Network Working Group A. K. Ravipati Internet-Draft RKR Networks Intended status: Informational 15 November 2026 Expires: 19 May 2027 The Synthetic-Twin HTTP Field draft-rkr-synthetic-twin-header-00 Abstract This document defines the Synthetic-Twin HTTP field. A sender adds the field to an HTTP request or response to say that the message was made by, or for, a synthetic twin and is not real user activity. A synthetic twin is a software model of a network estate, existing or planned. The field says which kind of synthetic message it is and can carry one opaque identifier. This lets operators keep synthetic messages apart from production messages in logs, analytics, billing, and alerting. Status of This Memo This Internet-Draft is submitted in full conformance with the provisions of BCP 78 and BCP 79. Internet-Drafts are working documents of the Internet Engineering Task Force (IETF). Note that other groups may also distribute working documents as Internet-Drafts. The list of current Internet- Drafts is at https://datatracker.ietf.org/drafts/current/. Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet-Drafts as reference material or to cite them other than as "work in progress." This Internet-Draft will expire on 19 May 2027. Copyright Notice Copyright (c) 2026 IETF Trust and the persons identified as the document authors. All rights reserved. Ravipati Expires 19 May 2027 [Page 1] Internet-Draft Synthetic-Twin Field November 2026 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 . . . . . . . . . . . . . . . . . . . . . . . . 2 1.1. Notational Conventions . . . . . . . . . . . . . . . . . 3 2. Terminology . . . . . . . . . . . . . . . . . . . . . . . . . 3 3. The Synthetic-Twin Field . . . . . . . . . . . . . . . . . . 4 3.1. Handling Unknown and Invalid Values . . . . . . . . . . . 4 4. Semantics . . . . . . . . . . . . . . . . . . . . . . . . . . 5 4.1. Meaning . . . . . . . . . . . . . . . . . . . . . . . . . 5 4.2. Senders . . . . . . . . . . . . . . . . . . . . . . . . . 5 4.3. Propagation . . . . . . . . . . . . . . . . . . . . . . . 5 4.4. Receivers . . . . . . . . . . . . . . . . . . . . . . . . 6 4.5. Responses and Caching . . . . . . . . . . . . . . . . . . 6 4.6. Intermediaries . . . . . . . . . . . . . . . . . . . . . 6 4.7. Relation to Other Fields . . . . . . . . . . . . . . . . 7 5. Examples . . . . . . . . . . . . . . . . . . . . . . . . . . 7 6. Security Considerations . . . . . . . . . . . . . . . . . . . 8 7. Privacy Considerations . . . . . . . . . . . . . . . . . . . 9 8. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 9 9. References . . . . . . . . . . . . . . . . . . . . . . . . . 9 9.1. Normative References . . . . . . . . . . . . . . . . . . 9 9.2. Informative References . . . . . . . . . . . . . . . . . 10 Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . 11 Author's Address . . . . . . . . . . . . . . . . . . . . . . . . 11 1. Introduction A synthetic twin is a software model of a network estate, existing or planned. Operators use a synthetic twin to try out changes and to run what-if tests before, or without, touching the real estate. This work produces HTTP messages that are not real user activity. Some of them stay inside the synthetic twin. Others are sent on purpose to real services. Today these messages are hard to tell apart from real traffic. As a result: * synthetic messages can distort analytics, billing, capacity data, and security alerts; Ravipati Expires 19 May 2027 [Page 2] Internet-Draft Synthetic-Twin Field November 2026 * a synthetic message that reaches a service it was not meant for is hard to spot; and * there is no common way to link a message back to the work that made it. Tools often solve this with private fields that differ from product to product, or with special User-Agent values. Neither works across products. This document defines one HTTP field, Synthetic-Twin, that carries a small, fixed set of facts: that the message is synthetic, which of two kinds it is, and an optional opaque identifier. It uses Structured Field Values [RFC9651], following the advice for new HTTP fields in [RFC9205]. The field is a label. It is not proof of anything, and it gives no extra rights to the sender (see Section 6). This document does not say when a sender may send synthetic messages to a service. That is a matter between the sender and the operator of the service. 1.1. Notational 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. This document uses the Structured Field types Item, Token, and String, and Parameters, from [RFC9651]. 2. Terminology Synthetic twin: A software model of a network estate, existing or planned. Modeled service: A service that exists only inside a synthetic twin. Real service: Any service that is not a modeled service. Synthetic message: An HTTP request or response made by, or for, a synthetic twin, and not by real user or production activity. Administrative domain: The set of networks and services run by one Ravipati Expires 19 May 2027 [Page 3] Internet-Draft Synthetic-Twin Field November 2026 operator, or by parties that have agreed to treat each other's synthetic messages as their own. 3. The Synthetic-Twin Field The Synthetic-Twin field is an Item Structured Field (Section 3.3 of [RFC9651]). It can appear in requests and in responses. Its value is a Token that gives the kind of synthetic message. This document defines two values: model: The message was made inside a synthetic twin and is addressed to a modeled service, or is a response from a modeled service. It is not meant to reach a real service. probe: The message is a synthetic request sent on purpose to a real service, or a response to such a request. It is not user activity. This document defines one Parameter: id: OPTIONAL. A String that identifies the work the message belongs to. The value is opaque: it has meaning only to the sender. It MUST be 1 to 64 characters long and MUST use only the characters a-z, A-Z, 0-9, "-", and "_". For example: Synthetic-Twin: probe; id="7f3a9c2104b8e6d1" Senders MUST NOT send more than one Synthetic-Twin field line in a message, and MUST NOT send the field in a trailer section (Section 6.5 of [RFC9110]). 3.1. Handling Unknown and Invalid Values Receivers MUST ignore Parameters they do not understand. This allows later documents to add Parameters. If the Token is a value the receiver does not understand, the receiver MUST still treat the message as a synthetic message, and MUST handle it as "probe". This gives one fallback: the message is synthetic, and it may be reaching a real service. If the id Parameter is not a String, or breaks the length or character rules above, the receiver MUST ignore the id Parameter and use the rest of the field. Ravipati Expires 19 May 2027 [Page 4] Internet-Draft Synthetic-Twin Field November 2026 If the field cannot be parsed as an Item, or the Item is not a Token, the receiver MUST ignore the field, as Section 4.2 of [RFC9651] requires for parse failures. This is also the result when a message has more than one Synthetic-Twin field line. A receiver MAY record that an invalid Synthetic-Twin field was present, but MUST NOT treat the message as a synthetic message for any other purpose. A receiver MUST ignore the field when it appears in a trailer section. 4. Semantics 4.1. Meaning The presence of a valid Synthetic-Twin field means: "the sender states that this message is a synthetic message of the given kind." It does not mean that the sender is trusted, allowed, or authenticated. The absence of the field means nothing. A message without the field is not proven to be production traffic. 4.2. Senders A synthetic twin, or a tool acting for it, SHOULD add the field to every synthetic HTTP request it sends: with the value "model" when the request is addressed to a modeled service, and with the value "probe" when it is addressed to a real service. A modeled service SHOULD add the field, with the value "model", to its responses. A sender that uses the id Parameter SHOULD use the same value for all messages that belong to the same piece of work, so that a receiver can group them. A user agent that follows [FETCH] treats this field as one that needs a preflight request when it is added to a cross-origin request. Tools that drive a browser need to allow for the extra request. 4.3. Propagation A service that makes further HTTP requests in order to handle a synthetic request SHOULD add the Synthetic-Twin field, unchanged, to those requests when they stay inside its administrative domain. Without this, the work done further along on behalf of a synthetic request looks like production activity. Ravipati Expires 19 May 2027 [Page 5] Internet-Draft Synthetic-Twin Field November 2026 A service MUST NOT add the field to a request that was not caused by a synthetic request. 4.4. Receivers A receiver MAY use the field to: * keep synthetic messages out of analytics, billing, and capacity data; * tag logs, traces, and flow records as synthetic, with the id; * send synthetic messages to a test path or a test data store; and * raise an alert when a "model" message reaches a real service. A real service that receives a "model" message SHOULD log it as misdirected. A receiver MUST NOT use the field alone to skip authentication, authorization, rate limits, input checks, or security monitoring. See Section 6. 4.5. Responses and Caching A real service that receives a request with a valid Synthetic-Twin field MAY add the same field, unchanged, to its response, to confirm that it saw the label. A response that carries the field, or that differs because the request carried the field, is not meant to be reused by a cache for a production request. A server that sends such a response MUST either: * mark it with the "no-store" response directive (Section 5.2.2.5 of [RFC9111]); or * list "Synthetic-Twin" in the Vary response field (Section 12.5.5 of [RFC9110]). When the id Parameter changes from one piece of work to the next, "no-store" is preferred, because Vary would keep a separate cache entry for each id value. 4.6. Intermediaries Proxies and gateways SHOULD forward the field unchanged inside an administrative domain. Ravipati Expires 19 May 2027 [Page 6] Internet-Draft Synthetic-Twin Field November 2026 A gateway that sends a message out of its administrative domain SHOULD remove the id Parameter, unless the operator has chosen to keep it. It MAY remove the whole field. A gateway that receives the field on a message from outside its administrative domain has no reason to believe it. The gateway SHOULD either remove the field or record that it arrived from outside, unless the sender has been authenticated as a known source of synthetic messages (see Section 6). 4.7. Relation to Other Fields The field does not replace trace context fields such as traceparent [TRACE-CONTEXT]. A message can carry both: traceparent links the message to a trace, and Synthetic-Twin says that the trace is synthetic. The baggage field [BAGGAGE] can carry any key and value chosen by an application, so two products rarely agree on one. Synthetic-Twin gives one registered name and one fixed meaning that proxies, log tools, and alert rules can match without knowing the application. The User-Agent field (Section 10.1.5 of [RFC9110]) names the software that made a request. It does not say whether the request is synthetic, and it is not sent on responses. 5. Examples A probe sent to a real service: GET /health HTTP/1.1 Host: app.example.net Synthetic-Twin: probe; id="7f3a9c2104b8e6d1" The real service confirms that it saw the label, and keeps the response out of caches: HTTP/1.1 200 OK Content-Type: application/json Cache-Control: no-store Synthetic-Twin: probe; id="7f3a9c2104b8e6d1" A response from a modeled service inside the synthetic twin: HTTP/1.1 200 OK Content-Type: application/json Synthetic-Twin: model; id="7f3a9c2104b8e6d1" Ravipati Expires 19 May 2027 [Page 7] Internet-Draft Synthetic-Twin Field November 2026 The smallest valid field: Synthetic-Twin: probe A field with a value this document does not define. The receiver treats the message as synthetic and handles it as "probe": Synthetic-Twin: other Two field lines. The combined value is not an Item, so the receiver ignores the field: Synthetic-Twin: probe Synthetic-Twin: model 6. Security Considerations The field is set by the sender and has no integrity protection of its own. Anyone can add it, remove it, or change it. This includes script in a web page, which can set the field on requests it makes (on cross-origin requests, after a preflight). Receivers need to think about four cases. Spoofing to hide: An attacker can add the field to real attack traffic so that it is dropped from alerts or analytics. For this reason a receiver MUST NOT relax security checks or monitoring because of the field alone. A receiver that wants to act on the field SHOULD first check that the message came from a known source of synthetic messages, for example by client certificate, network source, or an HTTP Message Signature [RFC9421] that covers the field. Spoofing to cause harm: An attacker can add the field to real user traffic so that it is dropped from billing, or sent to a test path. Receivers SHOULD only take such actions for authenticated sources of synthetic messages. Removal: An intermediary can remove the field, so synthetic messages look real. Operators that rely on the field SHOULD protect it, for example with TLS and with HTTP Message Signatures [RFC9421]. Unwanted load: Synthetic requests sent to a real service use its resources like any other request. The field does not change this. A real service applies its normal limits to labeled requests. The id Parameter is an identifier, not a credential. Senders MUST NOT put secrets, credentials, or tokens in it, and receivers MUST NOT treat knowledge of an id value as proof of anything. Ravipati Expires 19 May 2027 [Page 8] Internet-Draft Synthetic-Twin Field November 2026 Cached responses are covered in Section 4.5. 7. Privacy Considerations The field is designed to carry no personal data. Senders MUST NOT put personal data, user names, host names, or customer names in the id Parameter. The id SHOULD be a random or otherwise opaque value, not a descriptive name. An id that is reused across many messages links those messages together. This is the intended use inside an administrative domain. Outside it, the same link could help an observer follow a sender's activity. For this reason gateways remove the id Parameter from messages that leave the domain (see Section 4.6), in line with the advice on correlation in [RFC6973]. The presence of the field tells an observer that the sender uses a synthetic twin. Operators who do not want to reveal this outside their administrative domain can remove the whole field at the edge. 8. IANA Considerations IANA is requested to add the following entry to the "Hypertext Transfer Protocol (HTTP) Field Name Registry" at https://www.iana.org/assignments/http-fields (https://www.iana.org/assignments/http-fields), following the template in Section 16.3.1 of [RFC9110] and the Structured Type column defined in [RFC9651]: Field Name: Synthetic-Twin Status: provisional Structured Type: Item Reference: This document Comments: (none) This document does not create a new IANA registry. The values "model" and "probe" and the id Parameter are defined here only. New values and Parameters are added by documents that update this one. 9. References 9.1. Normative References Ravipati Expires 19 May 2027 [Page 9] Internet-Draft Synthetic-Twin Field November 2026 [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, . [RFC9110] Fielding, R., Ed., Nottingham, M., Ed., and J. Reschke, Ed., "HTTP Semantics", STD 97, RFC 9110, DOI 10.17487/RFC9110, June 2022, . [RFC9111] Fielding, R., Ed., Nottingham, M., Ed., and J. Reschke, Ed., "HTTP Caching", STD 98, RFC 9111, DOI 10.17487/RFC9111, June 2022, . [RFC9651] Nottingham, M. and P. Kamp, "Structured Field Values for HTTP", RFC 9651, DOI 10.17487/RFC9651, September 2024, . 9.2. Informative References [BAGGAGE] W3C, "Baggage", . [FETCH] WHATWG, "Fetch Living Standard", . [RFC6973] Cooper, A., Tschofenig, H., Aboba, B., Peterson, J., Morris, J., Hansen, M., and R. Smith, "Privacy Considerations for Internet Protocols", RFC 6973, DOI 10.17487/RFC6973, July 2013, . [RFC9205] Nottingham, M., "Building Protocols with HTTP", BCP 56, RFC 9205, DOI 10.17487/RFC9205, June 2022, . [RFC9421] Backman, A., Ed., Richer, J., Ed., and M. Sporny, "HTTP Message Signatures", RFC 9421, DOI 10.17487/RFC9421, February 2024, . [TRACE-CONTEXT] W3C, "Trace Context", W3C Recommendation, 23 November 2021, . Ravipati Expires 19 May 2027 [Page 10] Internet-Draft Synthetic-Twin Field November 2026 Acknowledgments The author thanks the reviewers at RKR Networks. Author's Address Ajay Kumar Ravipati RKR Networks Email: aj@rkr-networks.com Ravipati Expires 19 May 2027 [Page 11]