draft-ietf-netconf-configuration-tracing-00.xml

<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE rfc [
  <!ENTITY nbsp    "&#160;">
  <!ENTITY zwsp   "&#8203;">
  <!ENTITY nbhy   "&#8209;">
  <!ENTITY wj     "&#8288;">
]>
<?xml-model href="rfc7991bis.rnc"?>
<!-- <?xml-stylesheet type="text/xsl" href="rfc2629.xslt" ?> -->
<rfc
        xmlns:xi="http://www.w3.org/2001/XInclude"
        docName="draft-ietf-netconf-configuration-tracing-00"
        category="std"
        ipr="trust200902"
        obsoletes=""
        updates=""
        submissionType="IETF"
        xml:lang="en"
        tocInclude="true"
        sortRefs="true"
        symRefs="true"
        version="3" >
<front>
    <title abbrev="Configuration Tracing via trace-id">External Trace ID for Configuration Tracing</title>
    <seriesInfo name="Internet-Draft" value="draft-ietf-netconf-configuration-tracing-00"/>
    <author fullname="Jean Quilbeuf" initials="J" surname="Quilbeuf ">
      <organization>Huawei</organization>
      <address>
        <email>jean.quilbeuf@huawei.com</email>
      </address>
    </author> 
    <author fullname="Benoit Claise" initials="B" surname="Claise">
      <organization>Huawei</organization>
      <address>
        <email>benoit.claise@huawei.com</email>
      </address>
    </author>
    <author surname="Graf" initials="T" fullname="Thomas Graf">
      <organization>Swisscom</organization>
      <address>
        <postal>
          <street>Binzring 17</street>
          <city>Zurich</city>
          <code>8045</code>
          <country>Switzerland</country>
        </postal>
        <email>thomas.graf@swisscom.com</email>
      </address>
    </author>
    <author fullname="Diego R. Lopez" initials="D" surname="Lopez ">
      <organization>Telefonica I+D</organization>
      <address>
        <postal>
          <street>Don Ramon de la Cruz, 82</street>
          <city>Madrid  28006</city>
          <country>Spain</country>
        </postal>
        <email>diego.r.lopez@telefonica.com</email>
      </address>
    </author>
    <author fullname="Qiong Sun" initials="Q" surname="Sun">
      <organization>China Telecom</organization>
      <address>
        <email>sunqiong@chinatelecom.cn</email>
      </address>
    </author>
    <date year="2024" month="1" day="10"/>
    <area>General</area>
    <workgroup>OPSAWG</workgroup>
    <abstract>
      <t>
        Network equipment are often configured by a variety of network
        management systems (NMS), protocols, and teams.  If a network issue
        arises (e.g., because of a wrong configuration change), it is important
        to quickly identify the root cause and obtain the
        reason for pushing that modification.  Another potential network
        issue can stem from concurrent NMSes with overlapping intents, each
        having their own tasks to perform. In such a case, it is important to
        map the respective modifications to its originating NMS.
      </t>
      <t>
        This document specifies a NETCONF mechanism to automatically map the configuration
        modifications to their source, up to a specific NMS change request.  Such a mechanism is required,
        in particular, for autonomous networks to trace the source of a particular
        configuration change that led to an anomaly detection.  This mechanism facilitates the troubleshooting, the post mortem
        analysis, and in the end the closed loop automation required for
        self-healing networks.  The specification also includes a YANG module that is meant
        to map a local configuration change to the corresponding  trace id, up to the controller or even the orchestrator.
      </t>
    </abstract>
    <note removeInRFC="true">
      <name>Discussion Venues</name>
      <t>Source for this draft and an issue tracker can be found at
       <eref target="https://github.com/JeanQuilbeufHuawei/draft-quilbeuf-opsawg-configuration-tracing"/>.</t>
    </note>
  </front>
  <middle>
    <section anchor="intro" >
      <name>Introduction</name>
      <t>
        Issues arising in the network, for instance violation of some SLAs, might be due to some configuration modification.
        In the context of automated networks, the assurance system needs not only to identify and revert the problematic configuration modification,
        but also to make sure that it won't happen again and that the fix will not disrupt other services.
        To cover the last two points, it is imperative to understand the cause of the problematic configuration change.
        Indeed, the first point, making sure that the configuration modification will not be repeated, cannot be ensured if the cause for pushing the modification in the first place is not known.
        Ensuring the second point, not disrupting other services, requires as well knowing if the configuration modification was pushed in order to support new services.
        Therefore, we need to be able to trace a configuration modification on a device back to the reason that triggered that modification, for instance in a NMS, whether the controller or the orchestrator.
      </t>
      <t>
        This specification focuses only on configuration pushed via NETCONF <xref target="RFC6241"/> or RESTCONF <xref target="RFC8040"/>.
        The rationale for this choice is that NETCONF is better suited for normalization than other protocols (SNMP, CLI).
        Another reason is that the notion of trace context, useful to track configuration modifications, has been ported to NETCONF in <xref target="I-D.rogaglia-netconf-trace-ctx-extension"/> and RESTCONF in <xref target="I-D.rogaglia-netconf-restconf-trace-ctx-headers"/>.
      </t>
      <t>
        The same network element, or NETCONF <xref target="RFC6241"/> server, can be configured by different NMSs or NETCONF clients.
        If an issue arises, one of the starting points for investigation is the configuration modification on the devices supporting the impacted service.
        In the best case, there is a dedicated user for each client and the timestamp of the modification allows tracing the problematic modification to its cause.
        In the worst case, everything is done by the same user and some more correlations must be done to trace the problematic modification to its source.
      </t>
      <t>
        This document specifies a mechanism to automatically map the configuration modifications to their source, up to a specific NMS service request.
        Practically, this mechanism annotates configuration changes on the configured element with sufficient information to unambiguously identify the corresponding transaction, if any,
        on the element that requested the configuration modification.
        It reuses the concept of Trace Context <xref target="W3C-Trace-Context"/> applied to NETCONF as in <xref target="I-D.ietf-netconf-transaction-id"/>
        The information needed to trace the configuration is stored in a new YANG module that maps a local configuration change to some additional metadata.
        The additional metadata contains the trace ID, and, if the local change is not the beginning of the trace, the ID of the client that triggered the local-change.
      </t>
    </section>
    <section anchor="terminology">
      <name>Terminology</name>
      <t>
        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 <xref target="RFC2119"/> <xref target="RFC8174"/>
        when, and only when, they appear in all capitals, as shown here.
      </t>
      <t>
        This document uses the terms client and server from <xref target="RFC6241"/>.
      </t>
      <t>
        This document uses the terms transaction and Transaction ID from <xref target="I-D.ietf-netconf-transaction-id"/>.
      </t>
      <t>
        This document uses the term trace ID from <xref target="W3C-Trace-Context"/>.
      </t>
      <dl>
        <dt>Local Commit ID</dt>
        <dd>
          Identifier of a local configuration change on a Network Equipment, Controller, Orchestrator or any other device or software handling configuration.
          Such an identifier is usually present in devices that can show an history of the configuration changes, to identify one such configuration change.
        </dd>
      </dl>
    </section>
    <section anchor="use-cases">
      <name>Use cases</name>
      <t>
        This document was written with autonomous networks in mind.
        We assume that an existing monitoring or assurance system, such as described in <xref target="RFC9417"/>, is able to detect and report network anomalies ,
        e.g. SLA violations, intent violations, network failure, or simply a customer issue.
        Here are the use cases for the proposed YANG module.
      </t>
        <section anchor="configuration-mistakes">
          <name>Configuration Mistakes</name>
          <t>
            Taking into account that many network anomalies are due to configuration mistakes,
            this mechanism allows to find out whether the offending configuration modification was triggered by a tracing-enabled client/NMS.
            In such a case, we can map the offending configuration modification id on a server/NE to a local configuration modification id on the client/NMS.
            Assuming that this mechanism (the YANG module) is implemented on the controller, we can recursively find, in the orchestrator,
            the latest (set of of) service request(s) that triggered the configuration modification.
            Whether this/those service request(s) are actually the root cause needs to be investigated.
            However, they are a good starting point for troubleshooting, post mortem analysis, and in the end the closed loop automation, which is absolutely required for  for self-healing networks.
          </t>
        </section>
        <section anchor="concurrent-nms-configuration">
          <name>Concurrent NMS Configuration</name>
          <t>
            Building on the previous use case is the situation where two NMS's, unaware of the each other, are configuring a common router,
            each believing that they are the only NMS for the common router.
            So one configuration executed by the NMS1 is overwritten by the NMS2, which in turn is overwritten by NMS1, etc.
          </t>
        </section>
        <section anchor="conflicting-intents">
          <name>Conflicting Intents</name>
          <t>
            Autonomous networks will be solved first by assuring intent per specific domain; for example data center, core, cloud, etc.
            This last use case is a more specific "Concurrent NMS configuration" use case where assuring domain intent breaks the entire end to end service, even if the domain-specific controllers are aware of each other.
          </t>
        </section>
        <section anchor="not-use-case">
          <name>Not a use case: Onboarding</name>
          <t>
            During onboarding, a newly added device is likely to receive a multiple configuration message, as it needs to be fully configured.
            Our use cases focus more on what happens after the initial configuration is done, i.e. when the "stable" configuration is modified.
          </t>
        </section>
      </section>
    <section anchor="transaction-id-tracing">
      <name>Relying on W3C Trace Context to Trace Configuration Modifications</name>
      <section>
        <name>Existing configuration metadata on device</name>
        <t>
          This document assumes that NETCONF clients or servers (orchestrators, controllers, devices, ...) have some kind of mechanism to record the modifications done to the configuration.
          For instance, devices typically have an history of configuration changes and this history associates a locally unique identifier to some metadata, such as the timestamp of the modification, the user doing the modification or the protocol used for the modification.
          Such a locally unique identifier is a Local Commit ID, we assume that it exists on the platform.
          This Local Commit ID is the link between the module presented in this draft and the device-specific way of storing configuration changes.
        </t>
      </section>
      <section>
        <name>Client ID</name>
        <t>
            This document assumes that each NETCONF client for which configuration must be traced (for instance orchestrator and controllers)
               has a unique client ID among the other NETCONF clients in the network.
            Such an ID could be an IP address or a host name.
            The mechanism for providing and defining this client ID is out of scope of the current document.
        </t>
      </section>
      <section>
        <name>Instantiating the YANG module</name>
        <t>
            <xref target="I-D.rogaglia-netconf-trace-ctx-extension"/> defines a NETCONF extension providing the trace context from <xref target="W3C-Trace-Context"/>.
            Using this mechanism, the NETCONF server captures the trace-id, when available, and maps it to a local commit ID, by populating the YANG module.
        </t>
        <figure anchor="hierarchical-configuration" >
          <name>Example of Hierarchical Configuration. tx: transaction. tr: trace.</name>
          <artwork><![CDATA[
                         +---------------+
                         | Orchestrator  |
                         +---------------+
                                | tr-1, tx-1
                                v
                         +---------------+
                         |   Controller  |
                         +---------------+
                tr-1, tx-2 |           | tr-1, tx-3
                           v           v
                        +-----+     +-----+
                        | NE1 |     | NE2 |
                        +-----+     +-----+
            ]]>
          </artwork>
        </figure>
        <!--
        <t>
          A server considers as a Parent Transaction a transaction that modifies its configuration.
          On <xref target="hierarchical-configuration"/>, tx-1 is a Parent Transaction for the Controller.
        </t>
        <t>
          A client considers as a Child Transaction the modification of a server configuration.
          On <xref target="hierarchical-configuration"/>, tx-2 and tx-3 are Child Transactions for the Controller.
        </t>
        <t>
          If the set-tx-id feature is enabled (see open issue in <xref target="set-tx-id"/>), the client can specify its own transaction ID when sending the configuration ID for the server.
          In that case, the Controller in <xref target="hierarchical-configuration"/> could use the same transaction-id for both tx-2 and tx-3 and save a single Child Transaction ID for that commit.
          Otherwise, the server is the one generating the ID for the transaction between the client and the server.
          If the client has to configure several servers, for instance to enable a network service, then each of the configured servers might return a different ID.
          Therefore, for a configuration modification on the client might be implemented via several Child Transactions and thus might have several Child Transaction IDs.
        </t>
        <t>
          Our proposed solution is to store, on the server, a mapping between the existing local commit id and the Parent and Child Transactions related to that local configuration change.
          The mapping is read only and populated by the server at configuration time as follows:
        </t>
        <ul>
          <li>
            Parent Transaction:
            If the set-tx-id feature is available (see <xref target="set-tx-id"/>), the server MUST accept a transaction-ID and a client ID from client supporting configuration tracing.
            The server MUST store both entries as respectively Parent Transaction ID and Client ID, associated to the local configuration ID.
            If the set-tx-id feature is not available, the server MUST accept the client ID, generate a transaction ID, save both the transaction ID as Parent Transaction ID and the Client ID as Client ID, and send back the transaction ID to the client.
            If the client does not support configuration tracing, none of these entries are populated.
            In <xref target="hierarchical-configuration"/>, for the Controller, the Parent Transaction ID is the ID of tx-1.
          </li>
          <li>
            Child Transaction:
            If the set-tx-id feature is available (see <xref target="set-tx-id"/>), when a client has to configure servers in response to a local configuration change, then it MUST generate a Transaction ID, send it along with its ID to the configured servers, and save it as a Child Transaction ID.
            If the set-tx-id feature is not available, it MUST sent its own ID with the configuration, receive back the transaction ID from each server, and save all of them as Child Transaction ID.
            In <xref target="hierarchical-configuration"/>, for the Controller, the Child Transaction IDs are the IDs of tx-2 and tx-3.
          </li>
        </ul>
        <t>
          The two cases above are not mutually exclusive. A Controller can be configured by an Orchestrator and configure network equipment in turn, as shown in <xref target="hierarchical-configuration"/>.
          In that case, both the Parent Transaction ID, shared with the Orchestrator and the Child Transaction IDs, shared with the network equipments, are stored in the Controller.
          They are both associated to the corresponding configuration commit in the Controller.
        </t>  -->
        <t>
          It is technically possible that several clients push configuration to the candidate configuration datastore  and only one of them commits the changes to the running configuration datastore.
          From the running configuration datastore perspective, which is the effective one, there is a single modification, but caused by several clients, which means that this modification should have several corresponding client-ids.
          Although, this case is technically possible, it is a bad practice. We won’t cover it in this document.
          In other terms, we assume that a given configuration modification on a server is caused by a single client, and thus has a single corresponding client-id.
        </t>
      </section>
      <section anchor="module-usage">
        <name>Using the YANG module</name>
        <t>
          The YANG module defined below enables tracing a configuration change in a Network Equipment back to its origin, for instance a service request in an orchestrator.
          To do so, the Anomaly Detection System (ADS) should have, for each client-id, access to some credentials enabling read access to the YANG module for configuration tracing on that client.
          It should as well have access to the network equipment in which an issue is detected.
        </t>
        <figure anchor="configuration-tracing" >
          <name>Example of Configuration Tracing. tr: trace-id, C: Controller, O: orchestrator. The number between square brackets refer to steps in the listing below.</name>
          <artwork><![CDATA[
                                                  +---------------+
     .----------------[5]match tr-1-------------->|               |
     |                                            | Orchestrator  |
     | ----------------[6]commit-id---------------|               |
     | |                                          +---------------+
     | |                                                | tx-1
     | |                                                v
     | |                                          +---------------+
     | |   .-----------[3] match tr-1------------>|               |
     | |   |                                      |   Controller  |
     | |   | .-----------[4] c-id O tr-1----------|               |
     | |   | |                                    +---------------+
     | |   | |                                          | tx-2
     | v   | v                                          v
   +-----------+                                      +----+
   | Anomaly   |--[1] match commit-id before time t-->|    |
   | Detection |                                      | NE |
   | System    |<--------- [2] c-id C tr-1------------|    |
   +----------+                                       +----+
      ]]>
          </artwork>
        </figure>
        <t>
          The steps for a software to trace a configuration modification in a Network Equipment back to a service request are illustrated in <xref target="configuration-tracing"/>.
          They are detailed below.
        </t>
        <ol>
          <li>
            The Anomaly Detection System (ADS) identifies the commit id that created an issue, for instance by looking for the last commit-id occurring before the issue was detected.
            The ADS queries the NE for the trace id and client id associated to the commit-id.
          </li>
          <li>
            The ADS receives the trace-id and the client-id. In <xref target="configuration-tracing"/>, that step would receive the trace-id tr-1 and the id of the Controller as a result.
            If there is no associated client-id, the change was not done by a client compatible with the present draft, and the investigation stops here.
          </li>
          <li>
            The ADS queries the client identified by the client-id found at the previous step, looking for a match of the trace-id from the previous step.
            In <xref target="configuration-tracing"/>, for that step, the software would look for the trace-id tr-1 stored in the Controller.
          </li>
          <li>
            From that query, the ADS knows the local-commit-id on the client (Controller in our case).
            Since the local-commit-id is associated to a client-id pointing to the Orchestrator, the ADS continues the investigation.
          </li>
          <li>
            The ADS queries the Orchestrator, trying to find a match for the trace-id tr-1.
          </li>
          <li>
            Finally, the ADS receives the commit-id from the Orchestrator that ultimately caused the issue in the NE.
            Since there is no associated client-id, the investigation stops here.
            The modification associated to the commit-id, for instance a service request, is now available for further manual or automated analysis, such as analyzing the root cause of the issue.
          </li>
        </ol>
        <t>
          Note that step 5 and 6 are actually a repetition of step 3 and 4.
          The general algorithm is to continue looking for a client until no more client-id can be found in the current element.
        </t>
      </section>
    </section>
    <section anchor="yang-module">
      <name>YANG module</name>
      <t> We present in this section the YANG module for modelling the information about the configuration modifications. </t>
      <section anchor="module-tree">
        <name>Overview</name>
        <t> The tree representation <xref target="RFC8340"/> of our YANG module is depicted in <xref target="module-tree-fig"/> </t>
        <figure anchor="module-tree-fig" >
          <name>Tree representation of ietf-external-transaction-id YANG module</name>
          <artwork><![CDATA[
module: ietf-external-transaction-id
  +--ro external-transactions-id
     +--ro configuration-change* [local-commit-id]
        +--ro local-commit-id    string
        +--ro timestamp?         yang:date-and-time
        +--ro trace-parent
        |  +--ro version?       hex-digits
        |  +--ro trace-id?      hex-digits
        |  +--ro parent-id?     hex-digits
        |  +--ro trace-flags?   hex-digits
        +--ro client-id?         string
            ]]>
          </artwork>
        </figure>
        <t> 
          The local-commit-id represents the local id of the configuration changes, which is device-specific.
          It can be used to retrieve the local configuration changes that happened during that transaction.
        </t>
        <t>
          The trace-parent is present to identify the trace associated to the local-commit-id.
          This trace-parent can be transmitted by a client or created by the current server.
          In <xref target="module-usage"/>, the most important field in trace-parent is the trace-id.
          We also included the other fields for trace-parent as defined in <xref target="W3C-Trace-Context"/> for the sake of completion.
          In some cases, for instance direct configuration of the device, the device may choose to not include the trace-id.
        </t>
        <t>
          The presence of a client-id indicates that the trace-parent has been transmitted by that client.
          If the trace is initiated by the current server, there is no associated client-id.
        </t>
        <t>
          Even if this document focuses only on NETCONF or RESTCONF, the use cases defined in <xref target="use-cases"/> are not specific to NETCONF or RESTCONF and the mechanism described in this document could be adapted to other configuration mechanisms.
          For instance, a configuration modification pushed via CLI can be identified via a label, which could contain the trace-parent.
          As such cases are difficult to standardize, we won’t cover them in this document.
        </t>
      </section>
      
      <section anchor="module-code" >
        <name>YANG module ietf-external-transaction-id</name>
        <sourcecode name="ietf-external-transaction-id@2021-11-03.yang" type="yang" markers="true" ><![CDATA[
module ietf-external-transaction-id {
  yang-version 1.1;
  namespace
    "urn:ietf:params:xml:ns:yang:ietf-external-transaction-id";
  prefix ext-txid;

  import ietf-yang-types {
    prefix yang;
    reference
      "RFC 6991: Common YANG Data Types, Section 3";
  }

  organization
    "IETF NETCONF Working Group";
  contact
    "WG Web:   <https://datatracker.ietf.org/wg/netconf/>
     WG List:  <mailto:netconf@ietf.org>
     Author:   Benoit Claise  <mailto:benoit.claise@huawei.com>
     Author:   Jean Quilbeuf  <mailto:jean.quilbeuf@huawei.com>";
  description
    "This module enables tracing of configuration changes in a
     network for the sake of automated correlation between
     configuration changes and the external request that triggered
     that change.

     The module stores the identifier of the trace, if any, that
     triggered the change in a device. If that trace-id was provided
     by a client, (i.e. not created locally by the server), the id
     of that client is stored as well to indicated which client
     triggered the configuration change.

     Copyright (c) 2022 IETF Trust and the persons identified as
     authors of the code.  All rights reserved.

     Redistribution and use in source and binary forms, with or
     without modification, is permitted pursuant to, and subject
     to the license terms contained in, the Revised BSD License
     set forth in Section 4.c of the IETF Trust's Legal Provisions
     Relating to IETF Documents
     (https://trustee.ietf.org/license-info).
     This version of this YANG module is part of RFC XXXX; see the
     RFC itself for full legal notices.  ";

  revision 2022-10-20 {
    description
      "Initial revision";
    reference
      "RFC xxxx: Title to be completed";
  }

  typedef hex-digits {
    type string {
      pattern '[0-9a-f]*';
    }
    description
      "A string composed of hexadecimal digits. Digits represented by
       letters are restricted to lowercase so that a single
       representation of a given value is allowed. This enables using
       the string equality to check equality of the represented
       values.";
  }

  grouping trace-parent-g {
    description
      "Trace parent frow the W3C trace-context recommandation.
       Follows the format version 00.";
    leaf version {
      type hex-digits {
        length "2";
      }
      must "../version = '00'";
      description
        "Version of the trace context. Must be 00 to match the
         format described in this module.";
    }
    leaf trace-id {
      type hex-digits {
        length "32";
      }
      must "../trace-id != '00000000000000000000000000000000'";
      description
        "Trace ID that is common for every transaction that is
         part of the configuration chain. This value can be used
         to match a local commit id to a commit local to another
         system.";
    }
    leaf parent-id {
      type hex-digits {
        length "16";
      }
      description
        "ID of the request (client-side) that lead to configuring
         the server hosting this module.";
    }
    leaf trace-flags {
      type hex-digits {
        length "2";
      }
      description
        "Flags enabled for this trace. See W3C reference for the
         details about flags.";
    }
  }

  container external-transactions-id {
    config false;
    description
      "Contains the IDs of configuration transactions that are
       external to the device.";
    list configuration-change {
      key "local-commit-id";
      description
        "List of configuration changes, identified by their
         local-commit-id";
      leaf local-commit-id {
        type string;
        description
          "Stores the identifier as saved by the server. Can be used
           to retrieve the corresponding changes using the server
           mechanism if available.";
      }
      leaf timestamp {
        type yang:date-and-time;
        description
          "A timestamp that can be used to further filter change
           events.";
      }
      container trace-parent {
        description
          "Trace parent associated to the local-commit-id. If a
           client ID is present as well, the trace context was
           transmitted by that client. If not, the trace context was
           created locally.

           This trace-parent must come from the trace context of the
           request actually modifying the running configuration
           datastore. This request might be an edit-config or a
           commit depending on whether the candidate datastore is
           used.";
        uses trace-parent-g;
      }
      leaf client-id {
        type string;
        description
          "ID of the client that originated the modification, to
           further query information about the corresponding
           change.

           This data node is present only when the configuration was
           pushed by a compatible system.";
      }
    }
  }
}
    ]]></sourcecode>
      </section>
    </section>

    <section anchor="security">
        <name>Security Considerations</name>
    </section>
    <section anchor="iana">
      <name>IANA Considerations</name>
      <t>
        This document includes no request to IANA.
      </t>
    </section>
    <section>
      <name>Contributors</name>
    </section>
    <section>
      <name>Open Issues / TODO</name>
      <ul>
        <li>
          Indicate what to do with O-RAN apps, since each of them might be seen as a different client with a different client-id.
          This is actually a requirement that the client-id should be granular enough to distinguish between different controllers colocated on the same device.
          For instance, the IP address might not be a suitable client-id in that case.
        </li>
        <li>
          Define how to pass the client-id. Current leads are the trace-state from <xref target="W3C-Trace-Context"/> and <eref target="https://www.w3.org/TR/baggage/">W3C Baggage</eref>.
        </li>
        <li>
          The model and usage presented here focuses of the problem of tracing a configuration change back to its sources.
          As it relies on <xref target="W3C-Trace-Context"/>, we could also use associated mechanisms for collecting and representing trace data such as OTLP.
          For instance, we could define a YANG model matching the OTLP protobuffer definition (draft: <eref target="https://github.com/rgaglian/ietf-netconf-trace-context-extension/blob/main/ietf-netconf-otlp-protocol.tree"/>).
          In that case the client-id could be a specific attribute of the spans list.
        </li>
      </ul>
    </section>
  </middle>
  <back>
    <references title="Normative References">
      <?rfc include="http://xml.resource.org/public/rfc/bibxml/reference.RFC.2119.xml"?>
      <?rfc include="http://xml.resource.org/public/rfc/bibxml/reference.RFC.6241.xml"?>
      <?rfc include="http://xml.resource.org/public/rfc/bibxml/reference.RFC.8174.xml"?>
      <?rfc include="http://xml.resource.org/public/rfc/bibxml/reference.RFC.8340.xml"?>
      <?rfc include="http://xml.resource.org/public/rfc/bibxml3/reference.I-D.draft-ietf-netconf-transaction-id-01.xml"?>
      <?rfc include="http://xml.resource.org/public/rfc/bibxml3/reference.I-D.draft-rogaglia-netconf-trace-ctx-extension-03.xml"?>
      <?rfc include="http://xml.resource.org/public/rfc/bibxml3/reference.I-D.draft-rogaglia-netconf-restconf-trace-ctx-headers-00.xml"?>
      <reference anchor="W3C-Trace-Context" target="https://www.w3.org/TR/2021/REC-trace-context-1-20211123/">
        <front>
          <title>W3C Recommendation on Trace Context</title>
          <author >
            <organization></organization>
          </author>
          <date year="2021" month="November" day="23"/>
        </front>
</reference>
    </references>
    <references title="Informative References">
      <?rfc include="http://xml.resource.org/public/rfc/bibxml/reference.RFC.8040.xml"?>
      <?rfc include="http://xml.resource.org/public/rfc/bibxml/reference.RFC.9417.xml"?>
    </references>
    <?rfc needLines="100"?>
    <section>
      <name>Changes between revisions</name>
      <t>
        01 -> 02
      </t>
      <ul>
        <li> Switch to trace-parent instead of transaction id for tracing configuration</li>
      </ul>
      <t>
        00 -> 01
      </t>
      <ul>
        <li> Define Parent and Child Transaction</li>
        <li> Context for the "local-commit-id" concept</li>
        <li> Feedback from Med, both in text and YANG module</li>
      </ul>
    </section>
    <section anchor="usage">
      <name>Tracing configuration changes</name>
    </section>
    <section numbered="false">
      <name>Acknowledgements</name>
        <t>
           The authors would like to thank Mohamed Boucadair, Jan Linblad and Roque Gagliano for their reviews and propositions.
        </t>
    </section>
  </back>
</rfc>
<!-- Local Variables: -->
<!-- fill-column:72 -->
<!-- End: -->