record is released only to dispute participants.

An agent can authenticate and hold valid authorization yet still be unsafe to execute, due to missing evidence, an insufficient assurance stage, a stale authority epoch, an unresolved revocation, or an unmet human-approval requirement. It introduces the High-Assurance Agent Control Envelope (HAACE), a protocol-neutral profile binding an exact action to identity, authorization, policy, assurance, evidence, epoch, revocation, and oversight state, with fail-closed handling and non-bearer receipts. It complements OAuth, workload identity, and attestation.

Addresses a gap in SCIM: existing events report that a User was created, changed, activated, deactivated, or deleted, but none can say that one User has been retired in favor of another representing the same subject. It defines two SET-based provisioning events, one signaling that supersession and one correcting an earlier supersession. Supersession stays separate from attribute merging, and matching, identity proofing, and non-SCIM data are out of scope.

Presents an operator-perspective framework for building and running IPv6-only core networks spanning multiple interconnected Autonomous Systems. To carry remaining IPv4 traffic it proposes stateless IPv4/IPv6 address mapping as the basis for IPv4-as-a-Service, translating IPv4 packets at the network edge and forwarding them across the IPv6-only core without per-flow state. It is a problem statement, guidance, and requirements document covering applicability, trust boundaries, prefix allocation, and operational and security considerations.

Sets out standards for automatically configuring core infrastructure services on Autonomic Control Plane (ACP) nodes using GRASP. These services mainly bootstrap a network but continue running throughout its lifecycle, including secure remote access (SSH/NETCONF) to zero-touch bootstrapped ANI devices with RADIUS/Diameter authentication and authorization, plus ongoing autoconfiguration of syslog, NTP, and DNS for operations.

Describes how to use the GeneRic Autonomic Signaling Protocol (GRASP) for service announcement and discovery while reusing DNS Service Discovery (DNS-SD) features such as service names, instance names, and priority/weight selection. This avoids duplicating that work and keeps GRASP simple. The document also defines name discovery and reusable elements for GRASP objectives.

Describes a YANG data module for managing Service Level Agreement (SLA) assurance in optical transport networks. The module offers a standard way to record and report problems that could affect service and network availability, and gives consistent reporting of impairments and risks across optical transport domains. It is intended to support closed-loop operations where automated monitoring, analysis, and remediation help keep services reliable and SLA-compliant.

A practical guide to rolling out IPv6-only service on the user plane of 5G networks using 464XLAT, a mechanism for translating between IPv4 and IPv6. It also covers core 5G concepts and architectures, configuration methods, and operational challenges, aimed at helping mobile operators deploy IPv6-only user planes while still serving IPv4 content.

Addresses a gap in locally stored, individually signed records: each record verifies alone, but the collection can have records deleted, reordered, or added undetected. It specifies the Checkpointed Local Log (CLL), an append-only log kept by the producer as a Merkle Mountain Range under the COSE Receipts profile. A signed checkpoint commits to the whole history and is packaged as a SCITT Signed Statement, so independent Transparency Services can register it unchanged, turning point receipts into a stream with provable order, contemporaneity, and completeness.

Protocol implementations often mix specification checks, test vectors, security testing, independent reproduction, and release decisions, producing "evidence collapse" where a result at one assurance level is treated as proof of a stronger claim. It introduces Verification Promotion Boundaries (VPB), a machine-readable profile and state-transition procedure that keeps assurance levels separate under the rule "Unknown evidence never promotes." It defines no new cryptographic primitive, authorization protocol, or IETF certification program.

Proposes Agentic Authority Graph Validation (AAGV), a JSON interchange format and deterministic validation procedure for authority graphs. It models where authority starts, where consequential actions take effect, expected normal routes, revocable edge sets meant to cut off access, and each edge's current state. It rejects incomplete or vacuous graphs and returns indeterminate when relevant edge state is unknown. It does not define an authorization token, find hidden paths, or prove physical containment, distinguishing structural cut validity from operational closure.

Defines combinations of US NIST ML-KEM in hybrid with the traditional algorithms RSA-OAEP, ECDH, X25519, and X448, designed to meet security best practices and regulatory guidance. Composite ML-KEM applies to any application using X.509 or PKIX data structures that accept ML-KEM, and is most useful when an operator wants extra protection against breaks or catastrophic bugs in ML-KEM while post-quantum migration proceeds.

Describes how to send Forward Error Correction (FEC) repair data alongside source media in Media over QUIC (MoQ) sessions, so receivers recover from loss without waiting for retransmissions. It covers catalog-based signaling of FEC configuration (so receivers derive RaptorQ parameters without in-session messages), naming conventions and object formats for repair tracks, an ISO/IEC 23008-1 block profile, RaptorQ (RFC 6330) as the mandatory scheme with a registry for others, and compatibility with ATSC 3.0 and ARIB STD-B60 broadcast systems.

Defines how Media over QUIC (MoQ) works with IP multicast, including source-specific and any-source multicast, Automatic Multicast Tunneling (AMT), and TreeDN for scalable live streaming. It adds a multicast catalog extension for discovering endpoints and delivering across TV, mobile, and browser platforms. All multicast delivery uses MMTP packets, the same format used on MoQ QUIC streams and datagrams, and it introduces a manifest-based content-authentication profile to protect multicast delivery.

Specifies the use of MPEG Media Transport (MMT) as a container format for Media over QUIC (MoQ), explaining how MMT packets map to MoQ objects. It also covers how MMT works with Application-Layer FEC, how ATSC S-TSID signaling converts to and from MoQ catalogs, and how the approach fits ATSC 3.0 and ARIB STD-B60 broadcast systems across broadcast, multicast, and unicast delivery.

The JMAP base protocol (RFC 8620) lets clients upload and download arbitrary binary data ("blobs") over HTTP. This extension adds ways to create and access blobs through inline method calls within a JMAP request, a reverse lookup to find which objects reference a blob, support for large blobs built from chunks with server-side optimization, and server-side conversions such as image format changes, archive creation/extraction, compression/decompression, and delta/patch operations.

Extends HTTP Message Signatures (RFC 9421) for requests automated agents send for people. RFC 9421 identifies the signer but not whom the agent acts for or what it may do. The document defines an Agent-Delegation header carrying a chain of delegation links, each able only to narrow the scope, floors, and expiry of the previous link, with a verifier checking every link. Floors cover who stands behind the agent (via an Attestation Issuer) and what the agent may do (calls classed as r, w, or x). It defines no credential format, identity system, or revocation.

Proposes "dns-persist-01," a new ACME validation method where a Certification Authority verifies domain control by checking for a long-lived DNS TXT record identifying the CA and the ACME account. It targets settings where standard challenges are hard to use, such as multi-tenant hosting, enterprise DNS, and IoT. It is designed for security and robustness and meant to suit CAs under different policy regimes, including the CA/Browser Forum Baseline Requirements.

Proposes a single architecture for governing autonomous systems, whose core principle is that such systems must not act beyond their authorized task and grant. Consequential actions would run on a dedicated agents-only access plane with signed receipts proving execution, while HTTPS stays the human-facing channel. The design includes controls for freezing an individual system, freezing a network, and revoking an identity. It is an individual Experimental Internet-Draft regulating the channel agents act through rather than banning AI.

Describes TACET, a transparency map whose main output is a portable, offline-verifiable proof of absence over time: for a given subject, a proof that no record matching a public predicate appeared in the map or on the subject's monitored public surfaces, with the epoch range bounded below by a public randomness beacon and above by a Bitcoin block. It specifies a depth-256 sparse Merkle tree, signed epoch sheets, surface snapshots, and silence proofs at three strength levels. PNX applies the same construction to an AI agent's outbound traffic.

Describes a framework and protocol extensions for spreading routing-risk information and automating responses in BGP networks, building on RPKI Route Origin Validation. When a hijacked or abnormal route is found, the framework links it to related ASes or BGP peers so their paths can be deprioritized. It extends the RPKI-to-Router (RTR) protocol and BGP path attributes so risk indicators flow between relying parties and routers, and between routers, for coordinated cross-boundary mitigation.

Introduces the "iqa" URI scheme, which names the attestation standing of an identified subject as reported by one of three named authority organs. The URI carries no proof or credential, and no lookup or name-resolution service is consulted when it is used. The document sets out requirements for clients handling these URIs, addressing three risks: using the authority as a redirect target, using a registered protocol handler as a general-purpose launcher, and treating a parsed URI as a certification.

Specifies a small JSON attestation receipt for an AI inference decision. The receipt binds an outcome to a request hash, signed under a published Ed25519 public key, so a party that does not trust the issuer's infrastructure can verify offline that the issuer's signing key attested to the decision. The format is kept compact and version-selected so independent verifiers are easy to build and audit, for settings where an operator-controlled log alone is not sufficient evidence.

TLS handshakes normally authenticate peers with long-term credentials, but some cases also need assurance that a peer's runtime environment is secure. The document presents remote attestation, in which an entity produces Evidence about itself, and defines a TLS extension binding the TLS authentication key to a remote attestation session. This lets confidential workloads in a Trusted Execution Environment, or IoT devices, present fuller security metrics to peers. The design works with any attestation technology and topology and supports mutual attestation.

Standard RPKI Route Origin Validation confirms whether an AS may originate a prefix but does not catch an attacker in another region announcing a prefix the same origin AS is authorized to announce. It introduces "Regionalized ROA" (R-ROA), which attaches regional scope to ROAs, proposes RPKI-Router protocol extensions to carry that regional data from Trust Anchors to routers, and defines enhanced validation to detect and mitigate intra-AS cross-regional route hijacking.

Argues that "early attestation," where attestation happens inside the TLS handshake, is unsafe and unnecessarily complex. A formal analysis in the ProVerif tool, together with published CVEs and GitHub Security Advisories, shows such implementations can be exploited through relay attacks even without physical access to the target. Because continuous attestation is generally needed anyway, early attestation adds complexity without corresponding benefit. The draft cites CVE/GHSA identifiers with CVSS scores from 6.3 to 9.1 and names implementations it considers vulnerable; its artifacts are to be released under Apache-2.0. (Vendor and CVE claims are the authors'.)

Describes how software workloads authenticate each other over HTTP using HTTP Signatures, as part of the WIMSE architecture, relying on the Workload Identity Token. The method provides end-to-end protection of requests, and optionally responses, even when TLS proxies or load balancers sit between services. It applies only to HTTP traffic.

Updates the AUTH48 process described in RFC 7841. It adds automated validation milestones and access controls within the IETF Datatracker to block late technical changes after Working Group Last Call, in order to protect rough consensus. The document formally updates RFC 7841.

Defines the "rttp" URI scheme, which expresses an intended action toward a specific subject, where the subject's address is computed directly from the URI's authority component so no directory, registry, or DNS-style lookup is needed. It sets out required client behavior to prevent two misuses: treating the URI as a redirect destination, and letting a protocol handler open arbitrary URIs for any page.

Extends PPP's IPCP so a client can negotiate encrypted DNS resolver settings. It defines two configuration options, one for a primary resolver and one for a secondary, each carrying an Authentication Domain Name, a list of resolver IP addresses, and service parameters in RFC 9460 wire format. It covers DNS over TLS, DNS over HTTPS, and DNS over QUIC, and remains compatible with plaintext DNS configuration under RFC 1877.

Describes the file format used by the libpcap library to record captured packets to a file. Programs such as tcpdump use libpcap to read and write files in this format; the document standardizes the long-used "pcap" format so tools can interoperate reliably when exchanging packet captures.

Defines a YANG data model for network incident lifecycle management. The module offers a standard way to report, diagnose, and help reduce troubleshooting tickets and to resolve network incidents, in support of network service health and probable root-cause analysis, enabling more automated and consistent incident handling across management systems.

STAMP and its optional extensions can be used for Edge-to-Edge (E2E) active measurements, and In Situ OAM (IOAM) data fields can record and collect Hop-by-Hop (HBH) and E2E operational and telemetry information. The document extends STAMP to reflect IPv4/IPv6 headers as well as IPv6 extension headers, so HBH and E2E measurements such as those using IOAM data fields can be carried out.

Defines a deterministic conformance profile (eqr-core-1) and test vectors for Evidence Qualification Receipts. A closed algorithm checks receipts for JSON structure, SHA-256 digest encoding, UTC timestamps, evidence-reference integrity, requirement-state derivation, and a three-valued result (PASS, FAIL, INDETERMINATE). The profile is deliberately narrow: passing a check does not show evidence is accurate, that a scientific claim holds, that an issuer is trustworthy, or that an action is authorized; its goal is interoperable checking of receipt semantics.

Describes a minimal JSON evidence-record profile for checking artifact digests. It separates integrity verification from scientific assessment and defines a record structure, a validation algorithm, and negative test cases. It explicitly does not assert that an integrity-checked scientific claim is true, limiting its scope to verifiable integrity rather than scientific validity.

Segment Routing over IPv6 (SRv6) is the instantiation of Segment Routing on the IPv6 data plane, using IPv6 addresses to identify segments. The document discusses security considerations for SRv6 networks, including potential threats and possible mitigation methods. It does not define any new security protocols or extensions to existing protocols, serving instead as consolidated guidance for securing SRv6 deployments.

The Extensible Provisioning Protocol (EPP) carries DNSSEC parameters for domain names to registries, but registries and registrars restrict accepted values differently, so DNS operators struggle to predict which new algorithms will be accepted, slowing adoption. The draft recommends that registries and registrars base their acceptance policies on the IANA registries for DNSSEC algorithms and digest types, to make acceptance more predictable and speed algorithm rollout.

States that the Automatic Extended Route Optimization (AERO) and Overlay Multilink Network (OMNI) Interface functional specifications have reached a maturity ready for advancement in the RFC publication process. Updates to the base specifications are documented in this first amendment (AMEN) and any additional future amendments as necessary.

Defines Hamming Quasi-Cyclic (HQC) ciphersuites for use with COSE-HPKE, specifying twelve fully specified COSE algorithms for two modes (Integrated Encryption and Key Encryption). These use either pure HQC or a hybrid of HQC with a traditional elliptic-curve KEM, paired with the SHAKE256 key derivation function and AES-256-GCM. It also specifies the AKP key representation; message processing and context binding come from COSE-HPKE.

Extends the MOQT Streaming Format with an "mpeg2ts" packaging for carrying MPEG-2 Transport Stream packets over MOQT. It defines the catalog fields that describe such a track and the rules publishers and subscribers follow to carry the packet stream, enabling MPEG-2 TS content to be transported within Media over QUIC.

Explains the role of Endorsements in the IETF RATS architecture (RFC 9334), where a Verifier combines Evidence, Appraisal Policy, Endorsements, and Reference Values to produce Attestation Results for Relying Parties. It discusses factors in choosing a message format for Endorsements but does not define one. It updates RFC 9334 by classifying RATS conceptual messages as actual state, reference state, or appraisal policy, and adds guidance on conditionally endorsed values, endorsing verification keys, timeliness, and multiple Endorsements.

Defines the terminology for phrases such as "IPv6-Only" and "IPv6-Mostly" so they are used consistently in IETF and other documents. Its goal is that a reference to IPv6-Only describes the functionality actually in use within a given scope, not merely the protocol support that is installed, reducing ambiguity in operational and standards discussions.

Defines a well-known URI, "knowledge-linkset," where a web origin can publish a single list of the knowledge resources it offers, including graph data formats, a JSON-LD context, a plain-text file for AI agents, an export of content chunks, and a change log. Each entry can include a SHA-256 checksum in HTTP digest syntax so clients confirm a download matches the description. It introduces no new media type or link relation, using the existing "describedby" relation, and registers only a profile URI.

Explains how to use the Semantic Definition Format (SDF) to build digital twins, which are digital counterparts of physical devices and systems. SDF captures the data and interactions these twins rely on and can describe how each physical object within a larger twin behaves, what it can do, and how it communicates, giving a common modeling basis for IoT digital-twin deployments.

Segment Routing lets a node steer traffic along any path using an SR Policy, an ordered list of segments realized in SR-MPLS or SRv6 data planes. Maximum SID Depth (MSD) limits how many SIDs a node or link can support, first defined for SR-MPLS and later extended to SRv6. The draft notes that MSD can constrain SR-TE path computation by a PCE and specifies the MSD considerations a PCE should account for when computing SR policy paths.

Addresses a traffic-engineering need in networks using Layer 2 bundles such as LAGs, where a controller may need to learn how the member links of a bundle connect across the two ends. It specifies how OSPF and IS-IS can advertise the remote interface identifiers of those bundle members, and defines a matching extension to BGP-LS.

Proposes a framework for intelligent network measurement using the Model Context Protocol (MCP), where network devices act as MCP servers and controllers, management systems, or LLM-capable devices act as MCP clients, supporting natural-language-driven, AI-assisted measurement. It aims to provide real-time performance monitoring, intelligent fault diagnosis, topology discovery, and automated measurement workflows, and covers the architecture, use cases, and security considerations.

Proposes using the Model Context Protocol (MCP) for network troubleshooting, with devices acting as MCP servers that expose management data and actions and controllers acting as MCP clients. It extends the model to device-to-device collaboration so network elements can correlate faults among themselves when the controller is unreachable. The aim is an intent-based, conversational, secure approach to troubleshooting, configuration validation, and closed-loop remediation without new protocols or device agents.

NETCONF is strong for configuration transactions and YANG data models but struggles with AI-driven interpretation of requests, long-running work across domains, agreement among several agents, fast DevOps cycles, and transferring large non-configuration files. The document maps these gaps and proposes MCP and A2A as complementary protocols, with implementation guidance and ways for them to coexist with NETCONF.

LLMs are entering network operations through natural-language intent interfaces, but existing southbound protocols (NETCONF, RESTCONF, gNMI, MCP, A2A) were not designed for conversational, semantically rich, multi-agent orchestration. The document offers a systematic gap analysis of these protocols and identifies extension points for each to meet intent-based networking requirements.

LLMs and AI agents are changing how operators work with infrastructure, yet current management systems lack a standard, secure, intent-driven way for agents to find, call, and reason about network functions. The draft defines the problem of applying the Model Context Protocol to network management, covers use cases such as troubleshooting, measurement, security, and optimization, and lays out functional, security, and interoperability requirements for MCP-based architectures.

Reviews changes between Unicode 12.0.0 and Unicode 18.0.0 that affect the values the IDNA2008 algorithm produces. It identifies newly added code points needing individual examination and asks that they be marked "UNDER REVIEW" in the IANA tables until reviewed, as RFC 8753 provides. It supplies the table IANA needs to make its database consistent with Unicode 18.0.0, with all values computed directly from the Unicode Character Database using the RFC 5892 Section 3 algorithm.

Segment Routing lets operators define end-to-end paths by encoding an ordered list of instructions (segments), running on either an MPLS or IPv6 data plane. The Path Segment Identifier (PSID) already identifies SR paths in SR-MPLS networks, supporting uses such as end-to-end path protection and performance measurement. The document extends the PSID concept to identify SRv6 paths in IPv6 networks.

Explains how an Extended Community is derived from a Route Target, how the derived community corresponds to its Route Target, and how the two differ. It lists the code points already assigned for this purpose and requests the remaining assignments. It does not define how these derived communities are used, leaving that to the documents that specify each use.

Defines five standard risk categories for sensitive inputs in Model Context Protocol tool schemas, along with fixed rules and edge-case conditions for assigning them, so systems get risk information about a tool's inputs before it runs, independent of any gateway, routing, or capability system. The categories derive from a frozen census of publicly reachable MCP servers; the census and classification code are archived on Zenodo. It is an Informational individual submission others may cite as the source of these labels.

Describes how a QUIC endpoint can hand off sending Stateless Resets to a component that keeps running after the endpoint's QUIC stack is gone, such as an operating system resetting peers of an application that has exited. Because this is most useful for clients using zero-length Connection IDs, it also defines a transport parameter and a frame that let those clients share a Stateless Reset Token.

Defines a Packet Content Filter component for BGP FlowSpec (both FSv1 and FSv2). Existing IP filter components match only IP and transport-layer header fields, missing volumetric DDoS traffic that carries recognizable fixed payload patterns. The new component matches packet content at a specified offset using offset type, offset, content length, content, and mask fields (FSv1 also including an IP-version field), letting routers and switches handle simple volumetric attacks and traffic optimization while reducing processing cost.

Describes a gap in multi-topology routing across autonomous systems: under RFC 5120 (IS-IS) and RFC 4915 (OSPF), a topology in one AS may need to connect to a topology with a different MT-ID in a neighboring AS, but existing inter-AS Traffic Engineering extensions do not advertise this relationship. It proposes IS-IS and OSPF extensions that advertise the binding between local and remote MT-IDs on an inter-AS link.

Describes the VPN Prefix ORF mechanism, which lets a BGP router ask a neighbor to filter matching VPN routes. When route reflectors or ASBRs sit in the path, that neighbor may not be where the routes originated. The draft sets conditions under which an intermediate router may pass a matching filter request upstream, only after confirming no relevant downstream router still needs those routes, and defines an Interest Query/Response procedure for making that determination.

Analyzes the applicability of protocols related to a Computing-Aware Traffic Steering (CATS) system and describes how to build a CATS system by extending existing IETF protocols. It maps CATS functional needs onto available protocol mechanisms to guide implementers assembling computing-aware traffic steering from standardized building blocks.

Describes how measurement packets can follow predetermined paths in a Segment Routing domain, enabling new kinds of active measurement. It defines an experimental Type-p connectivity monitoring metric so central monitoring systems can supervise the state and performance of a connection or sub-path, supporting network-wide monitoring of SR path connectivity.

Specifies an experimental, JSON-based evidence envelope for exchanging source-grounded computational assertions, including provenance, reproducible operation identifiers, evidence sufficiency, and explicit abstention (declining to answer when support is lacking). The envelope separates syntactic validity, source integrity, mathematical correctness, and empirical truth. It does not define a general scientific inference engine and claims no experimental superiority.

By its name, this individual draft (part of the author's "verification/assurance" series) concerns FIC continuity, likely a continuity profile for fault/integrity-check evidence. The official abstract could not be retrieved on this run, so this description is approximate and will be grounded from the document's abstract on the next run.

By its name, this individual draft (part of the author's "verification/assurance" series) concerns CIF measurement, likely a measurement profile within a verification framework. The official abstract could not be retrieved on this run, so this description is approximate and will be grounded from the document's abstract on the next run.

Specifies FAMS-P, a peer-to-peer protocol letting federation gateways across administrative domains synchronize agent discovery metadata. It implements the "BGP-like structured peering" mode of the DAWN discovery architecture, targeting small closed federations of roughly 2 to 200 sites where DNS delegation does not fit. Gateways connect over mutually authenticated TLS 1.3, do an initial full sync then incremental updates, keep a local Federated Agent Directory, and apply signed records, TTL aging, replay protection, and per-peer export policy.

As cloud-native applications break into smaller services, those services can run on distributed edge nodes near users to improve experience. Under the circumstances analyzed, the draft proposes a Hybrid-Function-Chain (HFC) framework to allocate and schedule resources and services so that service delivery stays consistent from end to end across distributed edge deployments.

Defines five YANG 1.1 modules for configuring QUIC clients and servers. They cover basic settings for both roles, plus initial modules that mirror the IANA "QUIC Versions" and "QUIC Transport Parameters" registries, giving a standard configuration surface for QUIC-based transport in NETCONF-managed systems.

Describes four post-quantum password-authenticated key exchange (PAKE) protocols. OQUAKE and OQUAKE+ are standalone protocols resisting quantum-capable attackers: OQUAKE is symmetric for settings where both parties hold the password, while OQUAKE+ is the augmented client-server variant where only the client knows it. CPaceOQUAKE and CPaceOQUAKE+ combine classical CPace with OQUAKE in sequence so security holds if either component stays unbroken. The document recommends specific configurations for each.

Specifies Hamming Quasi-Cyclic (HQC) key encapsulation mechanisms for use with Hybrid Public Key Encryption (HPKE), covering the KEM interface, deterministic key-pair derivation, key encoding, and input validation for three HQC parameter sets. It also defines three hybrid KEMs pairing HQC with an elliptic-curve group (HQC with P-256, X25519, or P-384), requests six HPKE KEM identifiers, and leaves HPKE's key schedule, encryption operations, and export interface unchanged.

Describes how an Extensible Provisioning Protocol (EPP) connection is mapped onto HTTP. It specifies that EPP over HTTP (EoH) must use Transport Layer Security (TLS), i.e. HTTPS, to protect EPP data, giving registries and registrars a standardized HTTP transport binding for EPP as an alternative to the traditional EPP-over-TCP transport.

Introduces Flow Queue PIE (FQ-PIE), a hybrid packet scheduler and Active Queue Management algorithm aimed at bufferbloat. FQ-PIE uses hashing to classify incoming packets into different queues to provide flow isolation, then dequeues with a modified round-robin scheduler, with PIE managing each flow to keep link utilization high while holding queue delay near a target value.

Addresses how agent communication protocols blur different security claims (who authorized a task, which agent instance is acting, what a tool is, what authority was delegated, whether a session is continuous, and what evidence exists), which have different verifiers, freshness rules, and failure modes. Packing them into one token or record can make a protocol appear to grant more authority than a receiver can verify. It proposes a protocol-neutral verifier model and a reusable matrix, with outcome categories and dependency rules so a composite result cannot be stronger than its verified inputs.

Describes a flexible method for selecting candidate Segment Routing (SR) policy paths. Based on the real-time resource usage and forwarding quality of candidate paths, the head node can perform dynamic path switching among multiple candidate paths in the SR policy, improving responsiveness to changing network conditions.

Specifies a data model for synchronizing calendar data with a server using JMAP, letting clients efficiently read, write, and share calendars and events. It covers receiving push notifications for changes and event reminders, and tracking changes others make in a multi-user environment, providing a modern JMAP-based alternative to legacy calendar-access protocols.

A YANG configuration can express management intent, but its acceptance does not show that dependent entries in a programmable P4 forwarding pipeline took effect consistently. The draft proposes an experimental mapping and update profile for P4 covering a target-specific mapping contract, capability validation, generation-based activation, preservation of existing flow bindings, failure recovery, and operational reconciliation. It separates configuration acceptance, verified activation, and eventual convergence, defining no new encapsulation or YANG module.

Introduces JSON Schema as a JSON-based way to describe the structure of JSON data. Schemas can impose constraints on values, support extracting information from them, and define how software interacts with them. A second media type, for instances linked to a schema, offers richer integration than plain JSON documents can, moving JSON Schema toward standardization within the IETF.