Skip to main content

The Universal Zero-Port Interconnect Framework (UZPIF): An Identity-Centric Architecture for Post-Port Networking
draft-dpa-uzpif-framework-01

Document Type Active Internet-Draft (individual)
Author Benjamin Anthony Fisher
Last updated 2026-03-16
RFC stream (None)
Intended RFC status (None)
Formats
Stream Stream state (No stream defined)
Consensus boilerplate Unknown
RFC Editor Note (None)
IESG IESG state I-D Exists
Telechat date (None)
Responsible AD (None)
Send notices to (None)
draft-dpa-uzpif-framework-01
Network Working Group                                        B.A. Fisher
Internet-Draft                                                   DPA R&D
Intended status: Informational                             16 March 2026
Expires: 17 September 2026

  The Universal Zero-Port Interconnect Framework (UZPIF): An Identity-
             Centric Architecture for Post-Port Networking
                      draft-dpa-uzpif-framework-01

Abstract

   The Universal Zero-Port Interconnect Framework (UZPIF) describes a
   post-port networking model in which communication is established via
   outbound, identity-bound sessions to Rendezvous Nodes (RNs).  By
   removing publicly reachable listening ports at endpoints, UZPIF
   changes exposure assumptions and can reduce unsolicited ingress and
   Internet-wide scanning pressure, but it does not by itself guarantee
   privacy, decentralisation, or availability.

   This document outlines architectural motivation, a high-level
   security model, operational and economic considerations, a Pantheon
   trust and policy plane baseline, and an incremental migration
   approach.  It is part of an experimental, research-oriented
   Independent Stream suite and defines the current normative baseline
   for trust objects, validation rules, and security semantics within
   the suite framework.  Hard interoperability is expected for shared
   object semantics and validation rules.  Full wire-level, clustering,
   and proof-family interoperability is not claimed everywhere yet; the
   remaining details are intentionally profile-defined or deferred in
   companion drafts or profiles.  UZPIF is intended to be read alongside
   companion work describing the Universal Zero-Port Transport Protocol
   (UZP; [UZP]) and TLS-DPA ([TLS-DPA]).

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."

Fisher                  Expires 17 September 2026               [Page 1]
Internet-Draft                    UZPIF                       March 2026

   This Internet-Draft will expire on 17 September 2026.

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.
   Please review these documents carefully, as they describe your rights
   and restrictions with respect to this document.

Table of Contents

   1.  Scope and Status  . . . . . . . . . . . . . . . . . . . . . .   3
   2.  Executive Summary . . . . . . . . . . . . . . . . . . . . . .   4
   3.  Introduction  . . . . . . . . . . . . . . . . . . . . . . . .   5
     3.1.  Motivation and Deployment Context . . . . . . . . . . . .   6
     3.2.  Relationship to VPN-Based Approaches  . . . . . . . . . .   6
   4.  Terminology . . . . . . . . . . . . . . . . . . . . . . . . .   6
   5.  Common Signed Artefact Envelope . . . . . . . . . . . . . . .   7
     5.1.  Canonical Serialisation and Signature Coverage  . . . . .   9
     5.2.  Envelope Members and Extensions . . . . . . . . . . . . .  10
     5.3.  Suite-wide object_type Registry . . . . . . . . . . . . .  12
     5.4.  Signature Algorithm Identifiers . . . . . . . . . . . . .  14
     5.5.  Signature Entries, Ordering, and Detached Form  . . . . .  14
     5.6.  Validation Rules  . . . . . . . . . . . . . . . . . . . .  15
   6.  Core Architecture Overview (Informative)  . . . . . . . . . .  17
     6.1.  High-Level Flow: EP to RN to EP (Informative) . . . . . .  17
     6.2.  Pantheon Grant Verification (Informative) . . . . . . . .  17
     6.3.  Flow Stitching by the RN (Informative)  . . . . . . . . .  18
     6.4.  Multi-RN Stitching for High-Assurance Tenants
           (Informative) . . . . . . . . . . . . . . . . . . . . . .  18
   7.  Legacy Hardware Integration via HIL . . . . . . . . . . . . .  19
     7.1.  HIL Requirements  . . . . . . . . . . . . . . . . . . . .  19
     7.2.  Local Trust Boundary of HIL-Mediated Systems  . . . . . .  20
     7.3.  HIL Software and Supply-Chain Integrity . . . . . . . . .  21
     7.4.  Legacy Compatibility Classes  . . . . . . . . . . . . . .  21
   8.  Decentralisation and Authority Model  . . . . . . . . . . . .  22
     8.1.  Bootstrap Model Clarification . . . . . . . . . . . . . .  22
     8.2.  Bootstrap, Recovery, and Re-Enrolment . . . . . . . . . .  22
       8.2.1.  Bootstrap Trust-Transition Artefacts  . . . . . . . .  23
       8.2.2.  Baseline Trust-Transition Flow  . . . . . . . . . . .  26
       8.2.3.  Initial Trust Establishment . . . . . . . . . . . . .  27
     8.3.  Key Custody and Signing-Role Separation . . . . . . . . .  28
     8.4.  Client Sovereignty Model  . . . . . . . . . . . . . . . .  29
     8.5.  RN Controller Properties  . . . . . . . . . . . . . . . .  29

Fisher                  Expires 17 September 2026               [Page 2]
Internet-Draft                    UZPIF                       March 2026

       8.5.1.  RN Non-Sovereignty and Compromise Containment . . . .  30
       8.5.2.  Multi-RN Consistency and Failover Semantics . . . . .  31
     8.6.  RN Transparency Log Format  . . . . . . . . . . . . . . .  35
     8.7.  Proof of RN Misbehaviour  . . . . . . . . . . . . . . . .  36
   9.  Explicit Non-Goals  . . . . . . . . . . . . . . . . . . . . .  36
   10. Pantheon: Identity and Governance Model . . . . . . . . . . .  37
     10.1.  Identity Model . . . . . . . . . . . . . . . . . . . . .  37
     10.2.  Certificate Format . . . . . . . . . . . . . . . . . . .  38
     10.3.  Grant Structure  . . . . . . . . . . . . . . . . . . . .  38
     10.4.  Authority Identifiers and Metadata . . . . . . . . . . .  39
     10.5.  Issuer Discovery . . . . . . . . . . . . . . . . . . . .  40
     10.6.  Cross-Authority Validation . . . . . . . . . . . . . . .  40
     10.7.  Conflict and Split-Brain Handling  . . . . . . . . . . .  40
     10.8.  Attestation Model  . . . . . . . . . . . . . . . . . . .  41
     10.9.  Caching Rules  . . . . . . . . . . . . . . . . . . . . .  41
   11. Benefits and Trade-offs . . . . . . . . . . . . . . . . . . .  42
   12. Threat Model  . . . . . . . . . . . . . . . . . . . . . . . .  43
   13. Economics . . . . . . . . . . . . . . . . . . . . . . . . . .  44
   14. Migration Plan for Organisations  . . . . . . . . . . . . . .  44
   15. Security Considerations . . . . . . . . . . . . . . . . . . .  45
     15.1.  Transition of Network Security Roles . . . . . . . . . .  45
     15.2.  Compliance, Observability, and Intercept Boundaries  . .  46
     15.3.  Identity-Aware Observability Shift . . . . . . . . . . .  46
     15.4.  Time as a Trust Dependency . . . . . . . . . . . . . . .  47
     15.5.  Replay, Caching, and Stale Authority . . . . . . . . . .  48
     15.6.  Downgrade and Compatibility-Mode Abuse . . . . . . . . .  48
     15.7.  Break-Glass and Emergency Override Discipline  . . . . .  49
     15.8.  Metadata Leakage and Traffic Analysis  . . . . . . . . .  49
     15.9.  Rendezvous Concentration and Availability Pressure . . .  50
   16. IANA Considerations . . . . . . . . . . . . . . . . . . . . .  51
   17. Normative References  . . . . . . . . . . . . . . . . . . . .  52
   18. Informative References  . . . . . . . . . . . . . . . . . . .  52
   Author's Address  . . . . . . . . . . . . . . . . . . . . . . . .  53

1.  Scope and Status

   This Internet-Draft is part of an experimental, research-oriented
   suite prepared for the Independent Stream.  It is published to enable
   structured technical review, interoperability discussion, and
   disciplined specification development around the Universal Zero-Port
   Interconnect Framework (UZPIF).

   Within that suite, this document defines the current normative
   baseline for trust objects, validation rules, and security semantics
   governing the suite's shared identities, Grants, RN roles, and trust
   relationships.  Hard interoperability is expected for shared object
   semantics and validation rules.

Fisher                  Expires 17 September 2026               [Page 3]
Internet-Draft                    UZPIF                       March 2026

   The material is a research artefact.  It does not claim technical
   completeness, production readiness, or endorsement by the IETF or any
   other standards body, and it is not presented as a standards-track
   specification.

   Full wire-level, clustering, and proof-family interoperability is not
   claimed everywhere yet.  Companion wire encodings, clustering
   details, proof families, and deployment profiles remain intentionally
   profile-defined or deferred, so this draft should not be read as
   claiming a fully closed transport, proof, or availability model.

   It is designed for experimentation and profile-driven deployments
   within its target environment.  Privacy, decentralisation, and RN
   availability remain deployment- and profile-dependent properties.

   During conversion from internal research documents into IETF XML,
   care has been taken to:

   *  preserve a clear distinction between normative and informative
      content;

   *  use requirement language (e.g., "MUST", "SHOULD", "MAY") only
      where protocol behaviour is intentionally specified;

   *  avoid any implication of registry finalisation, mandatory
      implementation, or standard-track status; and

   *  maintain intellectual-property neutrality, with no implied patent
      grants or licensing commitments beyond the IETF Trust copyright
      licence applicable to Internet-Draft text.

   Ongoing research, implementation, performance validation, and real-
   world pilot work remain outside the scope of this Internet-Draft text
   and may be pursued separately.

2.  Executive Summary

   The Internet still commonly exposes services via publicly reachable
   transport ports, a legacy design choice that enables scanning and
   unsolicited connection attempts at global scale.  Operationally, this
   contributes to exposure for denial-of-service attacks, credential
   attacks, and lateral movement within networks.

   UZPIF (the framework) and UZP ([UZP]) (its transport protocol) remove
   the concept of exposed ports at endpoints.  Both endpoints initiate
   outbound, identity-anchored sessions to a Rendezvous Node (RN), which
   only stitches traffic when identity, context, and declared purpose
   align under policy issued by Pantheon, the identity and policy plane.

Fisher                  Expires 17 September 2026               [Page 4]
Internet-Draft                    UZPIF                       March 2026

   The intent is a model where discoverability and session establishment
   are policy-mediated rather than assumed by default, and where
   application traffic can be end-to-end authenticated and encrypted.
   UZP ([UZP]) is designed to support performance properties such as:

   *  block-level reliability,

   *  selective retransmission, and

   *  deterministic pacing.

   Legacy applications and non-UZPIF-capable hardware are intended to
   continue to operate via a Hardware Integration Layer (HIL) that acts
   as the explicitly modelled compatibility edge between port-based
   protocols and identity-centric sessions.

   A UZPIF-aligned design should be evaluated not only for its steady-
   state cryptographic model, but also for whether its bootstrap,
   recovery, downgrade, mediation, operational override, and
   observability paths reintroduce singular trust dependencies that the
   architecture is intended to avoid.

   UZPIF builds on transport, security, and identity work embodied in
   QUIC [RFC9000], TLS 1.3 [RFC8446], and the Host Identity Protocol
   [RFC7401], while aligning with modern zero-trust guidance (e.g., NIST
   SP 800-207 [NIST-SP800-207]) and post-quantum cryptography
   standardisation efforts (e.g., the NIST PQC project [NIST-PQC]).

3.  Introduction

   This document provides an architectural overview of UZPIF and the
   deployment conditions addressed by an identity-first, rendezvous-
   based model that avoids publicly reachable listeners.

   UZPIF should therefore be read as part of an experimental, research-
   oriented Independent Stream suite and as the current normative
   baseline for trust objects, validation rules, and security semantics
   within the framework.  Hard interoperability is expected for shared
   object semantics and validation rules.  Full wire-level, clustering,
   and proof-family interoperability is not claimed everywhere yet; the
   remaining details are intentionally profile-defined or deferred.
   Removing listeners alone does not by itself solve privacy,
   decentralisation, or RN availability.

Fisher                  Expires 17 September 2026               [Page 5]
Internet-Draft                    UZPIF                       March 2026

3.1.  Motivation and Deployment Context

   *  Investment in perimeter defences (e.g., DDoS mitigation and
      application firewalls) can yield diminishing returns as attackers
      automate scanning and exploit discovery at Internet scale.

   *  Zero Trust Network Access (ZTNA) and SASE deployments indicate
      demand for identity-first networking, yet many approaches still
      expose TCP/UDP ingress and rely on perimeter constructs.
      [NIST-SP800-207]

   *  Post-quantum cryptography efforts provide a path to identity-first
      transport without prohibitive performance regression as key
      encapsulation and signature schemes mature.  [NIST-PQC]

3.2.  Relationship to VPN-Based Approaches

   Conventional VPNs and overlay networks typically retain the
   assumption that services listen on IP:port tuples, even if those
   ports are only reachable within a private address space or through a
   gateway.  QUIC [RFC9000], TLS 1.3 [RFC8446], HIP [RFC7401], and
   systems such as Tor [Tor] demonstrate that identity, encryption, and
   rendezvous can be decoupled from raw addressing semantics, but they
   generally retain listener-based service reachability.

   *  *No listeners* at endpoints.

   *  *Identity-as-address* via identities (e.g., canonical and
      ephemeral identities) rather than IP:port.

   *  *Companion transport profiles* define session and transport
      behaviour rather than inheriting a VPN tunnel abstraction.

   *  *Pantheon policy plane* encoding purpose, context, and validity
      into every session.

4.  Terminology

   *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 [RFC2119] and
   [RFC8174] when, and only when, they appear in all capitals.

   This Internet-Draft is primarily architectural; requirement language
   is used sparingly and only where behaviour is intentionally
   specified.

Fisher                  Expires 17 September 2026               [Page 6]
Internet-Draft                    UZPIF                       March 2026

   EP  Endpoint.  A host or service that participates in UZPIF by
      initiating outbound sessions.

   RN  Rendezvous Node.  A mediator that accepts outbound sessions and
      stitches permitted flows.

   RN Controller  Control-plane software used to operate an RN
      deployment, including policy decisions and transparency log
      publication.

   Pantheon  An identity, attestation, and policy plane whose
      authorities bind identity, policy, and trust metadata to keys or
      selectors accepted under local policy, may validate or certify
      those bindings, and may issue credentials, Grants, and delegations
      over them; when federated, they rely on authority metadata, issuer
      discovery, cross-authority validation, and conflict handling.

   HIL  Hardware Integration Layer.  A constrained edge mediation
      component that exposes or terminates legacy port-based protocols
      on behalf of non-UZPIF-capable hardware, while participating in
      UZPIF as a policy-bound identity-aware gateway.

   CID  Canonical Identity.  A long-term cryptographic identity used to
      identify an EP (or a delegated sub-identity).

   EID  Ephemeral Identity.  A short-lived identity used for sessions,
      derived or issued under policy.

   ZPIT  Zero-Port Interconnect Tunnel.  An end-to-end encrypted tunnel
      stitched via one or more RNs.

   Proof of RN Misbehaviour  A cryptographically verifiable evidence
      package that demonstrates RN protocol deviation without requiring
      a central adjudicator.

5.  Common Signed Artefact Envelope

   This section defines the normative signed-object backbone for the
   UZPIF suite, including Pantheon Grants, Pantheon Certificates
   (PCerts), bootstrap and recovery artefacts, revocation signals and
   threshold evidence as profiled by TLS-DPA ([TLS-DPA]), Proof-of-
   Reachability evidence as defined by UZP ([UZP]), RN set
   advertisements, RN transparency entries, Discovery Grants, Signed
   Checkpoints, Indexer Receipts, and Revocation Acknowledgement
   artefacts as profiled by outbound indexing ([OUTBOUND-INDEXING]), and
   related auditable objects.  For interoperable exchange across the
   suite, such objects MUST use this common signed artefact envelope.

Fisher                  Expires 17 September 2026               [Page 7]
Internet-Draft                    UZPIF                       March 2026

   The envelope is wire-neutral: it does not define transport framing or
   carriage format.  It does define a single logical object model,
   canonical serialisation, signature coverage rule, extension rule, and
   validation baseline so that conforming implementations do not sign
   the same logical object differently.

   Each signed artefact consists of three top-level members: "envelope",
   "body", and "signatures".  The top-level member set is closed;
   unknown top-level members MUST cause rejection.  The "envelope"
   member carries the suite-wide semantics defined here.  The "body"
   member carries the object-specific semantics defined by the relevant
   draft.  The "signatures" member carries one or more embedded
   signatures over the canonical signature input defined below.  Object-
   specific drafts MAY extend the body, but they MUST NOT redefine the
   envelope semantics, canonical serialisation, exact signature
   coverage, extension handling, object identifier derivation, algorithm
   identifier handling, epoch-versus-sequence precedence, detached-
   signature policy, or ordering rules defined in this section.

   This section is normatively central for suite-interoperable signed
   objects.  Companion drafts define only the body profile, object-
   specific eligibility checks, and any explicitly deferred proof-family
   behaviour for a registered object type.  They MUST NOT redefine
   canonical serialisation, exact signature coverage, object identifier
   derivation, unknown-field handling, signature ordering, algorithm
   identifier comparison, epoch-versus-sequence precedence, or detached-
   signature policy.  If a companion draft conflicts with this section,
   this section controls for suite-interoperable exchange.

Fisher                  Expires 17 September 2026               [Page 8]
Internet-Draft                    UZPIF                       March 2026

   {
     "envelope": {
       "version": 1,
       "object_type": "grant",
       "object_id": "urn:uzpif:obj:v1:sha256:6f9c8a4c...",
       "issuer_authority_id": "authority:example",
       "subject_id": "cid:subject",
       "audience_id": "cid:audience",
       "scope": "tenant=alpha;service=payments;action=connect",
       "policy_id": "pantheon-connect",
       "policy_version": "2026-03-13",
       "issued_at": "2026-03-13T12:00:00Z",
       "not_before": "2026-03-13T12:00:00Z",
       "not_after": "2026-03-13T12:05:00Z",
       "epoch": 44,
       "sequence": 181,
       "nonce": "4f1c8a9d...",
       "key_id": "sig-main-2026q1",
       "prev_hash": "sha256:...",
       "log_ref": "log://rn.example/181"
     },
     "body": {
       "requested_peer": "cid:peer",
       "action": "connect",
       "qos_class": "interactive"
     },
     "signatures": [
       {
         "key_id": "sig-main-2026q1",
         "alg": "ed25519",
         "value": "base64..."
       }
     ]
   }

   Figure 1: Informative example of the common signed artefact envelope.

   This figure is illustrative only.  The normative envelope semantics,
   canonical serialisation rules, and validation requirements are
   defined by the text in this section.

5.1.  Canonical Serialisation and Signature Coverage

   Producers and verifiers MUST use the JSON Canonicalization Scheme
   (JCS) defined in [RFC8785] over UTF-8 JSON text.

Fisher                  Expires 17 September 2026               [Page 9]
Internet-Draft                    UZPIF                       March 2026

   Object production and verification use two exact canonical preimages.
   First, the producer derives "object_id" from the JCS serialisation of
   {"body": body, "envelope": envelope_without_object_id}, where
   envelope_without_object_id is the final envelope with only the
   "object_id" member omitted.  Second, after inserting the derived
   "object_id", the canonical signature input is exactly the UTF-8 octet
   string of the JCS serialisation of {"body": body, "envelope":
   envelope}. The top-level "signatures" member, any detached-signature
   container, and any transport wrapper are outside both canonical
   preimages.

   The "object_id" value MUST be formatted as
   "urn:uzpif:obj:v1:sha256:<lowercase-hex>", where the digest is the
   SHA-256 hash of the object-id preimage described above.  Verifiers
   MUST recompute and compare "object_id" before evaluating signatures.
   Adding, removing, or reordering signatures does not change
   "object_id".  Any change to envelope or body content, including
   extensions, validity bounds, policy references, or sequence metadata,
   MUST produce a different "object_id".

   Each accepted signature MUST cover the exact canonical signature
   input bytes and nothing else.  Suite-interoperable signatures MUST
   NOT be computed over pretty-printed JSON, partially selected members,
   alternate canonicalisation schemes, XML renderings, CBOR
   transcodings, log wrappers, transport records, or presentation-layer
   containers.

   If an object-specific schema admits optional defaults, producers MUST
   materialise the exact values to be signed before canonicalisation.
   Verifiers MUST recompute "object_id" and the signature input from the
   received values, not from locally inferred defaults or omitted
   aliases.

   Non-finite numbers, duplicate member names, semantically equivalent
   but non-canonical encodings, or values that cannot be represented
   under [RFC8785] MUST cause rejection.

5.2.  Envelope Members and Extensions

   Every common signed artefact MUST include "version", "object_type",
   "object_id", "issuer_authority_id", "subject_id", "scope",
   "issued_at", "not_before", "not_after", "key_id", and at least one
   entry in "signatures".  The "body" member MUST be a JSON object whose
   direct member set is closed by the declared "object_type": direct
   body members not defined for that object type, "extensions", or
   "critical_extensions" MUST cause rejection. "audience_id",
   "policy_id", "policy_version", "epoch", "sequence", "nonce",
   "prev_hash", and "log_ref" MUST be present when relevant to the

Fisher                  Expires 17 September 2026              [Page 10]
Internet-Draft                    UZPIF                       March 2026

   object's semantics, replay model, or transparency linkage.

   Authenticity alone is insufficient for reliance on a common signed
   artefact.  A valid signature proves origin and integrity, but a
   relying party MUST also evaluate freshness, scope, supersession,
   audience binding where relevant, and policy eligibility before
   treating the artefact as current authority.

   version  Identifies the envelope version so that parsers and relying
      parties can apply the correct validation rules.  Suite-
      interoperable objects defined by this revision MUST set "version"
      to 1.

   object_type  Classifies the artefact using the exact case-sensitive
      token registered in Section 5.3.  For suite-interoperable
      exchange, aliases or profile-specific synonyms are invalid.

   object_id  Provides the stable identifier for the logical signed
      object.  Producers MUST derive it according to Section 5.1.  The
      same envelope and body content MUST always produce the same
      object_id, regardless of signature count.  Producers MUST NOT
      reuse an object_id for different envelope or body content.

   issuer_authority_id  Identifies the authority context under which the
      object is issued.  It MUST map to signed authority metadata or
      other locally trusted issuer metadata.

   subject_id  Identifies the principal, resource set, or relationship
      primarily affected by the artefact.

   audience_id  Identifies the intended relying party or recipient when
      the artefact is not intended for universal consumption.

   scope  Defines the operational scope of the artefact, such as
      authorised actions, resource sets, tenant boundaries, RN context,
      or revocation domain.

   policy_id / policy_version  Name the policy basis and version under
      which the artefact was issued when policy tracking matters.

   issued_at / not_before / not_after  Define issuance time and validity
      bounds.  Artefacts outside their validity window MUST be rejected.

   epoch / sequence  Provide ordering and conflict-detection context.

Fisher                  Expires 17 September 2026              [Page 11]
Internet-Draft                    UZPIF                       March 2026

      If both are present, "epoch" takes precedence for semantic
      supersession and "sequence" orders objects within the same epoch
      or append-only stream.  A higher "epoch" supersedes any lower
      "epoch" regardless of "sequence".  If "epoch" is equal, a higher
      "sequence" is newer.  A lower-epoch object MUST NOT supersede a
      higher-epoch object merely because its "sequence" is larger.

   nonce  Carries replay-unique entropy for replay-sensitive artefacts
      such as Grants, PoR evidence, or session-bound authorisations.
      Re-usable objects that are not replay-sensitive MAY omit it.

   key_id  Identifies the primary issuer key used for key discovery.  It
      MUST match the first embedded signature entry after signature-
      array sorting and MUST match at least one embedded signature
      entry.

   prev_hash / log_ref  Link the artefact into an append-only sequence
      or an external transparency record when auditability or append-
      only verification is required.

   body members  The direct members of "body" are defined by the
      registered object-type profile.  Unknown direct body members MUST
      cause rejection unless they are carried under "body.extensions".

   body.extensions / body.critical_extensions  Object-specific extension
      values MUST appear only under "body.extensions", which, if
      present, MUST be a JSON object. "body.critical_extensions", if
      present, MUST be an array of unique extension keys sorted in
      ascending lexicographic order, and each listed key MUST exist in
      "body.extensions".  Extension keys MUST be unique namespaced ASCII
      strings.  Unknown extension values MAY be ignored only if they are
      not listed in "body.critical_extensions".  Unknown critical
      extensions MUST cause rejection.  Unknown members in "envelope",
      in an embedded signature entry, or as direct body members MUST
      cause rejection.

5.3.  Suite-wide object_type Registry

   The "object_type" member is a case-sensitive lowercase ASCII token.
   For suite-interoperable exchange, it MUST use one of the exact values
   registered in Table 1.  Companion drafts define only the body profile
   and additional validation for their registered object types; they
   MUST NOT rename, alias, or reinterpret a registered value.  New
   suite-wide object types require an explicit update to this registry.

   Detached signatures are not permitted for any registered object type
   in this revision.  A future suite revision MAY change that only by
   updating both this registry and Section 5.5.

Fisher                  Expires 17 September 2026              [Page 12]
Internet-Draft                    UZPIF                       March 2026

   +============================+===========================+==========+
   |object_type                 |Body profile               |Detached  |
   |                            |                           |signatures|
   +============================+===========================+==========+
   |grant                       |Pantheon Grant body        |No        |
   |                            |profile (Section 10.3)     |          |
   +----------------------------+---------------------------+----------+
   |pcert                       |Pantheon Certificate body  |No        |
   |                            |profile (Section 10.2)     |          |
   +----------------------------+---------------------------+----------+
   |bootstrap-seed-bundle       |Bootstrap trust seed       |No        |
   |                            |bundle (Section 8.2.1.1)   |          |
   +----------------------------+---------------------------+----------+
   |bootstrap-admission         |Bootstrap admission body   |No        |
   |                            |profile (Section 8.2.1.2)  |          |
   +----------------------------+---------------------------+----------+
   |recovery-reenrolment-bundle |Recovery / Re-Enrolment    |No        |
   |                            |body profile               |          |
   |                            |(Section 8.2.1.3)          |          |
   +----------------------------+---------------------------+----------+
   |rn-set-statement            |RN set advertisement body  |No        |
   |                            |profile (Section 8.5.2.1)  |          |
   +----------------------------+---------------------------+----------+
   |rn-transparency-entry       |RN transparency log entry  |No        |
   |                            |body profile (Section 8.6) |          |
   +----------------------------+---------------------------+----------+
   |misbehaviour-proof          |RN misbehaviour evidence   |No        |
   |                            |body profile (Section 8.7) |          |
   +----------------------------+---------------------------+----------+
   |revocation                  |TLS-DPA Revocation Signal  |No        |
   |                            |body profile ([TLS-DPA])   |          |
   +----------------------------+---------------------------+----------+
   |threshold-consensus-evidence|TLS-DPA threshold-         |No        |
   |                            |consensus body profile     |          |
   |                            |([TLS-DPA])                |          |
   +----------------------------+---------------------------+----------+
   |por-evidence                |UZP Proof-of-Reachability  |No        |
   |                            |body profile ([UZP])       |          |
   +----------------------------+---------------------------+----------+
   |discovery-grant             |Outbound indexing          |No        |
   |                            |Discovery Grant body       |          |
   |                            |profile                    |          |
   |                            |([OUTBOUND-INDEXING])      |          |
   +----------------------------+---------------------------+----------+
   |index-transparency-entry    |Outbound indexing          |No        |
   |                            |transparency-entry body    |          |
   |                            |profile                    |          |
   |                            |([OUTBOUND-INDEXING])      |          |

Fisher                  Expires 17 September 2026              [Page 13]
Internet-Draft                    UZPIF                       March 2026

   +----------------------------+---------------------------+----------+
   |signed-checkpoint           |Outbound indexing          |No        |
   |                            |checkpoint body profile    |          |
   |                            |([OUTBOUND-INDEXING])      |          |
   +----------------------------+---------------------------+----------+
   |indexer-receipt             |Outbound indexing receipt  |No        |
   |                            |body profile               |          |
   |                            |([OUTBOUND-INDEXING])      |          |
   +----------------------------+---------------------------+----------+
   |revocation-acknowledgement  |Outbound indexing          |No        |
   |                            |revocation-acknowledgement |          |
   |                            |body profile               |          |
   |                            |([OUTBOUND-INDEXING])      |          |
   +----------------------------+---------------------------+----------+

                Table 1: Registered suite-wide object types

5.4.  Signature Algorithm Identifiers

   The "alg" member is a case-sensitive ASCII token that identifies the
   exact signature verification procedure applied to the signature bytes
   in "value".  Comparison is byte-for-byte.  Aliases, case folding,
   numeric registry codes, OIDs, or implicit translation between naming
   schemes MUST NOT be used for suite-interoperable verification.

   Trusted issuer metadata MUST bind each signing key identifier to one
   or more exact "alg" tokens.  A verifier MUST accept a signature only
   when the signature entry's "alg" exactly matches a token bound to
   that key and local policy permits that algorithm for the declared
   object_type and time interval.

   Composite or hybrid constructions, if used, MUST have their own exact
   "alg" token and verification procedure.  They MUST NOT be represented
   by multiple "alg" values within one signature entry or by guessing
   from key type alone.

5.5.  Signature Entries, Ordering, and Detached Form

   The "signatures" array MUST contain at least one embedded signature.
   Each embedded signature entry MUST contain exactly "key_id", "alg",
   and "value".  The embedded-signature member set is closed.

Fisher                  Expires 17 September 2026              [Page 14]
Internet-Draft                    UZPIF                       March 2026

   The array order is significant for serialised interchange and MUST be
   sorted in ascending lexicographic order by "key_id", then "alg", then
   "value".  Duplicate entries with the same "key_id" and "alg", or
   duplicate complete entries, MUST cause rejection.  A received array
   that is not already in this order MUST cause rejection.  The envelope
   "key_id" MUST identify the first embedded signature entry after
   sorting.

   The "alg" value is mandatory and case-sensitive, and it is processed
   according to Section 5.4.  A signature entry whose "alg" is absent,
   unknown, unsupported, mismatched to the key metadata, or forbidden by
   local policy MUST be treated as invalid.  If the remaining valid
   signatures do not satisfy the required signature set, the object MUST
   be rejected.

   Embedded signatures are the only interoperable signature form for the
   registered object types in Table 1.  For those object types, detached
   signatures MUST NOT be used and MUST NOT count toward policy
   satisfaction.  If a future suite revision explicitly permits detached
   signatures for a specific object type, each detached signature MUST
   cover the same canonical signature input as the embedded form and
   MUST carry "object_id", "key_id", "alg", and "signed_object_hash",
   where "signed_object_hash" is "sha256:<lowercase-hex>" over the
   canonical signature input.  If both embedded and detached signatures
   are present for such a future object type, every signature MUST
   verify against the same canonical signature input and object_id; any
   mismatch MUST cause rejection.

5.6.  Validation Rules

   Relying parties validating a common signed artefact MUST verify at
   least the following:

   *  the top-level member set, envelope member set, embedded-signature
      member set, and object-type-specific direct body member set are
      well-formed and contain no unknown members except under
      "body.extensions";

   *  the envelope version is supported for the declared object_type,
      and registered suite-wide object types defined by this revision
      use "version" equal to 1;

   *  the declared object_type is an exact registered suite-wide token
      when interoperable suite exchange is claimed;

   *  object_id recomputation succeeds;

Fisher                  Expires 17 September 2026              [Page 15]
Internet-Draft                    UZPIF                       March 2026

   *  issuer_authority_id is locally trusted, validly delegated, or
      otherwise accepted under the applicable federation policy;

   *  the signing key identified by key_id is valid for the object_type
      and time interval, and the envelope key_id matches the first
      embedded signature entry after sorting;

   *  signature ordering and duplicate checks succeed, and every
      accepted signature verifies over the exact canonical signature
      input;

   *  each "alg" value exactly matches trusted key metadata and a
      locally permitted signature verification procedure for the
      declared object_type;

   *  the required signature set satisfies local policy, including any
      threshold or quorum rule;

   *  the current time falls within the declared validity interval;

   *  audience_id matches the relying party when the object is audience-
      bound;

   *  nonce uniqueness and epoch or sequence freshness checks succeed
      when replay resistance or state ordering matter;

   *  the artefact has not been superseded, rendered stale under local
      policy, or made policy-ineligible for the relevant scope and
      decision;

   *  prev_hash or log_ref, when present, are consistent with the
      relevant append-only or transparency evidence; and

   *  body.critical_extensions, when present, is unique, sorted,
      references only keys present in body.extensions, and contains no
      unsupported critical extension names.

   Objects sharing the same issuer_authority_id, object_type,
   subject_id, scope, and applicable epoch or sequence state, but with
   incompatible canonical signature inputs, MUST be treated as
   conflicting artefacts and processed according to Section 10.7.

Fisher                  Expires 17 September 2026              [Page 16]
Internet-Draft                    UZPIF                       March 2026

6.  Core Architecture Overview (Informative)

   This section is informative.  It provides high-level role and flow
   sketches only.  Authoritative session, handshake, stitching,
   failover, and proof semantics are defined in the companion UZP
   ([UZP]) and TLS-DPA ([TLS-DPA]) drafts.  Authoritative shared object
   semantics are defined by the common signed artefact envelope and the
   object-specific sections of this document and the relevant companion
   drafts.

   If any figure or shorthand message label in this section conflicts
   with the companion transport or handshake drafts, the companion
   drafts govern.

6.1.  High-Level Flow: EP to RN to EP (Informative)

   In UZPIF, both peers initiate outbound sessions towards an RN.  After
   policy evaluation and authorisation, the RN stitches the two sessions
   into a tunnel (ZPIT) that carries end-to-end protected application
   data.

   EP A (Initiator)        RN        EP B (Responder)
       |---- outbound ----->|              |
       |                    |<---- outbound|
       |<==== end-to-end encrypted ZPIT (stitched via RN) ====>|

          Figure 2: High-level communication pattern (informative)

   This figure shows both endpoints initiating outbound sessions to the
   RN, which stitches them into a single ZPIT.

   The labels in this figure are illustrative only; this document does
   not define authoritative wire messages or handshake ordering for the
   transport steps shown.

6.2.  Pantheon Grant Verification (Informative)

   Prior to stitching, an EP is expected to obtain a signed
   authorisation ("Grant") from Pantheon, as defined in Section 10.3.
   Grants bind identity to purpose and validity constraints, enabling
   RNs to make consistent policy decisions.  The authoritative Grant
   format and interoperability requirements are defined in Section 10.3
   and Section 5.

    EP             Pantheon             RN
     |-- Grant Request -->|              |
     |<- Signed Grant ----|              |
     |----------- Grant + CID/EID ----------------->|

Fisher                  Expires 17 September 2026              [Page 17]
Internet-Draft                    UZPIF                       March 2026

          Figure 3: Grant request and issuance flow (informative)

   This figure illustrates the basic grant request and issuance exchange
   between an EP and Pantheon.

   The message labels are illustrative only.  Authoritative Grant object
   semantics are defined in Section 10.3, and authoritative transport or
   handshake sequencing is defined in companion drafts.

6.3.  Flow Stitching by the RN (Informative)

   The RN establishes a stitched tunnel only if both peers present
   acceptable identities and authorisations.  The exact transport,
   stitching, failover, and protected-traffic semantics are defined in
   the companion UZP and TLS-DPA drafts.

   EP A                 RN                  EP B
    |-- Join Request -->|                    |
    |<-- Stitch OK -----|                    |
    |                    |<-- Join Request --|
    |                    |-- Stitch OK ----->|
    |
    |<====== end-to-end encrypted ZPIT (SessionID-bound) ==========>|

           Figure 4: Join and stitch establishment (informative)

   This figure shows the RN joining two authorised sessions into a
   stitched tunnel without learning plaintext.

   The message names and sequencing in this figure are illustrative
   only.  Authoritative stitching, failover, and session semantics are
   defined in the companion UZP and TLS-DPA drafts.

6.4.  Multi-RN Stitching for High-Assurance Tenants (Informative)

   UZPIF can be extended to multi-hop stitching, for example where a
   tenant requires multiple independently operated RNs and attestation
   chains.  End-to-end protection is expected to remain between
   endpoints.

   EP A -> RN1 -> RN2 -> EP B

   EP A <======== end-to-end AEAD protected traffic ========> EP B

        Figure 5: Multi-hop stitching with end-to-end authenticated
                          encryption (informative)

Fisher                  Expires 17 September 2026              [Page 18]
Internet-Draft                    UZPIF                       March 2026

   This figure depicts a multi-hop RN chain while end-to-end AEAD
   protection remains between endpoints.

   This subsection is a deployment sketch only and does not define
   authoritative multi-RN transport semantics.

7.  Legacy Hardware Integration via HIL

   Where attached equipment cannot natively participate in UZPIF, a
   Hardware Integration Layer (HIL) MAY mediate between legacy port-
   based protocols and UZPIF sessions.  Real-world examples include
   industrial devices, medical and laboratory equipment, older
   appliances, embedded controllers, legacy PBXs, telephony systems, and
   other equipment whose firmware cannot be remodelled around identity-
   first session initiation or removal of hard-coded inbound listeners.

   A HIL is a compatibility and containment boundary, not a redefinition
   of legacy protocols as identity-native.  A HIL is not an exception to
   UZPIF.  It is the explicitly modelled compatibility edge through
   which such equipment interacts with an identity-first deployment.  A
   HIL does not make a legacy protocol identity-safe; it constrains and
   evidences the points at which a legacy protocol is permitted to
   interact with an identity-first system.

   HILs MUST apply explicit policy, minimise exposed legacy surface, and
   produce auditable evidence for mediated actions.  A deployment using
   a HIL MUST treat the attached device as outside the native UZPIF
   trust model unless that device is separately enrolled with its own
   identity and policy context.  The HIL therefore terminates the trust
   boundary cleanly rather than treating legacy traffic as natively
   identity-bound.

7.1.  HIL Requirements

   A HIL used to bridge non-UZPIF-capable equipment:

   *  MUST be explicitly identified as a translation boundary;

   *  MUST terminate trust domains cleanly rather than representing
      legacy traffic as natively identity-bound;

   *  MUST enforce a narrow allow-listed protocol surface for the
      specific attached device or device class;

   *  MUST expose only the minimum necessary legacy port behaviour
      toward the attached equipment;

Fisher                  Expires 17 September 2026              [Page 19]
Internet-Draft                    UZPIF                       March 2026

   *  MUST NOT elevate unauthenticated legacy assertions into Pantheon,
      Grant, or attestation truth without validation;

   *  SHOULD support one-way or brokered operation where possible rather
      than transparent raw pass-through; and

   *  MUST be auditable as a security-critical adapter.

   When a HIL mediates an action into UZPIF, it SHOULD bind that action
   to auditable evidence including:

   *  the HIL identity;

   *  the attached device identity or slot or port identity;

   *  the translated protocol;

   *  the relevant time interval;

   *  the applicable policy or Grant; and

   *  an evidence-log reference or equivalent audit record.

7.2.  Local Trust Boundary of HIL-Mediated Systems

   A HIL secures the boundary between a UZPIF environment and a legacy
   device interface; it does not guarantee the integrity or
   confidentiality of the local physical segment behind that boundary.
   Deployments MUST treat the HIL-attached legacy domain as an
   independently exposed trust zone unless additional protections are
   present.

   The HIL's protection applies to the mediated protocol boundary, not
   automatically to the local wire, serial line, bus, switch port,
   backplane, maintenance interface, or other attached legacy segment
   behind it.  Compromise of that local segment may still permit false
   device input, spoofed local traffic, local firmware replacement, bus
   replay, serial-line manipulation, switch-port insertion, maintenance-
   port misuse, false sensor injection, or manipulation of translated
   events before they cross into the UZPIF-mediated side.

   HIL containment therefore reduces and evidences legacy exposure, but
   it does not retroactively make the attached hardware domain
   cryptographically trustworthy.  Where stronger assurance is required,
   deployments SHOULD consider physical controls, tamper detection,
   local attestation, secure serial or bus wrappers where available, and
   device-origin integrity checks where feasible.

Fisher                  Expires 17 September 2026              [Page 20]
Internet-Draft                    UZPIF                       March 2026

7.3.  HIL Software and Supply-Chain Integrity

   A compromised HIL is especially dangerous because it sits at the
   boundary between legacy protocol behaviour and native identity-bound
   operation.  A malicious or subverted HIL may spoof device status,
   fabricate translated events, widen the permitted legacy surface,
   create covert ingress, or launder insecure traffic into a deployment
   that appears to be operating under stronger trust assumptions.

   HIL software manifests, configurations, policy bundles, and updates
   SHOULD be signed and auditable.  No HIL software, configuration, or
   policy change SHALL become authoritative unless its manifest,
   version, signer set, scope, and policy eligibility validate
   successfully under the deployment's normal trust model.

   HIL update, provisioning, and recovery channels MUST NOT become
   hidden sovereign authority paths.  Local administrative overrides,
   emergency maintenance paths, or vendor tooling MUST NOT silently
   bypass the same validation model claimed for ordinary operation.
   This is aligned with the ICSD ([ICSD]) principle that privileged
   update or override paths must be modelled so that invalid states are
   rejected rather than normalised.

7.4.  Legacy Compatibility Classes

   This document distinguishes three compatibility classes for legacy
   integration so that deployments can state clearly whether a device is
   fully mediated, tightly brokered, or merely forwarded.

   Class A - Full Mediation  The HIL terminates the legacy protocol and
      exposes a modelled, policy-bound interface into UZPIF.  This is
      the RECOMMENDED integration model.

   Class B - Brokered Pass-Through  The HIL permits tightly scoped
      transport bridging for a specific session under explicit Grant,
      time, scope, and evidence rules.  This MAY be used where full
      mediation is not feasible, but it provides weaker containment than
      Class A.

   Class C - Direct Forwarding Compatibility Mode  The HIL forwards
      arbitrary port traffic.  This mode is outside the recommended
      security model, SHOULD NOT be used except as a bounded migration
      exception, and MUST NOT be represented as equivalent to native
      UZPIF participation or to Class A or Class B mediation.

Fisher                  Expires 17 September 2026              [Page 21]
Internet-Draft                    UZPIF                       March 2026

8.  Decentralisation and Authority Model

   This section defines constitutional constraints for UZPIF deployments
   and federation behaviour.  These constraints are normative.

   RN controllers are permissionless to deploy.  No central authorising
   body is required for RN controller deployment.

   UZPIF federation MUST be bilateral or multilateral among consenting
   parties and MUST NOT require a mandatory hierarchical control model.

   UZPIF MUST NOT depend on any single registry, single directory, or
   single governance entity for protocol operation.  UZPIF MUST NOT
   require a global mandatory trust root.

8.1.  Bootstrap Model Clarification

   Bootstrap MAY use multiple Bootstrap Seed Bundles, mirrored
   distribution paths, offline media, local ceremony, and DNS or similar
   mechanisms as transports for seed distribution.  The transport used
   to deliver seed material does not itself create trust; trust begins
   only after the device validates a Bootstrap Seed Bundle under local
   bootstrap policy.

   Bootstrap mechanisms MUST be replaceable across deployments and over
   time.  Bootstrap mechanisms MUST NOT depend on a single distribution
   source.  No single seed source SHALL be mandatory.

8.2.  Bootstrap, Recovery, and Re-Enrolment

   Bootstrap, recovery, factory reset, re-enrolment, and emergency
   rekeying are privileged trust transitions.  A deployment that appears
   decentralised during steady-state operation but depends on a singular
   bootstrap or recovery path has introduced a hidden trust anchor and
   does not satisfy the intended UZPIF sovereignty model.

   New-node bootstrap, client recovery after state loss, RN rejoin after
   corruption, initial HIL enrolment, and emergency rekeying after
   compromise MUST NOT require a singular globally mandatory authority
   or delivery path.  These workflows MUST NOT depend solely on one
   website, one operator, one Recovery / Re-Enrolment Bundle, one vendor
   endpoint, or one golden administrative key as the only path to valid
   trust state.

Fisher                  Expires 17 September 2026              [Page 22]
Internet-Draft                    UZPIF                       March 2026

   Recovery workflows MUST fail closed on missing, unsigned, expired,
   conflicting, or policy-ineligible recovery material.  Recovered trust
   state MUST remain independently verifiable through the signed
   artefacts, authority metadata, and transparency or quorum evidence
   defined below before it becomes authoritative.

   This document defines a minimal baseline model for those privileged
   transitions.  Baseline-conformant bootstrap or recovery support
   consists of validated Bootstrap Seed Bundle input, explicit first-
   contact admission through a Bootstrap Admission Object, explicit
   continuity-sensitive return through a Recovery / Re-Enrolment Bundle
   when prior state existed, and issuance of only bounded initial or
   recovered identity and Grant state under policy.

   Factory reset and re-enrolment MUST be treated as privileged trust
   transitions rather than convenience workflows.  A reset or re-
   enrolled endpoint, RN, or HIL MUST re-establish trust as a new or
   explicitly recovered principal under local policy, and previous trust
   state MUST NOT be silently reinstated absent verifiable recovery
   artefacts.

   Profiles MAY use multiple Bootstrap Seed Bundles, mirrored
   distribution paths, offline recovery packages, out-of-band
   provisioning, or quorum-approved rekey procedures.  However, no
   single bootstrap or recovery source SHALL be mandatory for global
   protocol validity.  This aligns with the invariant-closed design
   discipline described in ICSD ([ICSD]), where privileged recovery
   transitions are modelled so that invalid or unauthorised recovery
   states are rejected rather than normalised.

8.2.1.  Bootstrap Trust-Transition Artefacts

   For interoperable exchange, bootstrap and recovery trust-transition
   artefacts MUST use the common signed artefact envelope defined in
   Section 5.  This document defines three baseline object types:
   "bootstrap-seed-bundle", "bootstrap-admission", and "recovery-
   reenrolment-bundle".

8.2.1.1.  Bootstrap Seed Bundle

   A Bootstrap Seed Bundle carries the initial trust material a device
   uses before it has normal steady-state authority.  It is the artefact
   that introduces the first accepted authority metadata and the first
   accepted RN set or RN discovery hints into the device's trust
   evaluation.

   A Bootstrap Seed Bundle MUST carry at least:

Fisher                  Expires 17 September 2026              [Page 23]
Internet-Draft                    UZPIF                       March 2026

   *  the initial trusted authority metadata, or stable digests and
      retrieval references for that metadata;

   *  the accepted RN set, RN discovery hints, or both;

   *  a validity interval suitable for bounded bootstrap use;

   *  a bootstrap policy identifier or equivalent environment or trust-
      domain identifier;

   *  an epoch or sequence value sufficient for supersession and
      rollback detection; and

   *  a signature set sufficient for the local bootstrap policy.

   A Bootstrap Seed Bundle authorises initial trust discovery only.  It
   does not by itself admit a device, grant long-term authority, or
   silently create steady-state trust.

   At baseline, before first authority contact, the device MUST verify
   the Bootstrap Seed Bundle's signature set, bootstrap policy
   identifier, validity interval, authority metadata, and epoch or
   sequence freshness.  Manufacturer certificates, vendor-hosted seed
   material, or installer media MAY contribute to that evaluation only
   if explicitly accepted by local bootstrap policy; they are not
   sufficient by themselves to admit the device or create steady-state
   trust.

8.2.1.2.  Bootstrap Admission Object

   A Bootstrap Admission Object is the bounded initial enrolment
   artefact used to admit a new device or principal into first authority
   contact.  It is a narrow bootstrap credential, not a substitute for
   steady-state identity, Grant, or administrative authority.

   A Bootstrap Admission Object MUST carry at least:

   *  a binding to the device or principal being admitted, such as an
      attestation digest, presented public key, hardware identifier, or
      ceremony token;

   *  the intended bootstrap audience, such as the authority context,
      enrolment service, or admitted RN class;

   *  the permitted first-contact actions, such as initial identity
      issuance, initial Grant acquisition, or bootstrap-session
      establishment;

Fisher                  Expires 17 September 2026              [Page 24]
Internet-Draft                    UZPIF                       March 2026

   *  a narrow scope, expiry, and nonce, counter, or equivalent one-
      time-use bound;

   *  the bootstrap policy identifier under which admission is
      evaluated; and

   *  the approval signatures or approval evidence required by the local
      bootstrap policy.

   A Bootstrap Admission Object MUST be treated as bounded and
   bootstrap-only.  It MUST NOT be equivalent to unlimited long-term
   authority, a standing administrator credential, or an implicit right
   to bypass normal Grant, revocation, or transparency processing.

   At baseline, first authority contact MUST fail closed unless the
   Bootstrap Admission Object matches the presented device or principal
   binding, intended bootstrap audience, permitted first-contact action
   set, and one-time-use or bounded-validity constraints.  Successful
   admission MAY yield only bounded initial identity state, bounded
   initial Grant state, or both, and MUST NOT create broader authority
   than the object and policy explicitly permit.

   When onboarding is mediated by a HIL, the admission decision MUST
   still be made under bootstrap policy and MUST bind the admitted
   device or principal and, where relevant, the mediating HIL context.
   HIL presence, reachability, or vendor tooling is not a substitute for
   Day Zero trust admission.

8.2.1.3.  Recovery / Re-Enrolment Bundle

   A Recovery / Re-Enrolment Bundle is the artefact set used after state
   loss, corruption, key rotation, compromise response, or factory
   reset.  It defines how a device or principal re-establishes trust
   after prior state existed and therefore carries stronger continuity
   and incident-handling semantics than initial bootstrap admission.

   A Recovery / Re-Enrolment Bundle MUST carry at least:

   *  the recovered subject or replacement subject binding;

   *  the recovery reason or incident class, such as state loss,
      corruption, compromise, or planned rotation;

   *  references to the last known trusted state, such as prior object
      identifiers, issuer state, or highest accepted epoch or sequence
      values when available;

Fisher                  Expires 17 September 2026              [Page 25]
Internet-Draft                    UZPIF                       March 2026

   *  the recovery actions permitted, including whether identity
      continuity, replacement identity issuance, rekey, or Grant re-
      establishment is allowed;

   *  a bounded validity interval and intended recovery audience;

   *  the required approval signatures, quorum evidence, or recovery
      evidence references; and

   *  refreshed authority metadata, RN hints, or both when recovery
      requires trust-anchor or routing refresh.

   Recovery requires a distinct evaluation path from bootstrap.  A
   Recovery / Re-Enrolment Bundle MUST be evaluated under explicit
   recovery policy and MUST require stronger approval, continuity
   evidence, or incident classification than ordinary bootstrap
   admission when prior state existed or compromise is suspected.

   At baseline, a Recovery / Re-Enrolment Bundle MUST be evaluated
   against explicit prior-state linkage, incident classification,
   intended recovery audience, and the required approval or quorum
   evidence before any recovered identity or Grant state is accepted.
   Replaying bootstrap seed or bootstrap-admission artefacts without
   those tighter recovery checks MUST NOT be treated as valid recovery.

8.2.2.  Baseline Trust-Transition Flow

   The baseline bootstrap and recovery ceremony is as follows:

   1.  The device begins outside the closed trust domain and MUST NOT
       assume Pantheon-recognised identity, Grant state, accepted
       steady-state RN authority, or inherited session trust.

   2.  The device receives and validates one or more Bootstrap Seed
       Bundles through one or more policy-accepted distribution paths
       and verifies signatures, validity, scope, and supersession state
       before use.

   3.  Under bootstrap policy, the device performs first authority
       contact using the accepted seed material and a Bootstrap
       Admission Object, or a Recovery / Re-Enrolment Bundle when
       recovering existing trust state.

   4.  The contacted authority validates bootstrap or recovery
       artefacts, applies local bootstrap or recovery policy, and issues
       bounded initial or recovered identity and Grant state only within
       the bounds authorised for that transition.

Fisher                  Expires 17 September 2026              [Page 26]
Internet-Draft                    UZPIF                       March 2026

   5.  The device transitions into normal steady-state trust handling,
       after which ordinary UZPIF identity, Grant, revocation,
       transparency, scope, and policy validation rules apply.

   If the device is recovering after prior state loss, corruption, or
   compromise, it MUST NOT rely on a Bootstrap Admission Object alone
   when stronger recovery controls are required by policy.  Recovery
   admission MUST be explicit rather than hidden inside an apparently
   ordinary bootstrap ceremony.

   A HIL, manufacturer channel, or deployment installer MAY stage
   artefacts or carry first-contact traffic, but none of them is itself
   the baseline bootstrap mechanism.  The baseline mechanism is the
   validated trust-transition artefact chain plus the policy-controlled
   authority action described above.

8.2.3.  Initial Trust Establishment

   The zero-port trust model begins only after initial enrolment.  A
   device cannot originate within the closed trust domain; it must first
   be admitted to it through an explicit bootstrap process.  Bootstrap
   trust is therefore a first-class architectural concern and MUST NOT
   be treated as an implicit property of zero-port operation.

   A brand-new device, endpoint, RN, or HIL is not yet within the closed
   trust domain and may lack a Pantheon-recognised identity, Grant,
   trusted RN set, valid policy bundle, known authority set, secure time
   basis, prior session relationship, or attested operational state.
   Pre-enrolment identity establishment is therefore an explicit
   bootstrap transition rather than a mature steady-state protocol
   property.

   Initial enrolment depends on bootstrap artefacts and a bootstrap
   trust ceremony accepted under deployment policy.  At baseline, that
   means a validated Bootstrap Seed Bundle plus a Bootstrap Admission
   Object, or a Recovery / Re-Enrolment Bundle when prior trust state is
   being restored or replaced.  Deployments MUST define which bootstrap
   sources and Bootstrap Seed Bundles they accept rather than assuming
   any universal root of truth.  Acceptable bootstrap inputs MAY include
   manufacturer attestation, operator-installed seed trust, pre-
   provisioned deployment credentials, local physical ceremony,
   physically mediated enrolment, custodial onboarding authority,
   verified HIL-assisted enrolment, or enterprise MDM-style enrolment
   under signed policy.  Bootstrap authority MUST be explicit, bounded,
   and auditable.

Fisher                  Expires 17 September 2026              [Page 27]
Internet-Draft                    UZPIF                       March 2026

   Manufacturer identity or vendor material MAY be one acceptable
   bootstrap input, but it does not by itself solve enrolment, is not
   equivalent to a Bootstrap Admission Object or Recovery / Re-Enrolment
   Bundle, and MUST NOT silently become sovereign steady-state
   authority.  Admission into Pantheon or another local authority
   context still requires explicit acceptance under local policy via a
   Bootstrap Admission Object or stronger recovery artefact.

   A HIL MAY assist onboarding, staging, or mediated admission of non-
   native devices, but HIL presence does not by itself solve the
   bootstrap problem.  HIL-assisted enrolment still requires explicit
   acceptance of the HIL, its policy context, a validated Bootstrap Seed
   Bundle, and either a Bootstrap Admission Object or Recovery / Re-
   Enrolment Bundle under local deployment policy.

   Bootstrap trust does not, by itself, confer unlimited steady-state
   authority.  After enrolment, the normal UZPIF model for identity,
   Grant issuance, revocation, transparency, scope, and policy
   validation MUST apply.  Recovery after state loss, corruption, or
   compromise MUST transition through the Recovery / Re-Enrolment Bundle
   model rather than silently restoring prior authority as if no
   privileged transition had occurred.

8.3.  Key Custody and Signing-Role Separation

   Cryptographic separation of keys is not, by itself, sufficient if the
   same actor, machine, or operational role can exercise all practically
   decisive signing and administrative powers.  A deployment that
   concentrates issuance, policy, revocation, software-release, routing,
   and trust-bundle control into one custody domain has recreated a
   hidden sovereign control point even if each function uses a distinct
   key.

   Deployments SHOULD separate, where practical, at least some of the
   following roles:

   *  identity issuance;

   *  Grant or policy signing;

   *  revocation quorum participation;

   *  software release or patch manifest signing;

   *  RN operational administration; and

   *  transparency witnessing or equivalent append-only verification
      roles.

Fisher                  Expires 17 September 2026              [Page 28]
Internet-Draft                    UZPIF                       March 2026

   This document does not mandate one governance structure.  It does
   define the architectural principle that concentrated custody of all
   such roles SHOULD be avoided when stronger separation is practical,
   especially for deployments claiming high assurance, multi-party
   accountability, or resistance to unilateral control.

   Where small or local deployments combine roles, that concentration
   SHOULD be explicit, auditable, and bounded by stronger local controls
   such as threshold approval, independent logging, external witnessing,
   or stricter operational review.  Combined-role operation MUST NOT be
   treated as evidence that organisational separation is unnecessary at
   broader deployment scale.

8.4.  Client Sovereignty Model

   In UZPIF, clients are the ultimate trust arbiters at the edge.

   Clients MUST:

   *  validate RN transparency logs;

   *  verify Proof-of-Reachability (PoR) responses as defined by UZP
      ([UZP]); and

   *  independently determine trust decisions using local policy and
      cryptographic evidence.

8.5.  RN Controller Properties

   An RN Controller is the control-plane implementation used to operate
   one or more RNs.  The following requirements apply to RN Controller
   designs.

   RN Controllers MUST:

   *  be self-hostable without external approval;

   *  publish signed transparency logs;

   *  support cryptographic verifiability of routing behaviour; and

   *  allow client-side trust evaluation.

   RN Controllers MUST NOT:

   *  enforce global revocation unilaterally;

   *  depend on a single upstream authority; and

Fisher                  Expires 17 September 2026              [Page 29]
Internet-Draft                    UZPIF                       March 2026

   *  possess unilateral kill-switch capability.

8.5.1.  RN Non-Sovereignty and Compromise Containment

   Rendezvous Nodes are operationally significant, availability-
   sensitive infrastructure and MAY be high-value attack targets.
   However, UZPIF-compliant designs MUST ensure that RN compromise,
   local administrative access, or physical possession do not by
   themselves permit issuance of valid identity, Grant, revocation, or
   control-plane truth.  RN-relevant state transitions MUST be
   cryptographically validated, MUST fail closed on integrity mismatch,
   and MUST remain independently auditable.

   An RN MUST be treated as a traffic-stitching and rendezvous
   facilitator, a transparency-logged operator, and a cryptographically
   constrained participant.  It MUST NOT become a unilateral authority
   for identity, Grant issuance, revocation, software truth, or any
   other control-plane truth that would silently redefine system
   validity.

   RN behaviour MUST be externally verifiable through signed state,
   append-only transparency records, or other independently checkable
   evidence.  Compromise of one RN MUST NOT permit undetectable
   falsification of end-to-end identity or authorisation state.
   Physical control of RN infrastructure MUST NOT, by itself, be
   sufficient to create valid protocol artefacts, alter authoritative
   control-plane state, or impersonate endpoints without
   cryptographically detectable divergence.

   RN control-plane changes MUST follow deterministic validation rules.
   No RN SHALL apply unsigned or policy-ineligible control-plane state.
   No RN-observed traffic event SHALL be sufficient to mint valid
   endpoint identity or Grant state.  No local administrative path SHALL
   silently bypass the same verification rules required of normal
   control-plane transitions.

   RN software manifests, configurations, policy bundles, and updates
   SHOULD be signed by independent release authorities, and deployments
   SHOULD verify reproducible or attestable build identity where
   feasible.  No RN software, configuration, or policy update SHALL
   become authoritative unless its manifest, version, signer set, scope,
   and policy eligibility validate successfully.  Update or provisioning
   channels MUST NOT become hidden sovereign authority paths.  Clients
   and relying parties MUST NOT infer trust from RN software version
   claims alone.

Fisher                  Expires 17 September 2026              [Page 30]
Internet-Draft                    UZPIF                       March 2026

   Deployments SHOULD provide operator plurality, multi-RN alternatives,
   or client escape paths where practical so that no single RN or RN
   Controller is the sole required trust anchor for continued protocol
   validity.  Companion work on Invariant-Closed System Design (ICSD)
   ([ICSD]) is informative for structuring RN controller state so that
   rollback, drift, partial-write, unsigned-override, or otherwise
   invalid states are rejected rather than normalised.

8.5.2.  Multi-RN Consistency and Failover Semantics

   Deployments MAY use multiple RNs for resilience, locality, operator
   diversity, or high-assurance stitching.  In such deployments, RNs
   remain coordination points rather than sovereign issuers of identity,
   Grant, revocation, or other control-plane truth.  Divergent RN views
   MUST NOT, by themselves, create valid authorisation or identity
   truth.

   Deployments that require resilience SHOULD provision, discover, or
   otherwise maintain multiple candidate RNs acceptable under local
   policy.  RN identity, role eligibility, and participation in
   stitching MUST be independently verifiable from Pantheon artefacts,
   RN Set Statements, transparency evidence, and local policy rather
   than inferred merely from referral by another RN.  UZPIF does not
   require a single globally authoritative RN, and compliant deployments
   SHOULD avoid designs in which loss of one RN renders an otherwise
   healthy trust domain permanently unreachable.

8.5.2.1.  RN Set Advertisement and Discovery

   For interoperable RN discovery, a deployment SHOULD publish one or
   more RN Set Statements using the common signed artefact envelope with
   "object_type" set to "rn-set-statement".  A Bootstrap Seed Bundle or
   Recovery / Re-Enrolment Bundle MAY carry one or more RN Set
   Statements directly, or MAY carry references and digests that allow
   retrieval and validation of the current RN Set Statement under the
   same trust policy.

   Endpoints claiming resilient rendezvous MUST validate an RN Set
   Statement under the common envelope rules before using it for RN
   selection.  At baseline, that means verifying the signature set,
   trust-domain scope, validity interval, and epoch or sequence state,
   and preferring the highest eligible non-conflicting statement for the
   relevant scope.  Where policy and deployment topology permit,
   endpoints SHOULD maintain multiple currently eligible RNs rather than
   a single active candidate.

   An RN Set Statement MUST carry at least:

Fisher                  Expires 17 September 2026              [Page 31]
Internet-Draft                    UZPIF                       March 2026

   *  the issuing authority context or trust domain for which the RN set
      is valid;

   *  the eligible RN identifiers and sufficient contact or discovery
      hints for each RN;

   *  the RN roles, capabilities, or scope restrictions relevant to
      stitching, relay, or recovery use;

   *  the validity interval for the RN set;

   *  an epoch or sequence value sufficient for supersession and
      rollback detection;

   *  optional locality, priority, diversity, or operator-separation
      hints; and

   *  the signature set required by the deployment policy.

   RN referrals learned from an RN MAY be used as hints, but they MUST
   NOT by themselves establish RN eligibility.  Before an endpoint uses
   an alternate RN for authorisation-relevant operation, the RN MUST be
   accepted by a valid RN Set Statement, a Bootstrap Seed Bundle, a
   Recovery / Re-Enrolment Bundle, or another equally trusted signed
   bootstrap or recovery artefact accepted under local policy.

   An RN Set Statement is the baseline plurality artefact: it advertises
   which RNs are eligible for a given scope and role, but it does not by
   itself authorise a session or override current Grant, revocation,
   transparency, or other control-plane truth.

   If multiple RN Set Statements conflict for the same trust domain,
   scope, epoch, or sequence, endpoints MUST process them using the
   normal conflict rules for signed objects and MUST fail closed for
   authorisation-expanding actions until a non-conflicting eligible RN
   set is established.

8.5.2.2.  Alternate RN Eligibility and Failover Baseline

   An endpoint MAY treat an alternate RN as eligible only if all of the
   following are true:

   *  the RN appears in a currently valid RN Set Statement or equally
      trusted bootstrap or recovery artefact;

   *  the RN role and capability set are compatible with the requested
      stitching or recovery action;

Fisher                  Expires 17 September 2026              [Page 32]
Internet-Draft                    UZPIF                       March 2026

   *  the RN is not locally distrusted, revoked, or outside the
      currently accepted trust-domain scope; and

   *  the endpoint has enough fresh authority and continuity state to
      satisfy the chosen failover mode.

   Prior successful use of an RN does not by itself keep that RN
   eligible.  Eligibility MUST be re-evaluated at failover time against
   current RN Set Statements, current policy scope, current distrust or
   revocation state, and the exact stitching or recovery role the
   endpoint is attempting to use.

   Baseline failover triggers include RN loss, overload, local
   unreachability, policy-eligible alternate-RN preference, bounded
   continuity expiry, transparency inconsistency, partition suspicion,
   or signed or otherwise locally accepted evidence of RN misbehaviour.
   Triggering failover does not by itself alter identity, Grant
   validity, or authority truth.

   When a failover trigger fires, the endpoint enters failover handling,
   freezes new authorisation-expanding actions, and retains only
   portable state pending alternate-RN selection and revalidation.  It
   MUST NOT promote the failed or disconnected RN into a truth anchor
   merely because it was the last RN in use.

   The baseline portability rule is: identity, Grants, revocation state,
   Pantheon metadata, RN Set Statements, transparency checkpoints, and
   other externally verifiable authorisation artefacts are portable
   across RNs; RN-local queue placement, opaque stitch handles, pacing
   state, relay caches, and other non-verifiable relay state are not.

8.5.2.3.  Re-Stitching and Bounded Continuity State

   Fresh authorisation presentation is the default baseline for re-
   stitching at a new RN.  An alternate RN MAY accept bounded continuity
   state instead of full fresh authorisation only for non-expanding
   continuity when local policy permits it.

   Bounded continuity state MUST be externally verifiable or endpoint-
   verifiable and MUST be bound to the authenticated subject, peer,
   accepted Grant object identifiers or digests, relevant authority
   epoch or sequence state, and the accepted RN set.  It MUST carry its
   own expiry and MUST NOT be treated as standing authority.

   Failover MAY reuse bounded continuity state only when the resulting
   session scope is unchanged or narrower, the underlying Grant or
   equivalent authorisation remains valid, replay and freshness checks
   still succeed, and no unresolved RN disagreement or partition

Fisher                  Expires 17 September 2026              [Page 33]
Internet-Draft                    UZPIF                       March 2026

   condition would make the reused state ambiguous.  Otherwise, the
   endpoint MUST obtain fresh authorisation or perform a fresh re-stitch
   under current policy.

   Before an alternate RN returns a session to normal stitched
   operation, the endpoint and alternate RN MUST revalidate at least the
   chosen RN's current eligibility, the applicable RN Set Statement or
   equally trusted bootstrap or recovery artefact, current Grant and
   revocation state for the requested scope, any required transparency
   or misbehaviour evidence, and the binding, freshness, and expiry of
   any continuity state that will be resumed.

   Session continuity MUST NOT be confused with trust continuity.  Even
   when transport or application continuity is preserved, the alternate
   RN MUST revalidate the externally verifiable artefacts needed for the
   requested action and MUST NOT inherit opaque authorisation-relevant
   state solely because another RN previously held it.

8.5.2.4.  RN Disagreement and Partition Handling

   Loss, overload, unreachability, or partition of an RN is an
   availability event rather than an authorisation event.  Failure of
   one RN MUST NOT by itself invalidate endpoint identity, Grants,
   revocation state, or other externally verifiable control-plane truth.

   If two RNs or RN clusters present conflicting control views for the
   same session, subject, scope, epoch, or routing decision, endpoints
   and relying services MUST treat that disagreement as an availability
   or continuity issue rather than as proof that either view is newly
   authoritative.

   Under RN disagreement or suspected partition, endpoints MUST suspend
   new authorisation-expanding actions until the relevant artefacts are
   revalidated against externally verifiable authority.  Endpoints MAY
   continue narrowly scoped, already-established non-expanding operation
   only within an explicitly bounded continuity window allowed by local
   policy.  When the disagreement cannot be resolved within that window,
   implementations SHOULD prefer fresh re-stitching or clean session
   termination over silent continuity shortcuts.

   If only one RN remains reachable during disagreement or partition,
   endpoints MUST still treat external artefacts rather than that RN's
   local control view as authorisation truth.  A lone surviving RN is a
   remaining coordination path, not a newly elevated trust anchor.

Fisher                  Expires 17 September 2026              [Page 34]
Internet-Draft                    UZPIF                       March 2026

8.5.2.5.  RN Failover Continuity Constraints

   Failover MAY preserve continuity only if end-to-end trust assumptions
   are preserved.  An RN failover path MUST NOT silently weaken identity
   binding, freshness checks, replay resistance, transparency
   requirements, or downgrade posture merely to keep traffic flowing.
   When continuity and freshness are in tension, deployments SHOULD
   prefer fresh and independently verifiable authority for
   authorisation-expanding decisions, while allowing only bounded,
   policy-defined continuity behaviour for non-expanding recovery or
   session resumption.

   Session continuity across RN loss is not guaranteed.  A second RN or
   RN cluster MAY resume rendezvous only from externally verifiable
   artefacts and bounded continuity state accepted under local policy;
   it MUST NOT inherit opaque trust-relevant state solely because
   another RN previously held it.  When no safe resumable state exists,
   implementations SHOULD prefer fresh re-stitching over continuity
   shortcuts that weaken trust assumptions.

   This document does not require peer-to-peer fallback, HIL
   substitution, or reversion to inbound-port exposure as a universal
   recovery mechanism.  Profiles MAY define degraded or mediated
   recovery modes, but such modes MUST be explicit, policy-bound,
   auditable, and MUST NOT silently reintroduce weaker trust
   assumptions, hidden authority paths, or legacy open-listener exposure
   merely to preserve reachability.

8.6.  RN Transparency Log Format

   Each RN MUST publish a transparency log as an append-only sequence of
   signed entries.  Logs MUST be cryptographically verifiable,
   timestamped, and publicly retrievable.  Transparency entries used for
   interoperable verification MUST use the common signed artefact
   envelope defined in Section 5 with "object_type" set to "rn-
   transparency-entry".

   Each RN transparency log MUST contain signed records for:

   *  uptime attestations;

   *  session summaries at aggregated granularity; and

   *  error reports.

   Signed policy decisions are OPTIONAL but RECOMMENDED and SHOULD be
   published when policy constraints allow.

Fisher                  Expires 17 September 2026              [Page 35]
Internet-Draft                    UZPIF                       March 2026

   Each log entry MUST include at least a monotonically increasing
   sequence number, a timestamp, an entry type, a payload digest, and an
   RN signature.  In the common envelope, these properties are expressed
   via "object_type", "sequence", "issued_at", "prev_hash" or "log_ref",
   the object-specific body, and the embedded signature set.  Log
   interfaces MUST allow third parties to retrieve entries and
   independently verify append-only integrity.

   Evidence published for accountability SHOULD be sufficient for
   independent verification without unnecessarily exposing protected
   content, business-sensitive metadata, or broader activity graphs.

8.7.  Proof of RN Misbehaviour

   A Proof of RN Misbehaviour is a cryptographic evidence object that
   can be validated without a central judge.  Misbehaviour proofs used
   for interoperable exchange MUST use the common signed artefact
   envelope defined in Section 5 with "object_type" set to
   "misbehaviour-proof".

   Proofs MAY be constructed from one or more of the following evidence
   classes:

   *  signed evidence of misrouting;

   *  evidence that an RN failed to relay a required nonce during
      reachability procedures defined by UZP ([UZP]);

   *  log inconsistency proofs (for example, two different signed
      entries for the same sequence number) derived from the
      transparency material in Section 8.6; and

   *  signature replay detection evidence.

   Clients MUST be able to validate misbehaviour proofs locally using
   published keys and log material.  When proof validation succeeds,
   clients MAY blacklist the RN and SHOULD support policy controls for
   local blacklist retention and expiry.

   UZPIF MUST NOT require a token-slashing mechanism or any central
   adjudication entity for RN accountability.  Exclusion decisions are
   evidence-based and locally enforceable.

9.  Explicit Non-Goals

   To avoid governance ambiguity, the UZPIF stack does not define or
   require the following:

Fisher                  Expires 17 September 2026              [Page 36]
Internet-Draft                    UZPIF                       March 2026

   *  enforcement of a global content policy;

   *  mandatory jurisdictional compliance mechanisms;

   *  centralised identity arbitration; and

   *  global kill-switch functionality.

10.  Pantheon: Identity and Governance Model

   Pantheon is the identity, attestation, and policy plane for UZPIF.
   It MAY be deployed as a single authority or as a multi-authority
   federation.  As specified in Section 8, Pantheon federation is not a
   mandatory hierarchy and does not imply a single global authority.

   A deployment MUST NOT claim interoperable Pantheon federation unless
   relying parties can obtain authority metadata, discover issuers,
   apply cross-authority validation, and process conflicts or split-
   brain conditions using the rules in this section or a stricter
   profile.  Without those capabilities, Pantheon remains a local
   authority deployment rather than an interoperable federation.

   Pantheon objects exchanged across authorities MUST use the common
   signed artefact envelope defined in Section 5.

10.1.  Identity Model

   Pantheon authorities bind identity, policy, and trust metadata to
   keys or selectors accepted under policy.

   Deployments MAY use locally generated or externally attested keys if
   policy allows; Pantheon authorities validate or certify those
   bindings rather than being required to generate the underlying
   private keys.

   In the suite baseline, Pantheon authorities issue credentials,
   Grants, and delegations over accepted bindings rather than being
   required to generate or custody the underlying private keys.

   Pantheon commonly binds or delegates the following identity-bearing
   material:

   *  long-term public signing keys bound to CIDs;

   *  ephemeral session keys or ephemeral public keys bound to EIDs
      under policy; and

Fisher                  Expires 17 September 2026              [Page 37]
Internet-Draft                    UZPIF                       March 2026

   *  delegated sub-identities, selectors, or delegated credentials for
      services or microprocesses.

   This identity approach is conceptually aligned with HIP's separation
   of endpoint identities from locators [RFC7401], but elevated to a
   policy plane.

10.2.  Certificate Format

   Pantheon Certificates (PCerts) used for interoperable federation
   SHOULD include at least the following elements:

   *  an issuer authority identifier;

   *  a CID and public signing key (and optionally a post-quantum key
      encapsulation key);

   *  purpose tags (e.g., service, role, tenant);

   *  validity bounds (time or epoch); and

   *  optional attestation claims (e.g., hardware trust or enclave
      measurement).

   A PCert binds identity, scope, and trust metadata to an asserted
   public key.  It does not by itself state how the corresponding
   private key was generated or who holds custody of it unless explicit
   attestation claims in the object-specific body say so.

   Profiles MAY extend PCerts with additional fields, but interoperable
   federation requires enough issuer, subject, and validity information
   for cross-authority validation under Section 10.6.

   For interoperable exchange, a PCert MUST use the common signed
   artefact envelope with "object_type" set to "pcert".  The
   certificate-specific fields listed above populate the object-specific
   body and MUST NOT redefine envelope semantics.

10.3.  Grant Structure

   A Grant is described as a signed assertion binding:

   *  the CID/EID of the requester;

   *  the requested peer identity;

   *  purpose and action;

Fisher                  Expires 17 September 2026              [Page 38]
Internet-Draft                    UZPIF                       March 2026

   *  a time window and replay nonce; and

   *  an authorised quality-of-service (QoS) class.

   For interoperable exchange across the suite, a Grant MUST be
   represented as a signed artefact with "object_type" set to "grant",
   using the envelope defined in Section 5.

   This section is the canonical suite-level Grant definition for the
   architecture described in UZPIF.  Companion drafts such as UZP, TLS-
   DPA, and outbound indexing may add object-specific fields or
   application-specific constraints in the object-specific body, but
   they MUST preserve this baseline Grant semantics and the shared
   artefact envelope.

10.4.  Authority Identifiers and Metadata

   Each Pantheon authority participating in a federation MUST be
   identified by a stable authority_id.  An authority_id MUST remain
   stable across ordinary signing-key rollover and MUST be unique within
   the relying party's trust context.

   Each authority MUST publish signed authority metadata sufficient for
   relying parties and peer authorities to validate issued objects.  At
   a minimum, the metadata MUST identify:

   *  the authority_id;

   *  one or more current signing keys, and optionally future or
      retiring keys for rollover;

   *  service endpoints used for issuing or retrieving Pantheon objects;

   *  a metadata validity interval;

   *  the supported object types (for example PCerts, Grants,
      delegations, revocations, or attestation artefacts); and

   *  an optional transparency-log endpoint.

   Metadata MAY contain additional deployment-specific policy claims,
   but Pantheon federation MUST NOT depend on a single global metadata
   registry or a mandatory schema beyond the minimum semantics above.

Fisher                  Expires 17 September 2026              [Page 39]
Internet-Draft                    UZPIF                       March 2026

10.5.  Issuer Discovery

   Issuer discovery is deliberately non-centralised.  A relying party
   MAY discover an authority and its signed metadata through one or more
   of the following mechanisms:

   *  local trust bundles;

   *  signed referrals;

   *  cached authority metadata; and

   *  out-of-band provisioning.

   No global registry, mandatory directory, or single discovery service
   is required for Pantheon federation.  Deployments MAY combine
   discovery mechanisms and MAY continue to rely on cached metadata when
   fresh discovery is temporarily unavailable, subject to local expiry
   policy.  An authority without discoverable or locally provisioned
   metadata MUST NOT be treated as an interoperable federation
   participant.

10.6.  Cross-Authority Validation

   A client, RN, or other relying service evaluating a PCert, Grant,
   revocation statement, delegation, or other signed Pantheon object
   MUST accept that object only if its issuer is:

   *  locally trusted;

   *  delegated from a locally trusted authority, with a delegation
      chain valid for the relevant object type and scope; or

   *  accepted under a profile-defined federation rule.

   If none of these conditions holds, the object MUST be rejected.
   Validation MUST include signature verification, issuer metadata
   validity, object-type authorisation, scope, and time validity.

   Acceptance of one object type from an authority does not
   automatically authorise acceptance of other object types unless the
   applicable delegation or federation profile explicitly permits it.

10.7.  Conflict and Split-Brain Handling

   Pantheon federation MUST distinguish between issuer misbehaviour and
   federation divergence.  These cases are not equivalent and MUST NOT
   be processed using a single merge rule.

Fisher                  Expires 17 September 2026              [Page 40]
Internet-Draft                    UZPIF                       March 2026

   If the same authority issues conflicting signed objects for the same
   epoch and scope, relying parties MUST treat the condition as evidence
   of authority misbehaviour.  The conflicting objects SHOULD be
   retained as auditable evidence and MAY trigger local distrust,
   quarantine, or incident-response policy.

   If different authorities publish conflicting signed views for the
   same subject, scope, or revocation state, relying parties MUST treat
   the condition as federation divergence.  Such conflicts MUST NOT be
   auto-merged or silently normalised.  Resolution MUST be performed by
   local policy, an explicit federation profile, or quorum rules agreed
   by the deployment.

   Until resolved, implementations SHOULD fail closed for authorisation-
   expanding decisions and SHOULD surface the divergence to operators or
   relying services with enough issuer metadata to support audit and
   review.

10.8.  Attestation Model

   RNs and endpoints may publish attestations such as:

   *  hardware and software measurements;

   *  configuration hashes; and

   *  policy compliance data.

   Attestation artefacts SHOULD be published in transparency logs
   retrievable by relying parties.  Storage and serving MAY be provided
   by Pantheon services, RN Controllers, or both, mirroring transparency
   and verifiability goals in broader zero-trust guidance.
   [NIST-SP800-207]

10.9.  Caching Rules

   The following caching guidance is illustrative deployment guidance
   rather than a protocol-level interoperability requirement.
   Deployment profiles MAY tighten or relax these values according to
   their revocation, freshness, and transparency policies.

Fisher                  Expires 17 September 2026              [Page 41]
Internet-Draft                    UZPIF                       March 2026

   Authenticity alone is insufficient for continued reliance on cached
   authority.  Before cached metadata, PCerts, Grants, or attestation
   material are used for security-relevant decisions, implementations
   SHOULD evaluate freshness, scope, sequence or epoch where applicable,
   and current policy eligibility under local stale-state rules.
   Authorisation-expanding decisions SHOULD require revalidation once a
   locally permitted stale window has been exceeded or when superseding
   state may exist.

   Endpoints may cache:

   *  authority metadata for the shorter of its stated validity interval
      or a locally configured cache cap;

   *  PCerts for a locally configured interval appropriate to the
      deployment's revocation and freshness model (for example, up to 24
      hours in low-churn environments);

   *  Grants for the shorter of their validity window or session
      lifetime; and

   *  attestation proofs for the duration of RN handshake paths.

11.  Benefits and Trade-offs

   UZPIF and UZP ([UZP]) intentionally reuse established transport and
   cryptographic primitives, but change where and how they are bound to
   identity, policy, and reachability.  In particular, QUIC [RFC9000]
   and TLS 1.3 [RFC8446] demonstrate encrypted transports with modern
   handshake properties, while UZPIF shifts the reachability model away
   from listening endpoints.

Fisher                  Expires 17 September 2026              [Page 42]
Internet-Draft                    UZPIF                       March 2026

    +==============+======================+==============+============+
    | Dimension    | UZPIF/UZP ([UZP])    | Traditional  | QUIC/TLS   |
    |              |                      | TCP/TLS      |            |
    +==============+======================+==============+============+
    | Exposure     | No open ports        | Open ports   | Open ports |
    |              | (endpoints are not   | and/or       | at network |
    |              | publicly listening)  | proxies      | edge       |
    +--------------+----------------------+--------------+------------+
    | Identity     | Mandatory            | Typically    | TLS-level  |
    |              | cryptographic        | TLS-level    | identity   |
    |              | identity             | identity     |            |
    |              |                      | only         |            |
    +--------------+----------------------+--------------+------------+
    | Transport    | Defined by companion | TCP byte-    | QUIC       |
    | semantics    | transport profile    | stream       | stream     |
    |              |                      | transport    | transport  |
    +--------------+----------------------+--------------+------------+
    | RN trust     | Drop/delay only (no  | N/A          | N/A        |
    |              | end-to-end plaintext |              |            |
    |              | visibility expected) |              |            |
    +--------------+----------------------+--------------+------------+
    | Latency      | Deterministic pacing | Congestion   | Congestion |
    | control      | (design goal)        | control      | control    |
    |              |                      | variants     | variants   |
    |              |                      | (e.g., Reno/ | (e.g.,     |
    |              |                      | CUBIC)       | BBR/CUBIC) |
    +--------------+----------------------+--------------+------------+
    | Legacy       | Supported via HIL    | Native       | Often      |
    | applications | (intended)           |              | requires   |
    |              |                      |              | gateways   |
    |              |                      |              | or         |
    |              |                      |              | adaptation |
    +--------------+----------------------+--------------+------------+
    | Post-quantum | Designed for         | Inconsistent | Emerging   |
    | readiness    | cryptographic        | deployment   |            |
    |              | agility              |              |            |
    +--------------+----------------------+--------------+------------+

               Table 2: Comparison of transport architectures

12.  Threat Model

   This section sketches attacker classes and example controls.  It is
   not a complete security analysis and will evolve with implementation
   experience.

   *Attacker classes* include:

Fisher                  Expires 17 September 2026              [Page 43]
Internet-Draft                    UZPIF                       March 2026

   *  Internet-wide scanners;

   *  botnets seeking command-and-control beacons;

   *  malicious RNs (assumed capable of drop/delay/reorder);

   *  insiders with credentials; and

   *  traffic analysts performing correlation.

   Existing rendezvous and overlay systems (e.g., Tor [Tor]) and NAT
   traversal mechanisms based on STUN [RFC5389] and TURN [RFC5766]
   demonstrate the power of indirection, but they still assume exposed
   or discoverable listeners somewhere in the path.  UZPIF's design
   intent is to remove those listeners from the endpoint security model.

   Example controls discussed for UZPIF include:

   *  end-to-end authenticated encryption (AEAD);

   *  RN and endpoint attestation;

   *  puzzles and identity-bound rate limits;

   *  multi-RN stitching for higher assurance; and

   *  post-quantum readiness and cryptographic agility (see [NIST-PQC]).

13.  Economics

   *  Capital expenditure reduction: reduced reliance on perimeter
      appliances and complex DMZ designs.

   *  Operational expenditure reduction: fewer ACL/NAT rule changes and
      less inbound exposure management.

   *  Risk reduction: reduced externally visible attack surface.

   *  Potential service models: governance and RN validation as managed
      components.

14.  Migration Plan for Organisations

   UZPIF is intended for incremental deployment alongside existing TCP/
   TLS and QUIC-based stacks [RFC8446] and [RFC9000].

   1.  *Deploy an RN:* Introduce an outbound-only rendezvous node.

Fisher                  Expires 17 September 2026              [Page 44]
Internet-Draft                    UZPIF                       March 2026

   2.  *Deploy the HIL:* Install the Hardware Integration Layer at
       endpoints or device edges where legacy applications or hardware
       require mediated compatibility.

   3.  *Dual-stack operation:* Run UZP ([UZP]) alongside existing TCP/
       TLS.

   4.  *Cutover:* Migrate services gradually to zero-port operation.

15.  Security Considerations

   UZPIF's central security claim is that avoiding publicly reachable
   listeners at endpoints reduces exposure to scanning and unsolicited
   ingress.  However, the framework introduces reliance on identity,
   authorisation, and policy evaluation components (e.g., Pantheon and
   RNs) whose compromise or misconfiguration could impact availability
   and authorisation correctness.

   The threat model in Section 12 discusses attacker classes and
   candidate controls.  Future revisions of this document (and the
   companion UZP ([UZP]) and TLS-DPA ([TLS-DPA]) documents) are expected
   to provide a more systematic analysis, including key management,
   revocation, attestation trust, and traffic analysis resistance.

15.1.  Transition of Network Security Roles

   UZPIF shifts primary access control from network-coordinate filtering
   toward cryptographic identity, Grants, and invariant-bound policy
   enforcement.  Network address and open-port exposure therefore cease
   to be the primary admission model in a native UZPIF deployment.

   Traditional firewalls, DPI systems, and other middleboxes MAY remain
   useful for containment, telemetry, egress control, QoS enforcement,
   volumetric denial-of-service handling, forensic monitoring, HIL
   containment, regulatory inspection domains, and non-UZPIF legacy
   enclaves.  However, they MUST NOT be treated as authoritative truth
   sources for identity or policy validity.

   UZPIF does not assume that packet visibility implies authority.
   Where middleboxes remain deployed, they operate as bounded policy,
   containment, or observability components rather than as trust anchors
   that define whether an identity or Grant is valid.

Fisher                  Expires 17 September 2026              [Page 45]
Internet-Draft                    UZPIF                       March 2026

15.2.  Compliance, Observability, and Intercept Boundaries

   Some deployments will require explicit policy enforcement, enterprise
   inspection, regulated access, or other observability controls.  This
   document does not define a universal lawful-intercept or hidden
   inspection authority for UZPIF.

   Where compliance, monitoring, mediation, or inspection components are
   deployed, they MUST remain explicit, bounded, and cryptographically
   distinguishable from protocol truth.  The existence of such a
   component MUST NOT silently redefine endpoint identity, Pantheon
   authority, Grant validity, revocation state, or other authorisation-
   relevant truth in the UZPIF suite.

   A compliance or inspection function MAY enforce local policy, provide
   observability, or participate in a regulated deployment boundary, but
   it MUST be treated as a deployment-specific control surface rather
   than as implicit sovereign protocol authority.  Operator or
   institutional pressure to insert monitoring MUST NOT be satisfied by
   silently converting middleboxes, RNs, HILs, or auxiliary services
   into hidden trust anchors.

15.3.  Identity-Aware Observability Shift

   Network-address and port-oriented monitoring may lose significant
   explanatory power in UZPIF environments.  Legacy network-centric
   visibility therefore becomes incomplete rather than useless:
   defenders may see less value in source or destination address
   patterns, open-port exposure, unsolicited inbound attempts, protocol
   signatures, or packet-shape heuristics than in conventional port-
   based environments.

   This does not imply that UZPIF blinds defenders.  It does mean that
   security analytics SHOULD rely increasingly on identity-aware
   telemetry, authority artefacts, Grant usage patterns, rendezvous
   behaviour, and policy-verification events rather than on legacy
   assumptions about open-port exposure.  Otherwise, Shadow IT activity,
   stealthy misuse of valid outbound behaviour, unauthorised stitching
   requests, abuse of legitimate-looking identity objects, or
   compromised internal workloads acting within apparently allowed
   tunnel patterns may become harder to explain or detect.

   Privacy-related properties in UZPIF MUST be separated.  Reduced
   service discoverability means that publicly reachable listeners and
   routine unsolicited scanning lose some value.  Encrypted content
   protection means that authorised endpoints can protect payloads and
   application content from intermediaries.  Traffic-pattern privacy
   concerns whether observers can still infer communication existence,

Fisher                  Expires 17 September 2026              [Page 46]
Internet-Draft                    UZPIF                       March 2026

   counterpart relationships, cadence, timing, or operational
   significance from metadata.  Improvement in one of these properties
   MUST NOT be represented as automatically delivering the other two.

   Effective deployment profiles SHOULD expose auditable telemetry such
   as Grant issuance and usage logs, RN selection patterns, failed
   stitching attempts, revocation-consistency signals, identity or
   authority anomalies, HIL mediation events, downgrade or
   compatibility-mode activation, and unusual policy-bound session
   graphs.  Such telemetry SHOULD remain explicit, policy-bound, and
   reviewable rather than being inferred indirectly from legacy packet
   visibility alone.

15.4.  Time as a Trust Dependency

   Time is a security-relevant input in the UZPIF suite.  Grant expiry,
   revocation freshness, log checkpoints, replay defence, policy
   activation windows, and software or patch manifest validity all
   depend on time evaluation.  Implementations MUST therefore treat
   local time and time-derived freshness decisions as part of the
   trusted security boundary rather than as a purely operational
   convenience.

   Policy decisions that depend on time SHOULD tolerate bounded clock
   drift as defined by deployment policy or a stricter profile.
   However, no single unauthenticated time source SHALL become a
   mandatory truth anchor for continued protocol validity.  Deployments
   SHOULD use independently checkable, redundant, or otherwise policy-
   vetted time inputs where time correctness materially affects
   authorisation or freshness decisions.

   When time is uncertain, inconsistent, or outside locally permitted
   drift bounds, implementations SHOULD degrade safely rather than
   silently accepting stale or not-yet-valid authority.  Safe
   degradation MAY include rejecting time-sensitive artefacts, requiring
   revalidation, narrowing replay windows, surfacing operator alarms, or
   falling back to more conservative local policy.

   This document does not define a suite-wide time synchronisation
   protocol.  It does define the security posture that incorrect or
   manipulated time MUST NOT silently widen authority, revive expired
   Grants, delay emergency revocation effect, or normalise ambiguous
   checkpoint ordering without explicit local policy.

Fisher                  Expires 17 September 2026              [Page 47]
Internet-Draft                    UZPIF                       March 2026

15.5.  Replay, Caching, and Stale Authority

   Authenticity alone is insufficient for security-relevant artefacts in
   the UZPIF suite.  Grants, revocation signals and threshold evidence
   as profiled by TLS-DPA ([TLS-DPA]), PoR artefacts as defined by UZP
   ([UZP]), Proofs of RN Misbehaviour (Section 8.7), transparency
   checkpoints and receipts as profiled by outbound indexing
   ([OUTBOUND-INDEXING]), and similar signed objects MUST also be
   evaluated for freshness, scope, sequence or epoch, audience where
   relevant, and current policy eligibility before they influence
   authorisation or trust decisions.

   An old but well-signed artefact MUST NOT be accepted merely because
   its signature validates.  Implementations MUST determine whether the
   artefact has expired, has been superseded by a newer eligible
   artefact, falls outside the current policy scope, or remains within
   any locally permitted stale window for the decision being made.

   Profiles SHOULD define bounded stale windows for continued operation
   during outage or partition conditions.  Within such a window,
   implementations MAY continue to rely on cached authority for narrowly
   scoped, non-expanding behaviour if local policy permits it.  Outside
   that window, or when freshness cannot be established, systems SHOULD
   require revalidation before continuing security-relevant operation
   and SHOULD fail closed for authorisation-expanding decisions.

15.6.  Downgrade and Compatibility-Mode Abuse

   UZPIF deployments may support native operation, HIL-mediated
   operation, brokered compatibility modes, legacy transport fallback,
   optional transparency behaviour, or advisory revocation thresholds.
   These modes create downgrade risk if an attacker can force the system
   onto a weaker posture while preserving the appearance of normal
   operation.

   Fallback and degraded modes MUST be explicit, observable, and policy-
   bound.  An implementation MUST NOT silently downgrade from native
   identity-first operation to mediated, brokered, legacy-compatible,
   transparency-reduced, or revocation-weakened behaviour without an
   allowed policy rule and locally observable state indicating that the
   downgrade occurred.

Fisher                  Expires 17 September 2026              [Page 48]
Internet-Draft                    UZPIF                       March 2026

   Peers, operators, or relying services SHOULD be able to determine
   whether a session or decision is operating in native, mediated,
   brokered, or otherwise degraded mode.  Degraded operation MUST NOT
   silently inherit the same trust posture, authority scope, or audit
   assumptions as native operation.  Profiles SHOULD narrow permissions,
   reduce session lifetime, strengthen revalidation requirements, or
   require additional operator visibility when compatibility or degraded
   modes are active.

   Suppressed transparency checks, unavailable revocation evidence,
   weaker identity binding rules, or temporary compatibility exceptions
   MUST be treated as explicit trust reductions rather than harmless
   operational details.  If the required stronger mode cannot be
   established, implementations SHOULD fail closed for authorisation-
   expanding behaviour unless local policy explicitly permits the weaker
   mode for a bounded scope and duration.

15.7.  Break-Glass and Emergency Override Discipline

   Real-world deployments may require emergency action during outage,
   recovery, migration, or incident containment.  However, a break-glass
   path that silently disables normal trust checks is itself a high-risk
   authority path.

   Any emergency override MUST be explicit, narrowly scoped, time-
   bounded, auditable, and incapable of silently becoming normal steady-
   state policy.  A break-glass decision MUST NOT be normalised as an
   invisible standing exception to identity validation, Grant
   evaluation, revocation processing, transparency expectations, or
   other authorisation-relevant checks in the UZPIF suite.

   Deployment profiles SHOULD record the triggering reason, affected
   scope, approving authority, activation time, expiry condition, and
   review outcome for such overrides.  Emergency operation SHOULD
   surface clear local state to operators and relying parties when the
   stronger normal posture has been reduced.

15.8.  Metadata Leakage and Traffic Analysis

   UZPIF can materially reduce direct service and endpoint
   discoverability without guaranteeing metadata indistinguishability.
   Removing publicly reachable listening ports does not, by itself,
   provide strong privacy or anonymity.  Unscannable is not the same as
   unseen.  Rendezvous, transparency, and indexing infrastructure MAY
   still expose valuable metadata such as which parties communicate,
   which RN or RN cluster they use, how often they communicate, when
   sessions occur, how long they persist, retry patterns, topology
   clusters, discovery behaviour, revocation bursts, patch or rollout

Fisher                  Expires 17 September 2026              [Page 49]
Internet-Draft                    UZPIF                       March 2026

   waves, and whether bursts correlate with alarms, updates, or
   emergency operations.

   Even where payload plaintext remains protected, this metadata can
   support correlation, surveillance, tenant mapping, operational
   fingerprinting, or coercive analysis.  A Rendezvous Node or indexing
   service that is not a truth anchor may still become a surveillance
   anchor if deployments over-retain, over-expose, or casually
   centralise metadata visibility.

   Reduced discoverability, encrypted content protection, and traffic-
   pattern privacy are separate design axes.  UZPIF can reduce
   unsolicited inbound exposure and support protected payload transport
   while still leaving observable metadata patterns available to
   infrastructure operators or network observers.  Observers such as
   carriers, backbone operators, ISPs, or state-level monitors may still
   infer communication existence, cadence, rendezvous relationships, and
   operational timing from outbound rendezvous patterns unless
   additional privacy-preserving measures are deployed.  Claims about
   strong observation resistance or traffic-pattern privacy MUST
   therefore be bounded and profile-dependent.

   Operators SHOULD minimise metadata retention, restrict unnecessary
   visibility, and avoid collecting or exposing finer-grained metadata
   than is needed for security, accountability, or operational
   continuity.  Deployment profiles SHOULD define retention bounds,
   access controls, aggregation rules, and disclosure controls for
   security-relevant logs, receipts, checkpoints, and session records.

   Privacy-sensitive profiles MAY use identifier rotation, cover
   traffic, padding, batching, aggregation, timing obfuscation, multi-RN
   distribution, traffic shaping, delayed publication, or metadata-
   minimising RN policy to reduce correlation risk where those measures
   are compatible with the deployment's accountability and performance
   goals.  Removal of inbound ports alone MUST NOT be treated as
   sufficient support for broader privacy claims.

15.9.  Rendezvous Concentration and Availability Pressure

   UZPIF and its companion mechanisms introduce stateful verification
   paths, including handshake processing, PoR validation, Grant
   verification, transparency checking, authority discovery or
   resolution, repeated stitching attempts, and HIL-mediated translation
   work.  These paths may improve exposure relative to openly listening
   services while still creating denial-of-service or state-exhaustion
   risk if attackers can force expensive pre-authorisation work at
   scale.

Fisher                  Expires 17 September 2026              [Page 50]
Internet-Draft                    UZPIF                       March 2026

   Removing inbound exposure from endpoints does not eliminate denial-
   of-service risk; it shifts strategic pressure toward rendezvous and
   control-plane infrastructure.  UZPIF therefore redistributes attack
   surface away from broad endpoint exposure and toward smaller sets of
   highly defended coordination infrastructure.  UZPIF-compliant
   deployments MUST therefore treat RN availability as a first-class
   design concern rather than as an operational afterthought.

   RNs are likely focal points for volumetric, state-exhaustion, and
   coordination-disruption attacks.  Endpoint hardening alone does not
   solve network-wide availability.  Deployments SHOULD assume
   deliberate RN-targeted overload attempts and SHOULD use plurality,
   geographic dispersion, bounded-state processing, rate controls,
   admission pacing, queue isolation, and failure isolation where
   practical.  The availability of RN infrastructure also depends partly
   on underlying transport and routing ecosystems that are outside the
   protocol's direct control, so protocol and deployment design SHOULD
   minimise the consequences of temporary RN overload, routing
   impairment, or RN loss.

   Expensive cryptographic verification, policy evaluation, federation
   lookups, or translation work SHOULD occur as late as safely possible
   in the processing path.  Stateless screening, bounded-prestate
   admission checks, puzzles, quotas, or other low-cost filters SHOULD
   be preferred where practical before allocating scarce state or
   invoking high-cost validation.

   RNs, HILs, and related control-plane services SHOULD rate-limit,
   queue-bound, or otherwise constrain untrusted requests before high-
   cost cryptographic or policy evaluation where feasible.
   Implementations SHOULD separate basic survivability controls from
   richer verification paths so that overload of one validation function
   does not automatically collapse unrelated session handling or local
   containment.

   Basic survivability SHOULD NOT depend on a single always-available
   validation bottleneck such as one authority-resolution path, one
   transparency endpoint, or one high-cost policy service.  Deployment
   profiles SHOULD define bounded-failure behaviour, degraded but
   observable operation, and local overload policy so that exhaustion
   pressure does not silently widen authority or force unsafe fallback.

16.  IANA Considerations

   This document has no IANA actions.

   UZPIF is an architectural framework and does not define protocol
   parameters requiring registries.

Fisher                  Expires 17 September 2026              [Page 51]
Internet-Draft                    UZPIF                       March 2026

17.  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>.

   [RFC8785]  Rundgren, A., Jordan, B., and S. Erdtman, "JSON
              Canonicalization Scheme (JCS)", RFC 8785,
              DOI 10.17487/RFC8785, June 2020,
              <https://www.rfc-editor.org/info/rfc8785>.

18.  Informative References

   [ICSD]     Fisher, B. A., "Invariant-Closed System Design (ICSD)",
              Work in Progress, Internet-Draft, draft-dpa-icsd,
              <https://datatracker.ietf.org/doc/html/draft-dpa-icsd>.

   [NIST-PQC] National Institute of Standards and Technology, "NIST
              Post-Quantum Cryptography Standardization Project", 2022,
              <https://csrc.nist.gov/Projects/post-quantum-
              cryptography>.

   [NIST-SP800-207]
              Rose, S., Borchert, O., Mitchell, S., and S. Connelly,
              "Zero Trust Architecture", NIST SP 800-207, 2019,
              <https://doi.org/10.6028/NIST.SP.800-207>.

   [OUTBOUND-INDEXING]
              Fisher, B. A., "Outbound Indexing over UZPIF: Consent-
              First Discovery and Indexing", Work in Progress, Internet-
              Draft, draft-dpa-uzpif-outbound-indexing,
              <https://datatracker.ietf.org/doc/html/draft-dpa-uzpif-
              outbound-indexing>.

   [RFC5389]  Rosenberg, J., Mahy, R., Matthews, P., and D. Wing,
              "Session Traversal Utilities for NAT (STUN)", RFC 5389,
              DOI 10.17487/RFC5389, October 2008,
              <https://www.rfc-editor.org/info/rfc5389>.

Fisher                  Expires 17 September 2026              [Page 52]
Internet-Draft                    UZPIF                       March 2026

   [RFC5766]  Mahy, R., Matthews, P., and J. Rosenberg, "Traversal Using
              Relays around NAT (TURN): Relay Extensions to Session
              Traversal Utilities for NAT (STUN)", RFC 5766,
              DOI 10.17487/RFC5766, April 2010,
              <https://www.rfc-editor.org/info/rfc5766>.

   [RFC7401]  Moskowitz, R., Ed., Heer, T., Jokela, P., and T.
              Henderson, "Host Identity Protocol Version 2 (HIPv2)",
              RFC 7401, DOI 10.17487/RFC7401, April 2015,
              <https://www.rfc-editor.org/info/rfc7401>.

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

   [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>.

   [TLS-DPA]  Fisher, B. A., "TLS-DPA: An Identity-Bound Security
              Protocol for Traditional, Overlay, and Zero-Port
              Transports", Work in Progress, Internet-Draft, draft-dpa-
              tls-dpa,
              <https://datatracker.ietf.org/doc/html/draft-dpa-tls-dpa>.

   [Tor]      Dingledine, R., Mathewson, N., and P. Syverson, "Tor: The
              Second-Generation Onion Router", 2004.

   [UZP]      Fisher, B. A., "UZP: Universal Zero-Port Transport
              Protocol", Work in Progress, Internet-Draft, draft-dpa-
              uzp-transport, <https://datatracker.ietf.org/doc/html/
              draft-dpa-uzp-transport>.

Author's Address

   Benjamin Anthony Fisher
   DPA R&D Ltd (https://www.dpa-cloud.co.uk)
   Email: b.fisher@dpa-cloud.co.uk
   URI:   https://orcid.org/0009-0004-4412-2269

Fisher                  Expires 17 September 2026              [Page 53]