Network Working Group A. Bisht Internet-Draft Cisco Intended status: Informational 18 November 2026 Expires: 22 May 2027 Client-Visible Relay Diagnostics for Media over QUIC Transport (MOQT) draft-altanai-moq-relay-diagnostics-00 Abstract Media over QUIC Transport (MOQT) deployments typically place one or more relays between publishers and subscribers. When delivery stalls, degrades, or fails, a client observes only its own leg of the path, and often cannot determine whether the cause lies with the original publisher, an intermediate relay, path conditions, authorization policy, or its own behavior. This document collects use cases and requirements for exposing a limited, relay-controlled set of diagnostic information to authorized MOQT clients. It examines three candidate mechanisms: an explicit request and response exchange on the control path, diagnostic information carried in existing responses and error codes, and advertisement of diagnostic capability during session establishment. Because such information can reveal deployment topology, operational state, or the activity of other users, authorization, data minimization, and relay discretion over what is disclosed are treated as first-order requirements rather than later refinements. Related operator examples describe capacity and incident observations within a separately authorized administrative scope. They motivate correlation with existing metrics and logging systems without defining a relay management API. The goal is diagnostics that are actionable at runtime and interoperable across implementations, not a general-purpose telemetry channel within MOQT. Information intended primarily for offline analysis remains the province of logging and observability systems such as qlog. This document defines the problem space and the design constraints; a concrete wire format is left to a follow-on document once the working group has settled on an approach. Status of This Memo This Internet-Draft is submitted in full conformance with the provisions of BCP 78 and BCP 79. Bisht Expires 22 May 2027 [Page 1] Internet-Draft MOQT Relay Diagnostics November 2026 Internet-Drafts are working documents of the Internet Engineering Task Force (IETF). Note that other groups may also distribute working documents as Internet-Drafts. The list of current Internet- Drafts is at https://datatracker.ietf.org/drafts/current/. Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet-Drafts as reference material or to cite them other than as "work in progress." This Internet-Draft will expire on 22 May 2027. Copyright Notice Copyright (c) 2026 IETF Trust and the persons identified as the document authors. All rights reserved. This document is subject to BCP 78 and the IETF Trust's Legal Provisions Relating to IETF Documents (https://trustee.ietf.org/ license-info) in effect on the date of publication of this document. Please review these documents carefully, as they describe your rights and restrictions with respect to this document. Code Components extracted from this document must include Revised BSD License text as described in Section 4.e of the Trust Legal Provisions and are provided without warranty as described in the Revised BSD License. Table of Contents 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 3 1.1. Note to Readers . . . . . . . . . . . . . . . . . . . . . 3 1.2. Conventions . . . . . . . . . . . . . . . . . . . . . . . 3 1.3. Terminology . . . . . . . . . . . . . . . . . . . . . . . 4 2. Problem Statement . . . . . . . . . . . . . . . . . . . . . . 5 3. Use Cases . . . . . . . . . . . . . . . . . . . . . . . . . . 5 3.1. Distinguishing a Local Request Outcome . . . . . . . . . 5 3.2. Diagnosing a Delivery Stall . . . . . . . . . . . . . . . 5 3.3. Explaining a Relay Decision . . . . . . . . . . . . . . . 6 3.4. Supporting an Operator-Assisted Incident . . . . . . . . 6 3.5. Learning Whether Diagnostics Are Available . . . . . . . 6 4. Illustrative Diagnostic Observations . . . . . . . . . . . . 6 4.1. Counting and Measurement Scope . . . . . . . . . . . . . 7 4.2. Example and Disclosure Limits . . . . . . . . . . . . . . 8 5. Operator Diagnostics at Operational Scale . . . . . . . . . . 9 5.1. Aggregation and Access Boundaries . . . . . . . . . . . . 11 6. Requirements . . . . . . . . . . . . . . . . . . . . . . . . 11 6.1. Diagnostic Scope . . . . . . . . . . . . . . . . . . . . 12 6.2. Privacy and Authorization . . . . . . . . . . . . . . . . 12 6.3. Safety and Operational Behavior . . . . . . . . . . . . . 13 Bisht Expires 22 May 2027 [Page 2] Internet-Draft MOQT Relay Diagnostics November 2026 7. Candidate Mechanisms . . . . . . . . . . . . . . . . . . . . 13 7.1. Explicit Request and Response on the Control Path . . . . 13 7.2. Diagnostic Information in Existing Responses and Error Codes . . . . . . . . . . . . . . . . . . . . . . . . . . 14 7.3. Capability Advertisement During Session Establishment . . 14 8. Relationship to Other Work . . . . . . . . . . . . . . . . . 14 9. Security Considerations . . . . . . . . . . . . . . . . . . . 15 10. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 16 11. References . . . . . . . . . . . . . . . . . . . . . . . . . 16 11.1. Normative References . . . . . . . . . . . . . . . . . . 16 11.2. Informative References . . . . . . . . . . . . . . . . . 16 Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . 17 Author's Address . . . . . . . . . . . . . . . . . . . . . . . . 17 1. Introduction Media over QUIC Transport (MOQT) [MoQTransport] distributes media through a publish/subscribe model in which relays sit between publishers and subscribers. A subscriber therefore terminates only the last hop of a multi-hop path, and the signals available to it locally are rarely sufficient to explain why delivery is degraded or has failed. The term "diagnostics" comes from [Issue1692], which asks how a client can get diagnostic data from a relay. Discussion following IETF 126 helped shape the use cases and requirements for any future wire format. 1.1. Note to Readers This revision is a problem statement and requirements document. It defines no protocol elements. Comments are welcome on the MOQ working group mailing list. 1.2. Conventions The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in BCP 14 [RFC2119] [RFC8174] when, and only when, they appear in all capitals, as shown here. Bisht Expires 22 May 2027 [Page 3] Internet-Draft MOQT Relay Diagnostics November 2026 1.3. Terminology A *client-visible relay diagnostic* is a limited report of what one relay observed or did while handling a particular request. For example, when a subscription stalls, the relay may report that it handed Objects to its outgoing transport for that subscription. Whether those Objects reached the subscriber or why playback stalled may remain unknown. This document uses *operator telemetry* for the metrics, logs, and traces used to understand relay or service behavior within an authorized administrative scope. The transport roles follow [MoQTransport]. * A *publisher* publishes a Track and sends its Objects to a subscriber. * A *subscriber* requests or accepts delivery of Objects from a Track. * A *relay* terminates MOQT sessions and forwards Objects between peers. It can act as a subscriber upstream and as a publisher downstream, and may cache Objects. This document uses the following operational roles. * A *relay operator* runs a relay or a fleet of relays. * A *service operator* operates a delivery service and uses authorized telemetry to assess delivery or investigate user reports. * A *tenant operator* administers delivery for a tenant and can see the observations authorized for that tenant. * A *telemetry collector* is a system that receives or retrieves metrics, logs, or traces for an operator. When it subscribes to a MOQT telemetry Track, it acts as a subscriber for that Track. An organization can fill several of these roles. For example, a service operator can run relays and publish content. Each role has its own scope of access under the deployment's policy. Bisht Expires 22 May 2027 [Page 4] Internet-Draft MOQT Relay Diagnostics November 2026 2. Problem Statement MOQT clients observe only their own sessions and cannot reliably determine why delivery stalls or fails across a relay chain. An authorized client needs a small, request-scoped account of what its relay observed and did, sufficient to guide waiting, retrying, or reporting a problem. That account can be incomplete and cannot establish end-to-end delivery or identify causes outside the relay's knowledge. Disclosure must protect other clients, topology, cache state, and operational policy, while keeping diagnostic processing and query rates bounded. 3. Use Cases The following use cases describe the question a client or operator wants to answer. They do not require a relay to disclose every fact that could help answer it. In each case, a relay may provide a coarse, request-scoped result, withhold details, or report that it cannot determine the cause. 3.1. Distinguishing a Local Request Outcome A subscriber issues a request for a track or a range of content and receives an error or observes no useful progress. The application needs to distinguish between a request that was rejected or could not be served at this relay and a request that was accepted but has not yet produced content. A useful answer might identify the outcome category and whether retrying the same request is likely to help. It need not expose the relay's policy rules, cache keys, or internal request identifiers. 3.2. Diagnosing a Delivery Stall A subscriber has an active subscription, but expected objects are not arriving within the application's tolerance. The subscriber or its support tooling wants to know whether this relay has observed relevant upstream content, has forwarded content downstream, or cannot determine where progress stopped. A bounded summary for that subscriber's request can guide whether to wait, retry, change tracks, or collect local evidence. It should not reveal other subscribers, shared cache contents, or the identities of upstream peers. Bisht Expires 22 May 2027 [Page 5] Internet-Draft MOQT Relay Diagnostics November 2026 3.3. Explaining a Relay Decision A relay may decline, defer, or limit work for a request because of its local authorization or resource policy. An authorized client wants an explanation that is actionable, such as whether the failure is likely to persist without a change in credentials or request parameters. A relay can communicate a coarse outcome without disclosing confidential policy thresholds, tenant configuration, load measurements, or details that would help an attacker tune requests around policy. 3.4. Supporting an Operator-Assisted Incident A subscriber reports a delivery problem to the service operator. A diagnostic summary for the subscriber's own request or a short-lived correlation value can help the service operator match the report to the relay's handling of that request. The value should be short- lived and refer only to the event, while the subscriber's access remains limited to its own request. More detailed logs may be available to the service operator through a separately authorized observability system. 3.5. Learning Whether Diagnostics Are Available A client may need to know whether a relay supports a diagnostic feature before requesting it. A capability indication can prevent unsupported requests, but it should reveal only the feature category needed for interoperability. It should not fingerprint a relay implementation or disclose deployment topology. 4. Illustrative Diagnostic Observations The following proposed metrics describe one client's existing, authorized request. Their names and semantics are starting points for discussion, not standardized MOQT fields, mandatory measurements, or a selected wire format. Counter and Gauge have the numeric meanings described in [MoQMetrics]. A relay would disclose a metric only when it can attribute the observation to that request using bounded local state and its disclosure policy permits access. Bisht Expires 22 May 2027 [Page 6] Internet-Draft MOQT Relay Diagnostics November 2026 +==================================+==============================+ | Suggested metric | Proposed meaning | +==================================+==============================+ | request_objects_observed_total | Cumulative number of | | | distinct, complete data | | | Objects accepted into this | | | request's forwarding path | | | after its filters and policy | | | checks. | +----------------------------------+------------------------------+ | request_objects_forwarded_total | Cumulative number of | | | distinct, complete data | | | Objects handed to the | | | outgoing transport for this | | | request. This local handoff | | | does not establish client | | | receipt, decoding, or | | | playback. | +----------------------------------+------------------------------+ | request_queued_objects | Complete, eligible Objects | | | still held by this relay | | | awaiting completion of | | | transport handoff for this | | | request, including any | | | retained unsent portion of | | | an Object. Excludes | | | transport-internal buffering | | | and shared cache inventory. | +----------------------------------+------------------------------+ | request_last_forward_age_seconds | Monotonic elapsed time since | | | the last complete-Object | | | handoff for this request, | | | rounded upward to a | | | disclosed precision when | | | policy requires. Omitted if | | | no such handoff is known. | +----------------------------------+------------------------------+ Table 1 4.1. Counting and Measurement Scope The two Counters start at zero when observation begins for the request and increase within one measurement epoch. A restart, reset, or change of measurement scope starts a new epoch, distinguished by a short-lived, session-local value. A 10-second count is the difference between two samples from the same epoch; a _total value is not a rolling 10-second count. Samples from different epochs cannot Bisht Expires 22 May 2027 [Page 7] Internet-Draft MOQT Relay Diagnostics November 2026 be subtracted. The Counters use independent acceptance and handoff boundaries. An Object accepted before an epoch or comparison window can complete handoff inside it, incrementing only the forwarding Counter for that interval. Their difference is therefore not a balance and can be negative over a comparison window. An Object is identified within a request by its Full Track Name, Group ID, and Object ID, and contributes at most once to each Counter per epoch. Incomplete Objects, status markers, duplicates, and transport retransmissions do not increment these Counters. Cached Objects can count when accepted into or handed off for this request; the report does not identify their source or expose shared cache activity. A relay omits a metric if bounded attribution or duplicate suppression is unavailable, including under subscription aggregation. The queue Gauge can decrease when Objects expire or are discarded, as well as when they are handed off. The Counter difference therefore does not necessarily equal queue depth and does not establish packet loss or a cause. The age Gauge uses a local monotonic clock and discloses no exact event timestamp. For example, with upward rounding in 5-second steps, a scalar value of 5 means an age greater than zero and at most 5 seconds; an interval such as [1, 5] is not a numeric Gauge value. 4.2. Example and Disclosure Limits This conceptual report samples an epoch 10 seconds after observation began. Its Counter values are totals since that epoch began; later window counts would require comparable earlier samples. The age value is rounded upward in 5-second steps. All field names, epoch and precision metadata, and JSON encoding below are illustrative and carry no protocol allocation or interoperability requirement. json { "scope": "existing_request", "measurement_epoch": "session- local-example", "measurement_elapsed_seconds": 10, "metrics": { "request_objects_observed_total": 120, "request_objects_forwarded_total": 117, "request_queued_objects": 2, "request_last_forward_age_seconds": 5 }, "age_rounding_seconds": 5, "progress": "forwarding_observed", "cause": "unknown", "coverage": "partial" } Unsupported, unavailable, withheld, or unattributable measurements are omitted, not replaced with zero. Zero represents a measured zero, such as no handoff counted in this epoch. For the age Gauge, no known handoff means omission, not zero. Any indication of why a value is absent is also subject to disclosure policy. Approximate Bisht Expires 22 May 2027 [Page 8] Internet-Draft MOQT Relay Diagnostics November 2026 request-specific bytes forwarded or a coarse progress category, such as waiting or unknown, can supplement the four metrics without exposing peer identities or asserting an upstream cause. Reports exclude cache-hit status, topology, relay-wide load, and other users' activity. Query rates, processing, response size, and retention need limits; producing a measurement must not initiate upstream retrieval or subscriptions. The working group would need to agree these semantics, precision metadata, authorization and request binding, and any mapping to [MoQMetrics] before implementations could claim interoperability. 5. Operator Diagnostics at Operational Scale Relay operators use operational measurements to assess capacity and behavior across the relays they run. Service operators use authorized observations to assess service health and correlate user reports with local delivery events. The metrics below illustrate those administrative views. A telemetry collector can use [MoQMetrics] data, [MoQLog] records, [QLOG] traces, or deployment observability data within its authorized collection scope. +=====================================+============================+ | Suggested operator observation | Operational use and | | | measurement scope | +=====================================+============================+ | relay_active_sessions | Current established MOQT | | | sessions at one relay; | | | helps identify connection | | | load and changes during | | | draining. | +-------------------------------------+----------------------------+ | relay_active_subscriptions | Current established | | | subscriptions at that | | | relay; counts local | | | subscription instances, | | | not unique viewers or end- | | | to-end subscriptions. | +-------------------------------------+----------------------------+ | relay_requests_total | Local request attempts | | | reaching initial | | | acceptance, rejection, | | | timeout, or cancellation, | | | counted once per attempt. | | | Bounded labels distinguish | | | request type, direction, | | | and outcome; each retry is | | | a separate attempt. | Bisht Expires 22 May 2027 [Page 9] Internet-Draft MOQT Relay Diagnostics November 2026 +-------------------------------------+----------------------------+ | relay_queued_payload_bytes | Unsent payload retained in | | | relay-managed forwarding | | | queues, counted per | | | pending downstream | | | delivery; measures | | | delivery backlog rather | | | than physical memory or | | | transport-internal | | | buffers. | +-------------------------------------+----------------------------+ | relay_delivery_timeout_resets_total | Locally initiated data- | | | stream terminations for | | | delivery timeout, counted | | | once per stream; helps | | | detect deadline pressure | | | without treating every | | | affected Object as a | | | separate reset. | +-------------------------------------+----------------------------+ | relay_cpu_utilization | Process CPU consumption | | | divided by available | | | process CPU capacity over | | | a disclosed sample | | | interval; the interval and | | | capacity basis accompany | | | the observation. | +-------------------------------------+----------------------------+ | relay_process_resident_memory_bytes | Resident memory of the | | | relay process, with | | | collection method | | | identified; useful for | | | capacity assessment | | | alongside queue and cache | | | observations. | +-------------------------------------+----------------------------+ | relay_cached_payload_bytes | Total payload length of | | | distinct cached Objects, | | | counted once per Full | | | Track Name, Group ID, and | | | Object ID; excludes | | | metadata and is not | | | physical process memory. | +-------------------------------------+----------------------------+ Table 2 Bisht Expires 22 May 2027 [Page 10] Internet-Draft MOQT Relay Diagnostics November 2026 For example, a relay operator can investigate scheduling or downstream capacity when queue bytes and delivery-timeout resets rise together. A rise in failed upstream request attempts, observed where the relay acts as a requester, can guide examination of that relay's upstream connections. High CPU or resident memory with growing load can motivate capacity planning or draining. These are correlations to investigate, not proof of a cause or of the quality experienced by every viewer. 5.1. Aggregation and Access Boundaries Every operator sample needs an administrative scope, collection interval or capture time, freshness, and coverage. Counters have local measurement epochs; rates or window deltas are calculated per relay before aggregation so restarts are not mistaken for traffic changes. Gauges are combined only across disjoint scopes with compatible meanings. Fleet reports disclose missing or stale contributors instead of presenting partial coverage as a complete service view. Fleet CPU fractions are weighted by the disclosed capacity basis rather than summed. Capacity thresholds and alert policies remain deployment decisions. Operator authorization is role- and scope-specific. A tenant operator sees only that tenant's authorized observations; node-wide CPU, shared-cache inventory, or another tenant's activity needs separate permission. Collector access, retention, and incident correlation need auditing and access controls. Labels use bounded categories such as request type and outcome, avoiding tokens, user identifiers, full content names, or arbitrary peer addresses. Detailed logs or traces are retrieved only for an authorized, bounded incident context, with sensitive fields minimized. These operational views are not returned through an ordinary client's private diagnostic view. 6. Requirements These requirements describe design goals for a future mechanism. They do not specify MOQT messages or require relays to expose operational information. Requirements to identify an individual request or report from one relay apply to the client-visible view. Operator aggregates instead identify their administrative scope, contributing relays, coverage, and freshness as described above; they retain the applicable access and resource limits. Bisht Expires 22 May 2027 [Page 11] Internet-Draft MOQT Relay Diagnostics November 2026 6.1. Diagnostic Scope * A mechanism needs to identify the request or delivery context to which a diagnostic applies, without requiring a stable identifier that can be used to correlate a client's activity across sessions. * A diagnostic needs to distinguish, where the relay has sufficient evidence, between a relay-observed outcome and an unknown or unobserved cause. * A mechanism needs to permit an implementation to return no diagnostic or a less detailed result when the relay lacks reliable evidence. * A diagnostic needs to avoid implying facts about upstream endpoints or network segments that the reporting relay cannot observe. 6.2. Privacy and Authorization * A relay needs to authenticate and authorize access to diagnostics according to its deployment's policy. A client authorized to request content is not automatically authorized to inspect general relay health or other clients' activity. * Diagnostic responses need to be scoped to the requesting client's own session and request, except where a separately defined and authorized operator interface is used. * Operator collection needs explicit administrative and tenant scope, role-based access, bounded retention, and auditability. Authorization for content delivery or a client diagnostic view does not grant operator access. * A mechanism needs to support data minimization: disclose the least information useful for the stated diagnostic purpose, and allow the relay to omit sensitive fields. * Diagnostics need not disclose relay topology, upstream identities, cache keys or contents, tenant names, policy rules, precise resource utilization, or information about other clients. * Any correlation value needs to be scoped, short-lived, and difficult to use as a cross-session or cross-service tracking identifier. Bisht Expires 22 May 2027 [Page 12] Internet-Draft MOQT Relay Diagnostics November 2026 * A client needs to be able to distinguish unsupported, unavailable, and withheld information where doing so does not itself create an unacceptable information leak. A mechanism can deliberately collapse these outcomes when policy requires it. 6.3. Safety and Operational Behavior * A mechanism needs bounded response size and bounded processing cost. * A relay needs discretion to rate-limit, defer, or refuse diagnostic requests without affecting ordinary content delivery. * Diagnostic requests need to avoid causing new upstream subscriptions, cache fills, or other expensive work solely to generate a diagnostic. * A diagnostic needs to be clearly identified as a point-in-time report from one relay. It may be incomplete, delayed, or based on an approximation. * Diagnostics need not be enabled by default. Capability discovery and use should be compatible with deployments that choose not to expose them. * Clients need to treat diagnostic information as advisory and continue to apply their own timeouts, retry limits, and security policy. 7. Candidate Mechanisms 7.1. Explicit Request and Response on the Control Path An explicit request lets a client ask for a diagnostic when a particular request is stalled or has failed. It can carry a request- scoped context and allow the relay to return a result only when asked. This avoids sending diagnostics for every request and gives the relay an opportunity to apply authorization and rate limits. The mechanism also adds a new request/response exchange and associated state. It needs rules for matching a diagnostic to a request, handling unavailable or withheld information, and preventing repeated queries from becoming a resource or probing attack. A diagnostic request must not trigger expensive upstream work solely to answer the query. Bisht Expires 22 May 2027 [Page 13] Internet-Draft MOQT Relay Diagnostics November 2026 7.2. Diagnostic Information in Existing Responses and Error Codes Adding a small diagnostic category to an existing response or error can make an outcome immediately actionable and avoid a separate round trip. It is a natural fit when the relay is already returning a result for the request. This approach is less suitable for later investigation of an ongoing stall, and risks turning response fields into a channel for detailed operational data. Error details may also be propagated or cached by relays. Any details would therefore need to remain compact, request- specific, privacy-preserving, and safe to forward or retain for the lifetime defined by the base protocol. 7.3. Capability Advertisement During Session Establishment Capability advertisement can tell a client whether a relay understands a diagnostic mechanism before the client uses it. It can support gradual deployment and avoid sending requests that cannot be understood. An advertisement says that a capability is supported; it does not authorize a particular client or guarantee that a relay will disclose a diagnostic for a particular request. Detailed capability lists can fingerprint implementations or deployments, so discovery should identify only the protocol feature needed for interoperability. The mechanisms are not necessarily exclusive. A later design could use capability discovery together with either an explicit query or compact response details. Selection requires working-group agreement on the target use cases and privacy model. 8. Relationship to Other Work This work concerns request-scoped explanations that can help an endpoint make a runtime decision or help a user report a problem. It is not a general metrics format, a relay management API, or a logging protocol. [MoQMetrics] provides a numeric data model, live subscriptions, and retrieval of cached measurements. Its Counter and Gauge types are candidates for the observations illustrated here. The proposed request semantics, attribution, and disclosure rules would need an agreed mapping, including naming and authorization, rather than a competing metrics transport. Bisht Expires 22 May 2027 [Page 14] Internet-Draft MOQT Relay Diagnostics November 2026 [MoQLog] provides structured operational logging over MOQT, while [QLOG] provides event representations useful for protocol analysis and incident correlation. Existing metrics, logs, and traces can support either audience when authorized. The operator examples here motivate reuse of those systems; they do not define a replacement collector, dashboard, or management API. The following are out of scope: exposing a relay's complete internal state; distributing network-wide or cross-tenant telemetry to ordinary clients; proving end-to-end delivery; standardizing operator dashboards; and defining a general-purpose query language for relay state. 9. Security Considerations Relay diagnostics create a deliberate information channel from a relay to a client. The channel can reveal operational state, deployment relationships, traffic patterns, and policy decisions. An attacker may use repeated or carefully varied requests to infer whether content is cached, whether another client is active, how a tenant is configured, or when a relay is under load. Even coarse fields can become identifying when combined with timing and client- side observations. A future protocol design should apply data minimization at the relay, not rely only on clients to protect received data. Diagnostics should be tied to the requesting client's authorization context and scoped to that client's own request. General relay health, cross- client statistics, and operator correlation should use separately authorized interfaces. Deployments may choose not to provide client- visible diagnostics or may provide only a small set of coarse outcomes. Operator observations need separately scoped authorization, tenant isolation, access auditing, and retention controls. Aggregation does not by itself make data anonymous, and small groups or repeated queries can still reveal an individual client's activity. Correlation with detailed logs or traces must remain within the operator's authorized incident scope. Implementations should avoid returning stable relay or client identifiers, precise timestamps, exact cache keys, upstream addresses, internal error text, or raw policy and resource measurements unless a separate, explicit trust model justifies them. Any event correlation value should be short-lived and non-reusable across sessions. A relay may deliberately make "unknown", "withheld", and "not available" indistinguishable if differentiating them would expose sensitive policy or state. Bisht Expires 22 May 2027 [Page 15] Internet-Draft MOQT Relay Diagnostics November 2026 Diagnostic queries can consume CPU, memory, or control-stream capacity, and may be abused to probe policy or amplify traffic. Responses and processing need limits; relays need rate limiting and discretion to refuse queries. Generating a diagnostic should not trigger content retrieval, cache filling, or upstream subscription work. Where diagnostic messages can be replayed, cached, or propagated by other relays, their scope and lifetime need to be defined so that one client's result is not disclosed to another. Clients should treat diagnostics as advisory, potentially incomplete observations. They should not infer that a named upstream component is at fault based solely on a relay response, and should not use the absence of a diagnostic as proof that delivery is healthy. Operators should consider the privacy of diagnostic data in logs, support tickets, and telemetry pipelines. 10. IANA Considerations This document has no IANA actions. 11. References 11.1. Normative References [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate Requirement Levels", BCP 14, RFC 2119, DOI 10.17487/RFC2119, March 1997, . [RFC8174] Leiba, B., "Ambiguity of Uppercase vs Lowercase in RFC 2119 Key Words", BCP 14, RFC 8174, DOI 10.17487/RFC8174, May 2017, . 11.2. Informative References [Issue1692] "moq-transport issue #1692: client-visible relay diagnostics", 2026, . [MoQLog] Jennings, C., "Logging over Media over QUIC Transport", 2026, . [MoQMetrics] Jennings, C., "Metrics over MOQT", 2026, . Bisht Expires 22 May 2027 [Page 16] Internet-Draft MOQT Relay Diagnostics November 2026 [MoQTransport] IETF MOQ WG, "Media over QUIC Transport", 2026, . [QLOG] IETF QUIC WG, "Main Logging Schema for qlog", 2026, . Acknowledgments The author thanks the participants in the Media over QUIC Transport working group for discussion of relay diagnostics and issue #1692. Additional acknowledgments will be added as the document develops. Author's Address Altanai Bisht Cisco Email: altanai@outlook.com Bisht Expires 22 May 2027 [Page 17]