]> Qwilt

6, Ha'harash Hod HaSharon 4524079 Israel nir@apache.org

Qwilt

6, Ha'harash Hod HaSharon 4524079 Israel ori.finkelman.ietf@gmail.com

Verizon

13100 Columbia Pike Silver Spring MD 20904 USA sanjay.mishra@verizon.com

Qwilt

275 Shoreline Dr Ste 510 Redwood City CA 94065 USA jayrobertson@acm.org

2you.io

1295 Beacon Street Unit 249 Brookline MA 02446 USA alan.arolovitch@gmail.com

This document obsoletes RFC8007. The document describes the part of Content Delivery Network Interconnection (CDNI) Control interface that allows a CDN to trigger activity in an interconnected CDN that is configured to deliver content on its behalf. The upstream CDN MAY use this mechanism to request that the downstream CDN preposition, invalidate, and/or purge metadata and/or content. The upstream CDN MAY monitor the status of activity that it has triggered in the downstream CDN.

introduces the problem scope for Content Delivery Network Interconnection (CDNI) and lists the four categories of interfaces that may be used to compose a CDNI solution (Control, Metadata, Request Routing, and Logging). expands on the information provided in and describes each of the interfaces and the relationships between them in more detail. The CDNI Control Interface / Triggers 1st edition , deprecated by this document, describes the "CI/T" interface -- "CDNI Control Interface / Triggers". It does not consider those parts of the Control interface that relate to the configuration, bootstrapping, or authentication of CDN Interconnect interfaces. Section 4 of identifies the requirements specific to the CI/T interface; requirements applicable to the CI/T interface are CI-1 to CI-6. This document is a second edition of the CDNI Control Interface / Triggers, which defines a new version, "v2", of the interface objects. The new version aims to support REST architectural style in a way that improves the interface's flexibility, extensibility, and interoperability, and allows encoding of the interface using OpenAPI . The new objects replace the main CI/T objects as follows: The "ci-trigger-command" object and its matching "ci-trigger-status" object are replaced with the "ci-trigger.v2" object representing a trigger resource The "ci-trigger-collection" object is replaced with the "ci-trigger-collection.v2" that is expanded to support filtering by trigger state and trigger labels The second edition of the CI/T interface further allows the use of separate Control interface endpoints for content and metadata. The document also provides a trigger extension mechanism that MAY be used to provide further instruction on the trigger execution. This second edition also includes cascaded CDN error propagation and extended trigger status reporting for improved trigger execution monitoring, as well as the use of external object lists for improved scale and integration of trigger-based APIs with existing content workflows. outlines the model for the CI/T interface at a high level. defines the CI/T interface offered by the downstream CDN. defines the encoding of the standard CI/T objects and introduces trigger spec and trigger extension types. describes the FCI capabilities objects used to inform on the supported CI/T-related capabilities. contains example messages.

This document reuses the terminology defined in and uses "uCDN" and "dCDN" as shorthand for "upstream CDN" and "downstream CDN", respectively. This document also introduces additional terminology extending , defined as follows: Node: A Node is a device or function participating in content delivery within a Content Delivery Network (CDN). A Node extends the Surrogate defined in , by additionally allowing requests originating from other Nodes, performed to satisfy requests ultimately initiated by User Agents. A Node therefore includes cache systems operating at any position within a CDN delivery topology, regardless of whether they directly receive requests from User Agents. Terminal State: A trigger state indicating that processing of the trigger has concluded and no further execution progress is expected. Terminal States are "complete", "processed", "failed", and "cancelled". Transit CDN (tCDN): A CDN that receives a CI/T trigger from an upstream CDN and forwards (redistributes) it to one or more downstream CDNs as part of a cascaded CDN deployment. The keywords "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 when, and only when, they appear in all capitals, as shown here.

The CI/T interface utilizes the HTTP/1.1 protocol and follows the principles of the Representational State Transfer (REST) architectural style. The uCDN, in its capacity as a CI/T interface client, requests the dCDN to carry out an action ("trigger") related to metadata and/or content stored by the dCDN on behalf of the uCDN. The dCDN, as a CI/T interface server, governs the triggers as a set of resources, which can be dynamically created and deleted, and whose state can be retrieved and/or modified by the uCDN. Each such trigger is identified by a unique Uniform Resource Identifier (URI) as defined in Section 4.2 of . Once a trigger is created, the uCDN can retrieve its representation from the dCDN or request the trigger to be modified by transferring an updated representation of it to the dCDN. The CI/T interface supports the representation of trigger resources using JSON . This RESTful data model, built around a common "trigger" resource, replaces the command-oriented model of , wherein the uCDN passed commands to the dCDN using "ci-trigger-command" objects, and the dCDN generated "ci-trigger-status" objects in response.

Section 9.3 of defines the set of methods in HTTP. The CI/T interface uses some of these methods for resource creation, retrieval of resource state, modification of resources, and deletion of resources. The HTTP methods not listed here are not supported by the CI/T interface. GET - retrieves a representation of the target resource. GET is a "safe" method: it is not intended to cause state changes on the server. POST - requests that the target resource process the representation enclosed in the request. When processing results in the creation of a new resource, the server responds with 201 ("Created") and a Location header field containing the URI of the newly created resource. DELETE - requests removal of the target resource. HEAD - identical to GET, except that the server MUST NOT return a message body. HEAD is used to inspect response metadata (e.g., to verify resource existence, or to check Last-Modified, ETag, or Content-Length headers) without transferring the representation.

The uCDN requests creation of a trigger resource to instruct the dCDN to perform an action. If the dCDN accepts the request, it creates a new trigger resource and returns its unique URI to the uCDN. The uCDN MUST use this URI for all requests associated with the created trigger resource. Note that the version of the trigger resources that the uCDN requests to create MUST match the version of CI/T trigger objects reported as supported by the dCDN. The CI/T interface supports the following types of trigger action: preposition - used to instruct the dCDN to fetch metadata from the uCDN or content from any origin, including the uCDN. invalidate - used to instruct the dCDN to revalidate specific metadata and/or content before its next use. purge - used to instruct the dCDN to delete specific metadata and/or content. Note that additional action types can be defined and registered in the future. The trigger resource has a "state" attribute. The dCDN creates new triggers in the "pending" state. Once the dCDN starts processing a pending trigger, the trigger state is set to "active". The uCDN MAY explicitly request the trigger to be created in the "active" state. If accepted by the dCDN, it MAY create the new trigger in the "active" state and start its processing immediately upon creation. Once the trigger processing is complete, the state is set to either "complete" or "failed", depending on the processing outcome. The uCDN MAY request the cancellation of a trigger from the dCDN. If such a request is accepted, the trigger state is changed to "cancelling", and when the cancellation is complete, the trigger state changes to "cancelled". For a full description of the trigger resource, please refer to .

The dCDN MUST only allow the uCDN access to the trigger resources it created. The dCDN MUST ensure that triggers created in response to a uCDN request apply only to content and metadata objects associated with that uCDN. In case of content prepositioning, the dCDN MUST be able to associate content objects referenced in a trigger created by the uCDN with delivery CDNI metadata objects in its possession that are associated with the same uCDN. These CDNI metadata objects include HostIndex, HostMatch, HostMetadata, PathMatch, PatternMatch, and PathMetadata, as described in Section 3.1 of . For example, if the dCDN has no MI metadata objects that enable the dCDN to respond to requests for video.example.com, it MUST NOT allow prepositioning of content objects with this hostname in the object URL. In case of metadata prepositioning, the prepositioned metadata objects MUST be consistent with pre-existing metadata, e.g., prepositioning of the MI HostMatch object in the absence of MI HostIndex would be rejected. If such association between a trigger and pre-existing delivery metadata cannot be established, the dCDN MUST reject it by responding with 400 ("Bad Request") HTTP status code. Furthermore, the dCDN SHOULD similarly reject a trigger from the uCDN A that seeks to preposition delivery metadata objects that are in conflict with the pre-existing metadata objects belonging to another uCDN B, e.g., if it could cause the dCDN to match a content request with metadata objects from multiple uCDNs.

The Trigger Index is the top-level resource that references all trigger resources belonging to a particular uCDN. The dCDN MUST maintain one Trigger Index resource for each uCDN, and MUST enforce that each uCDN has access only to its own Trigger Index. The Trigger Index contains references to Trigger Collection resources. Each collection contains triggers that are optionally filtered by parameters - for example, all triggers in the "pending" state or all triggers labeled "video". The same trigger resource can be present in more than one collection. The supported trigger collection representations are listed in and include filtering of triggers by state and label. The dCDN selects which filtered collections it publishes, as described in . Note that additional trigger collection representations can be defined in the future.

Figure 1 is an example showing the basic message flow in a CI/T interface session used by the uCDN to trigger activity in the dCDN and for the uCDN to discover the status of that activity. Only successful triggering is shown. Please note that the example below uses simplified trigger identifiers for brevity. It is RECOMMENDED that the actual implementations use unique UUID identifiers as specified in . Examples of the messages are shown in .

[ ]--+ | [ ] | (2) | (3) HTTP 201 Response [ ]<-+ [ ] <-------------------------------------------------- [ ] | Loc: https://dcdn.example/cit/uCDN/123 | | | . . . . . . . . . | | | (4) GET https://dcdn.example/cit/uCDN/123 | [ ] --------------------------------------------------> [ ] | [ ] | (5) HTTP 200 Trigger resource representation [ ] [ ] <-------------------------------------------------- [ ] | | | | | (6) DELETE https://dcdn.example/cit/uCDN/123 | [ ] --------------------------------------------------> [ ]--+ | [ ] | (7) | (8) HTTP 204 No Content [ ]<-- [ ] <-------------------------------------------------- [ ] | | ]]>

The steps in Figure 1 are as follows: The uCDN requests creation of a new trigger resource by POSTing its representation to the trigger index resource with a well-known URI "https://dcdn.example/cit/uCDN". The dCDN authenticates the request, validates the trigger resource in it, and if the request is accepted, creates a new trigger resource. The dCDN responds to the uCDN with an HTTP 201 ("Created") response status and the location of the trigger resource. The uCDN MAY query, possibly repeatedly, the trigger resource in the dCDN. The dCDN responds to each query with the current trigger resource representation, including the trigger state that reflects the progress of the uCDN request. Once the trigger reaches a terminal state, the uCDN MAY request deletion of the trigger resource. The dCDN validates the request and the trigger resource state. If successful, the trigger resource is removed by the server, and subsequent requests for this resource MUST result in 404 ("Not Found"). The dCDN responds to the deletion request with a 204 ("No Content") status code. This section provides an overview of a regular session. For detailed discussions of trigger modification, cancellation, and deletion, see , , and .

The uCDN MAY place limits on the timing and order of execution of a trigger through optional TimePolicy and/or ExecutionPolicy extensions. If neither of these extensions are present in the trigger resource, the timing and order of the trigger execution is under the dCDN's control, including the start time, pacing of the activity in the network, and order in which the dCDN chooses to process pending triggers. The CI/T "invalidate" and "purge" trigger actions MUST be applied to all data acquired before the dCDN begins the trigger processing (i.e., enters "active" state). The dCDN implementation SHOULD apply "invalidate" and "purge" triggers to content acquisition that is in progress when the trigger becomes active, to avoid placing purged or invalidated content into the cache upon completion of the content acquisition. The dCDN SHOULD NOT apply CI/T "invalidate" and "purge" actions to data acquired after the trigger processing started, but this may not always be achievable, so the uCDN SHOULD NOT rely on it. If the uCDN wishes to invalidate or purge content and then immediately preposition replacement content at the same URLs, it SHOULD ensure that the dCDN has processed the invalidate/purge before initiating the prepositioning. Otherwise, there is a risk that the dCDN prepositions the new content, then immediately invalidates or purges it (as a result of the two uCDN requests running in parallel). The uCDN MAY use the Execution Policy extension to condition the start of preposition trigger processing on completion of the earlier invalidate/purge trigger(s).

Each trigger can operate on multiple metadata and/or content elements. These elements are targeted by specifying both their subject (i.e., "metadata" or "content") as well as specification method (e.g., URL Regexes) and value. Multiple representations of an HTTP resource may share the same URL. Triggers that invalidate or purge metadata and/or content apply to all resource representations with matching URLs.

Possible trigger states are defined in . Trigger state MUST NOT be reported as "complete" until all operations listed in the trigger have been completed successfully. In case of CDN cascading, the completion of operations includes processing of the trigger in downstream CDNs. For detailed discussion of the cascading use case, see . The reasons for failure, and URLs or patterns affected, SHOULD be made available in the trigger state representation. For more details about error handling, see .

The CDNI Control Interface / Triggers 1st edition defines a set of properties and objects used by the trigger commands. This 2nd edition defines an extension mechanism to the triggers interface that enables applications to add instructions for finer control over the trigger execution, for example indicating a time window in which to execute the trigger. This document specifies a generic trigger extension object wrapper for managing CDNI trigger extensions in a uniform manner. All trigger extensions are OPTIONAL, and it is thus the responsibility of the extension specification to define a consistent default behavior for the case the extension is not present. All trigger extensions MUST have their type registered in the IANA "CDNI CI/T Trigger Extension Types" registry (see ). This document also defines an initial set of trigger extension types and registers them in the IANA "CDNI CI/T Trigger Extension Types" registry: JSON string Description location-policy Allowing the control over the locations in which the trigger is executed. time-policy Allowing the scheduling of a trigger to run in a specific time window. execution-policy Allowing the control over the order and timing in which triggers are executed.

In a network of interconnected CDNs, a uCDN distributes a trigger referencing metadata and/or content items to one or more dCDNs. When a dCDN further distributes the trigger to downstream dCDNs, it acts in the capacity of a Transit CDN (tCDN). The dCDN that creates trigger resources at the request of a transit CDN MUST associate the triggers with the transit CDN from which it receives the request, regardless of where the trigger request may have originated. If the dCDN is also acting as the uCDN in a cascade, it MUST forward trigger requests to any dCDNs that may be affected. The trigger state MUST NOT be reported as "complete" by a transit CDN until it is "complete" in all of its dCDNs and in the transit CDN itself. If a trigger is reported as "processed" in the transit CDN or any one of its dCDNs, transit CDNs MUST report the trigger as "processed" as well. If a trigger is reported as "failed" by the transit CDN or any one of its dCDNs, the transit CDN MUST report the trigger as "failed" only after its processing is finished in it and all of its dCDNs. A cancelled trigger MUST be reported as "cancelling" until it has been reported as "cancelled", "complete", or "failed" by all cascaded dCDNs. Security considerations are discussed further in .

A "diamond" configuration is one where the dCDN can potentially acquire metadata and content originated in one uCDN from that uCDN itself and a transit CDN, or via more than one transit CDN. The "diamond" configuration may cause configuration consistency problems, where the dCDN may end up in possession of multiple, potentially conflicting metadata objects belonging to the multiple uCDNs that match the same content request. The conflict may arise due to the differences in trigger processing by the transit CDNs and/or variances in trigger propagation delay across different paths in the "diamond" topology. Because of this, the "diamond" configuration is considered a configuration error. A dCDN that receives identical trigger creation requests from different uCDNs SHOULD reject duplicate trigger requests, as described in .

Given three CDNs, A, B, and C, if CDNs B and C delegate delivery of CDN A's content to each other, CDN A's trigger creation requests could be passed between CDNs B and C in a loop. More complex networks of CDNs could contain similar loops involving more hops. When such CDN topologies become possible, it is RECOMMENDED that CDNs participating in it utilize a CDN Provider ID (PID) to detect and prevent loops as follows: The uCDNs that originate a new trigger request SHOULD specify their CDN PID using the trigger "cdn-path" attribute (see for details). A dCDN that receives a trigger creation request that contains a "cdn-path" attribute SHOULD check it for its own CDN PID. If the dCDN's PID is already present, and the dCDN is not the CDN initiating the trigger, this condition likely indicates a loop. In such case, the dCDN MUST reject the trigger with status code 400 ("Bad Request"), which would result in a trigger rejection being returned to the originating uCDN. If the dCDN receives a trigger that it itself originated, the dCDN MAY process the trigger as required. A dCDN that acts as a transit CDN by cascading trigger requests to additional dCDNs SHOULD NOT reject triggers that contain CDN PIDs matching any of its downstream CDNs in their CDN path, allowing each CDN to do its own loop detection. Transit CDNs MUST append their CDN PID to the CDN path of a trigger before sending it to its downstream CDNs. The dCDNs SHOULD advertise their CDN PID to the uCDNs using the "cdn-id" attribute of the trigger index resource.

This section describes an interface to enable the uCDN to trigger activity in the dCDN. The CI/T interface builds on top of HTTP, so the dCDNs MAY make use of any HTTP feature when implementing the CI/T Interface. For example, the dCDN SHOULD make use of HTTP's caching mechanisms to reduce the uCDN's trigger status polling overhead by indicating the modification status of a requested resource representation. The dCDNs MAY implement separate CI/T interfaces per , i.e., one CI/T interface for trigger operations on metadata and another for operations on content. In this case, the dCDN MUST advertise separate interface endpoints via . All dCDNs implementing CI/T MUST support the HTTP GET, HEAD, POST, and DELETE methods as defined in . The only resource representation specified in this document is JSON . It MUST be supported by both the uCDN and the dCDN. The CI/T interface uses a root URI for the retrieval of the trigger index resource and creation of new triggers. The mechanism for discovery of that URL is part of the CI/T interface bootstrapping and is outside the scope of this document. The uCDN requests to create a new trigger resource by POSTing its representation to the trigger index resource URI, discovered at the time of interface bootstrapping, e.g., "https://dcdn.example/cit/ucdn/triggers". If the request is accepted by the dCDN, it creates a new trigger resource and returns its URI to the uCDN in an HTTP 201 ("Created") response. Once created, the new trigger URI also becomes available via the trigger collection resources described in . Additionally, the uCDN MAY discover the URIs of multiple trigger collection representations by retrieving the trigger index resource, which is accessible at the interface root URI. This means that the URIs for all trigger resources and trigger collection representations can be discovered by the uCDN through the top-level trigger index resource, allowing dCDNs to use any URI structure they choose for CI/T resources. Therefore, uCDNs MUST NOT make any assumptions regarding the structure of CI/T URIs or the mapping between CI/T objects and their associated URIs. The URIs used in the examples in this document are purely illustrative and are not intended to impose a definitive structure on CI/T interface implementations.

To create a new trigger, the uCDN makes an HTTP POST request with the trigger representation to the trigger index resource URI. The trigger representation MUST include the mandatory attributes of the trigger resource. The uCDN MAY also specify trigger v2 specification attributes, namely trigger labels and trigger extensions, as well as the "cdn-path" attribute of the trigger resource. The dCDN MUST validate the trigger resource representation sent by the uCDN. If the representation is malformed or the uCDN does not have sufficient access rights, the dCDN MUST respond with an appropriate 4xx HTTP error code. The new trigger resource is created in a "pending" state. If successful, The HTTP response to the uCDN trigger creation request MUST have status code 201 ("Created") and MUST convey the URI of the newly created trigger resources in the Location response header field . The HTTP response SHOULD include the updated representation of the trigger resource. This is particularly important in cases where the dCDN processed the trigger immediately. Once a trigger resource has been created, the dCDN MUST NOT reuse its URI, even after the trigger resource has been fully removed. It is therefore RECOMMENDED that the dCDN utilize unique UUID identifiers as specified in . The dCDN SHOULD respond with updated trigger resource representations to the subsequent uCDN requests sent to the created trigger URI. If the dCDN is unable to do that, it MUST indicate that it has accepted the request but will not be providing further status updates, setting the trigger state to "processed" at creation time. In this case, the dCDN SHOULD continue processing as if it were a request in the "complete" state. It is RECOMMENDED that the dCDN provide an estimate of trigger completion ("etime") when the trigger state is set to "processed". The uCDN MAY request the new trigger to be created in the "active" state so that its processing begins immediately. If agreed to by the dCDN, the dCDN MUST start the new trigger processing immediately. Otherwise, the dCDN MUST set the state of the new trigger to "pending". Once trigger processing has started, the status MUST be changed to "active". Finally, once the trigger processing is complete, the trigger state MUST be set to "complete" or "failed". Once created, trigger resources can be cancelled, modified, or deleted by the uCDN, subject to the constraints described below.

Modification of existing triggers is useful for the uCDN to correct an error in trigger specification or trigger extension(s) that may govern when the trigger is to be processed. The uCDN can request modification of an existing trigger resource by sending an updated trigger representation to the trigger URI using HTTP POST command. The POST method is used here to mutate an existing resource; no new resource is created. The dCDN MAY accept modifications of the trigger specifications, trigger extensions and trigger labels, when the trigger is in a "pending" state, i.e., the dCDN hasn't started its processing yet. The dCDN MAY also accept a request to change the trigger state subject to the following constraints: the requested state is "cancelled", and the trigger was in either "pending" or "active" state when the dCDN receives the request the requested state is "active", and the trigger was in a "pending" state when dCDN received the request describes the processing of trigger cancellation requests in detail. The uCDN MAY request to set the trigger state to "active" to prompt the dCDN to re-examine the trigger resource and start its processing immediately. The dCDN MUST respond to the trigger modification request appropriately. Thus, the HTTP status code 200 ("OK") SHOULD be returned if the modification has been processed, 202 ("Accepted") if the command has been accepted but the modification is not fully complete yet, 404 ("Not Found") when the trigger resource does not exist, 409 ("Conflict") when the trigger resource is in a state that doesn't allow the requested modification, 501 ("Not Implemented") if the modification is not supported by the dCDN or an appropriate 4xx HTTP error code in case of a malformed request. In case of successful 2xx response, the dCDN MUST provide the updated trigger resource representation in the response body.

The uCDN MAY request cancellation of a trigger by requesting its state to be set to "cancelled", as described in . The dCDN MUST respond to such requests, however, the actual cancellation of a trigger resource is OPTIONAL to implement. The dCDN MUST respond to the trigger cancellation request with an appropriate HTTP response status code as documented in If cancellation of a "pending" trigger is accepted by the dCDN, the dCDN SHOULD NOT start the processing of that activity. However, a uCDN's observation of the "pending" state does not guarantee that the trigger will still be "pending" when the cancellation request is processed by the dCDN; consequently, processing may still occur. If cancellation of an "active" or "processed" trigger is accepted by the dCDN, the dCDN SHOULD stop processing the trigger. However, as with the cancellation of a "pending" trigger, the dCDN does not guarantee that the trigger processing doesn't run to completion in the meantime. If the dCDN cannot stop the trigger processing immediately after receiving the request from the uCDN to do so, it MUST set the trigger state to "cancelling" and provide this state in the trigger representation in its response. If the trigger processing is stopped before its normal completion, the trigger state MUST be set to "cancelled". Cancellation of a "complete", "failed" or "cancelled" trigger requires no processing in the dCDN. Its state MUST NOT be changed.

The uCDN has two ways to check the progress of its triggers' processing, as described in and . To enable the uCDN to use client-side caching of the trigger index resource, as well as all trigger and trigger collection resources, each resource representation sent by the dCDN SHOULD include at least one of the following HTTP response headers: "ETag" or "Last-Modified". The dCDN SHOULD respond with the HTTP 304 ("Not Modified") status code and no response body for conditional resource requests using the 'If-None-Match' and/or 'If-Modified-Since' headers, as specified in Section 13 of , if it does not have a more recent resource representation. The dCDN SHOULD also use the cache control headers for responses to GET requests for its resources to indicate the frequency at which it recommends that the uCDN and/or intermediate proxies should poll for change. If provided, the uCDN should match the frequency of polling to the cache control information provided by the dCDN.

The uCDN MAY fetch the Trigger Collection that contains all of its triggers, or one of the collections that filter triggers based on a parameter such as state or label. This makes it possible for the uCDN to poll the status of all trigger resources or selected trigger subsets. The set of filtered trigger collections is chosen by the dCDN, and the uCDN cannot request collections with other filters. If the uCDN requires a subset of triggers for which the dCDN publishes no matching collection, the uCDN retrieves the narrowest published collection containing that subset, or the unfiltered collection, and selects the relevant triggers itself using the "state" and "labels" attributes of each trigger.

The dCDN provides the uCDN with a trigger resource URI at creation time. Alternatively, the uCDN MAY discover the URI by retrieving the trigger index resource and the appropriate trigger collection referenced therein. The uCDN MAY obtain an up-to-date representation of the trigger resource at any time using an HTTP GET request, including the current trigger state.

If the dCDN advertises support for extended status, the uCDN MAY request an extended representation of trigger resources and trigger collections. The extended representation provides additional information beyond the default resource representation. When applied to a trigger collection resource, the extended representation embeds full representations of trigger resources within the collection. This allows retrieval of complete trigger resource representations for a specific trigger state (e.g., all triggers in a "pending" state). When applied to an individual trigger resource, the extended representation includes the list of objects derived by the dCDN during trigger processing, in addition to the default trigger resource attributes. The uCDN MAY request the extended representation by passing the query string parameter "status=extended" when requesting either a trigger collection resource or a trigger resource. The dCDN SHOULD respond with HTTP status 200 ("OK") when such a request can be satisfied, 501 ("Not Implemented") if the capability has not been implemented or advertised, and 400 ("Bad Request") for a malformed query. By default, entries in a trigger collection represent trigger resources using only their resource URI, and individual trigger resources do not include derived object lists unless the extended representation is requested.

The uCDN MAY request the deletion of trigger resources at any time using the HTTP DELETE method, as defined in the CDNI Control Interface / Triggers 1st edition , Once deleted, the deleted trigger MUST be removed from all trigger collections. Subsequent requests to the trigger resource URI MUST be rejected by the dCDN with HTTP error 404 ("Not Found"). The effect of deletion is similar to cancellation, except that the trigger resource becomes unavailable after the deletion is complete. For this reason, the uCDN SHOULD cancel triggers rather than delete them when it needs to access the trigger status after processing has terminated. If deletion of a "pending" trigger is accepted by the dCDN, the dCDN SHOULD NOT start processing that activity. However, a uCDN's observation of the "pending" state does not guarantee that the trigger will still be "pending" when the deletion request is processed by the dCDN; consequently, processing may still occur. When an "active" or "processed" trigger is deleted, the dCDN SHOULD stop processing it. However, as with the deletion of a "pending" trigger, the dCDN does not guarantee this. Deletion of a "complete", "cancelled", "cancelling", or "failed" trigger MUST result in no further processing by the dCDN. The dCDN MUST respond to the trigger deletion request appropriately. The dCDN MUST respond with status code 204 ("No Content") without a response body if the trigger has been deleted immediately, 202 ("Accepted") if the command has been accepted but the trigger has not yet been deleted, 404 ("Not Found") if the trigger resource does not exist, or 501 ("Not Implemented") if deletion is not supported by the dCDN. The trigger state MUST be set to "cancelling" while the dCDN is processing a deletion request asynchronously.

The dCDN MAY automatically delete trigger resources sometime after they reach a terminal state. In this case, after the dCDN has removed such a trigger, it MUST respond to subsequent requests for it with the HTTP error 404 ("Not Found") and remove it from all trigger collections. A dCDN SHOULD NOT expire triggers in the "processed" state while they can reasonably be expected to remain active, such as when execution or redistribution to other dCDNs may still be ongoing. If the dCDN does remove triggers in a terminal state automatically, it MUST report the expiry timeout period, using an attribute "staleresourcetime" of the trigger index resource (see for details. It is RECOMMENDED that the dCDN sets the value of the "staleresourcetime" attribute to at least 24 hours. It is further RECOMMENDED that the uCDN sets its trigger polling period to less than this period, so it doesn't miss trigger status updates before the "complete" or "failed" triggers are expired by the dCDN.

The dCDN SHOULD reject invalid CI/T interface requests with a 4xx or 5xx HTTP status code. For example, the dCDN MAY respond with 400 ("Bad Request") if the request is malformed, or 403 ("Forbidden") or 404 ("Not Found") if the request could not be properly authenticated or if the uCDN is trying to act on another CDN's resources. The HTTP status codes used by this interface are summarized in . The response body for a rejected request SHOULD contain an Error.v2 Description object identifying the reason for the rejection. Providing the full Error.v2 Description allows a transit CDN that receives the rejection when propagating a trigger to relay the error upstream without loss of information (see ). A rejected request does not create a trigger resource. The response body is sent with the MIME media type "application/cdni" and the payload type "ci-trigger-error.v2" (). The dCDN SHOULD include the response body irrespective of the "Accept" header field of the rejected request. Once a request has been accepted, processing errors are reported within the trigger as a list of Error.v2 Descriptions. Each Error.v2 Description reports errors against one or more of the URLs or patterns in the trigger specification. If any part of the trigger processing fails, the trigger SHOULD be reported as "failed" once its activity completes, or once no further errors will be reported. The "errors" property in the trigger enumerates which actions failed and why. It MAY be present while the trigger is still "pending" or "active" if processing is still running for some of the URLs or patterns in the trigger specification. If the dCDN experiences an internal outage or is unable to propagate a trigger to one or more cascaded dCDNs, it SHOULD report an error for the affected trigger. The temporary unavailability of individual nodes within the dCDN is an internal operational condition. The dCDN SHOULD handle this internally, and such unavailability SHOULD NOT, by itself, affect the reported state of the trigger. The dCDN SHOULD ensure that nodes returning to service are brought into a state consistent with previously executed trigger operations before those nodes resume normal operation.

This section summarizes the HTTP status codes referenced by the CI/T interface. Unless otherwise specified, the semantics of HTTP status codes are defined in . The dCDN uses HTTP status codes to indicate the outcome of requests performed by the uCDN. The table below provides a summary of the status codes used by this specification and their typical usage within the CI/T interface. HTTP Status Codes Used by the CI/T Interface

Code	Reason Phrase	Usage in this Specification
200	OK	Returned when a request has been successfully processed.
201	Created	Returned when a trigger resource is successfully created. The Location header conveys the URI of the new trigger resource.
202	Accepted	Returned when a request has been accepted for processing but the operation has not yet completed.
204	No Content	Returned for successful DELETE requests when no response body is provided.
304	Not Modified	Returned for conditional GET requests when the requested resource representation has not changed.
400	Bad Request	Returned when a request is malformed or otherwise invalid.
403	Forbidden	Returned when the requester is not authorized to act on the specified resource.
404	Not Found	Returned when the referenced trigger resource does not exist or has been removed.
409	Conflict	Returned when the trigger resource state does not allow the requested operation.
501	Not Implemented	Returned when the requested operation is not supported by the dCDN.

The 501 status code indicates a lack of support for the operation itself, for example, deleting or modifying a trigger, or returning an extended representation of a trigger or trigger collection. A trigger whose type, spec, subject or extension the dCDN does not support is rejected with a 400 status code.

CDNI triggers may be propagated over a chain of downstream CDNs. For example, an upstream CDN A ("uCDN-A") delegates to a downstream CDN B ("dCDN-B"), and dCDN-B in turn delegates to a downstream CDN C ("dCDN-C"). Triggers sent from uCDN-A to dCDN-B may be redistributed from dCDN-B to dCDN-C, and errors can occur anywhere along the path. It is therefore desirable for uCDN-A, which originated the trigger, to be able to trace an error back to the downstream CDN where it occurred. This document adds a mechanism to convey the PID of the dCDN at which a fault occurred back to the uCDN by including that PID in the Error.v2 Description. Errors arising within a cascaded CDN may be either synchronous request rejections or asynchronous processing failures. A downstream CDN that accepts a trigger creation request with a 2xx status code reports any subsequent processing failures asynchronously, through Error.v2 Descriptions in the trigger status resource. However, when such a CDN propagates the trigger further downstream, the next CDN in the chain may reject the propagated request with a 4xx or 5xx HTTP status code. Because the upstream request has already been accepted, this rejection cannot be reported upstream as an HTTP status code. Instead, the CDN that received the rejection MUST translate it into an Error.v2 Description and report it in its trigger status resource. Accordingly, when dCDN-B propagates a trigger to dCDN-C, it MUST propagate back the errors associated with that trigger, whether dCDN-C rejected the propagated request synchronously or reported processing failures asynchronously, by adding the corresponding Error.v2 Descriptions to the "errors" array in the trigger status resource it returns to uCDN-A. When doing so, dCDN-B MAY include dCDN-C's PID in each propagated Error.v2 Description, indicating the CDN at which the error originated. uCDN-A then receives an "errors" array containing the errors that occurred across all downstream CDNs along the execution path, where each error MAY identify the CDN at which it occurred. below is an example showing the message flow used by uCDN-A to trigger activity in dCDN-B, followed by dCDN-C, as well as the discovery of the status of that activity, including the Error Propagation.

[ ]--+ | | [ ] | (2) | | [ ]<-+ | | (3) HTTP 201 Response. [ ] | |<----------------------------[ ] | | Loc: [ ] | | https://dcdn-b.com [ ] (4) POST | | /cit/ucdn-a/123 [ ] https://dcdn-c.com | | [ ] /cit/dcdn-b | (5) | [ ]--------------------------->[ ]--+ | | [ ] | | | [ ]<-+ | | (6) HTTP 400 Response. [ ] | [ ]<---------------------------[ ] | | | | [ ]--+ | | [ ] | (7) | | [ ]<-+ | . . . . . . . . . | (8) GET | | | https://dcdn-b.com | | | /cit/ucdn-a/123 | | [ ]--------------------------->[ ] | | [ ] | | (9) HTTP 200 [ ] | | Trigger resource [ ] | [ ]<---------------------------[ ] | ]]>

The steps in are as follows: uCDN-A creates a trigger in dCDN-B by POSTing a new trigger representation to "https://dcdn-b.com/cit/ucdn-a". dCDN-B authenticates the request, validates the trigger creation request, and, if it accepts the request, creates a new trigger resource. dCDN-B responds to uCDN-A with an HTTP 201 ("Created") response status and the location of the newly created trigger. dCDN-B creates a trigger in dCDN-C by POSTing the newly received trigger representation to "https://dcdn-c.com/cit/dcdn-b". dCDN-C authenticates the request and validates the trigger creation request. dCDN-C does not support an element of the request (for example, the trigger definition contains an "action" type that dCDN-C does not support). dCDN-C rejects the request with a 4xx HTTP status code (for example, 400 "Bad Request"). The response body contains an Error.v2 Description with the error code "eunsupported" and the offending portion of the trigger specification. No trigger resource is created at dCDN-C. dCDN-B receives the rejection from dCDN-C and translates it into an Error.v2 Description, setting the state of its own trigger resource to "failed". uCDN-A queries, possibly repeatedly, the trigger resource in dCDN-B. dCDN-B responds with the updated trigger resource representation. The trigger state is "failed", and the Error.v2 Description carries the error code "eunsupported" received from dCDN-C, with dCDN-C's PID identifying where the error originated. Including dCDN-C's PID is OPTIONAL; dCDN-B MAY instead use its own PID, thereby obfuscating the identity of dCDN-C from the upstream CDN.

The Trigger, Trigger Index, and Trigger Collection resources and their respective properties are encoded in JSON format. When sending the JSON-based representation of these resources, the MIME media type "application/cdni" MUST be used, with parameter "ptype" values as defined below and in . Names in JSON are case-sensitive. The names and literal values specified in the present document MUST always use lowercase. JSON types, including "object", "array", "number", and "string", are defined in . Unrecognized attributes in JSON objects SHOULD NOT be treated as an error by either the uCDN or the dCDN. They SHOULD be ignored during processing and passed on by the dCDN to any further dCDNs in a cascade.

Trigger resource is encoded as a JSON object and MUST use a MIME media type of "application/cdni; ptype=ci-trigger.v2". Please note that the dCDN MUST include all existing trigger attributes in the trigger resource representation when requested by the uCDN. The trigger resource contains the following attributes: Description: Defines the type of the CI/T trigger action. Value: Trigger action type, as defined in . Mandatory-to-Specify: It is optional for trigger update requests sent by the uCDN, otherwise mandatory. Description: Array of trigger specs representing the trigger's targets, as described in . Value: Array of GenericTriggerSpec objects (see ). Mandatory-to-Specify: It is optional for trigger update requests sent by the uCDN, otherwise mandatory. Furthermore, when mandatory, the list MUST NOT be empty. Description: Array of trigger extensions, as described in . Value: Array of GenericTriggerExtension objects (see ). Mandatory-to-Specify: No. The default is no extensions. Description: Trigger labels, as described in . Value: A JSON object, whose members are the trigger labels. Each member name is a label key and each member value is the corresponding label value; both are JSON strings. A label key and a label value MUST each be no more than 63 characters in length, MUST begin with a letter or a number, and MAY contain letters, numbers, hyphens, dots, and underscores. Mandatory-to-Specify: No. The default is no labels. Description: The chain of CDN PIDs of CDNs that have already created this trigger resource. Value: An array of JSON strings, where each string is a CDN PID as defined in . Mandatory-to-Specify: No. The default is no CDN path. Description: The time at which the trigger resource was received by the dCDN. The time is determined by the dCDN; there is no requirement to synchronize clocks between interconnected CDNs. Value: Time, as defined in Section 4.3.4 of . Mandatory-to-Specify: The dCDN MUST specify in trigger status representations. It is ignored when included in trigger representations sent by the uCDN. Description: The time at which the trigger resource was last modified. The time is determined by the dCDN; there is no requirement to synchronize clocks between interconnected CDNs. Value: Time, as defined in Section 4.3.4 of . Mandatory-to-Specify: The dCDN MUST specify in trigger status representations. It is ignored when included in trigger representations sent by the uCDN. Description: The estimate of the time at which the dCDN expects to complete the trigger processing. Time is determined by the dCDN; there is no requirement to synchronize clocks between interconnected CDNs. Value: Time, as defined in Section 4.3.4 of . Mandatory-to-Specify: The dCDN MAY specify in trigger status representations. It is ignored when included in trigger representations sent by the uCDN. Description: The current trigger state. Value: Trigger state, as defined in . Mandatory-to-Specify: The dCDN MUST include trigger state in the trigger resource representations it sends. The trigger state defaults to "pending" when a trigger is created and is optional in trigger update requests sent by the uCDN. Description: A human-readable explanation for the object state. Value: A JSON string, the human-readable reason. Mandatory-to-Specify: No. The dCDN MAY include the trigger reason in the trigger resource representations it sends. Description: Descriptions of errors that have occurred while processing the trigger. Value: An array of Error.v2 Descriptions, as defined in . An empty array is allowed and is equivalent to omitting the "errors" attribute from the object. Mandatory-to-Specify: No. The dCDN SHOULD include this attribute in the trigger resource representations it sends when the trigger is in a "failed" state. Description: Total aggregate number of objects affected by the trigger, e.g., number of objects purged, invalidated or prepositioned as a result of trigger processing. Value: Integer. Mandatory-to-Specify: No. This attribute is "optional-to-implement". When supported, the dCDN SHOULD include this attribute in the trigger resource representations requested by the uCDN. Description: The total number of unique dCDN nodes affected by the trigger. Value: Integer. Mandatory-to-Specify: No. This attribute is "optional-to-implement". When supported, the dCDN SHOULD include this attribute in the trigger resource representations requested by the uCDN. Description: Total aggregate size of objects affected by the trigger, in bytes. Value: Integer. Mandatory-to-Specify: No. This attribute is "optional-to-implement". When supported, the dCDN SHOULD include this attribute in the trigger resource representations requested by the uCDN. Description: List of objects derived by the dCDN when processing the trigger. When the "content-objectlist" trigger specification is used, the trigger MAY include one or more manifest files that form a hierarchical structure (e.g., media HLS playlists referenced by master HLS playlists, or JSON object lists referenced by other JSON object lists). The purpose of this optional attribute is to provide a single, flattened list of content objects derived from the input trigger. Value: An array of ContentObject objects. The dCDN SHOULD provide the list of objects it used as input for processing the trigger with , provided that the dCDN advertised support for extended status. An empty array is allowed and is equivalent to omitting "objects" from the trigger representation. This field is intended to provide the list of all objects used in processing. The objects that failed to process SHOULD be specified using the Error.v2 Description resource. Mandatory-to-Specify: No. The dCDN SHOULD only send this attribute in the trigger resource representation when it advertises support for the extended representation via FCI and and the extended representation is requested by the uCDN. The attributes "total-objects-count", "total-nodes-count", and "total-objects-size" are cumulative progress indicators. It is RECOMMENDED that the dCDN maintains these attributes as cumulative counters across all of its nodes, as well as any cascaded dCDNs, without de-duplicating the same objects processed in multiple nodes. The dCDN MAY update these values as trigger processing progresses while the trigger is in the "active" state. The primary purpose of these attributes is to help the uCDN detect abnormal trigger processing results, e.g., a purge or preposition trigger that impacted a lower or higher number of objects than expected.

A trigger action is used in a trigger resource to describe trigger actions. It was previously referred to in as "Trigger Type". All trigger actions MUST be registered in the IANA "CDNI CI/T Trigger Types" registry (see ). A dCDN that receives a request containing a trigger action it does not recognize or support MUST reject the request with a 4xx HTTP status code (for example, 400 "Bad Request"), as described in . Such a request does not create a trigger and therefore produces no Error.v2 Description. The following trigger actions are defined by this document: JSON string Description preposition A request for the dCDN to acquire metadata and/or content. invalidate A request for the dCDN to invalidate metadata and/or content. After servicing this request, the dCDN will not use the specified data without first revalidating it using, for example, an "If-None-Match" HTTP request. The dCDN need not erase the associated data. purge A request for the dCDN to erase metadata and/or content. After servicing the request, the specified data MUST NOT be held on the dCDN (the dCDN MUST reacquire the metadata and/or content from the uCDN if it needs it). The dCDN MUST support at least one of the trigger actions. An "invalidate" or "purge" trigger that does not match any objects known to the dCDN MUST NOT be treated as an error condition. In such cases, the trigger MAY be completed successfully even if no objects were affected.

The CDNI Control Interface / Triggers 1st edition defines a set of properties and objects used by the trigger commands to specify the targets upon which the trigger is applied. This document modifies the trigger interface objects so that it has a list of trigger specs. Such structure improves the interface's extensibility and flexibility. Furthermore, the document defines a generic trigger spec object that acts as a wrapper for managing individual CDNI trigger specs in an abstract manner, allowing future extension of the interface. All trigger specs MUST be registered in the IANA "CDNI CI/T Trigger Specs" registry (see ). A dCDN that receives a trigger creation request containing a trigger specification it does not recognize or support MUST reject the request with a 400 ("Bad Request") HTTP status code. The response body SHOULD contain an Error.v2 Description with the error code "espec", including the portion of the trigger specification that caused the rejection, as defined in , and a human-readable description identifying the unsupported or unrecognized element. Such a request does not create a trigger resource. This document defines an initial set of trigger spec objects and registers them in the IANA "CDNI CI/T Trigger Specs" registry: JSON string Description urls Allowing the specification of trigger targets via URLs. ccids Allowing the specification of trigger targets via CCIDs content groupings, as defined in section 4.2.8 . uri-pattern-match Allowing the specification of trigger targets via URI patterns. uri-regex-match Allowing the specification of trigger targets via regexes matching their URI, as defined in . content-objectlist Allowing the specification of trigger targets using lists of objects. The dCDN MUST support the "urls" trigger spec. Support for all other trigger specs is OPTIONAL. Each trigger usually refers to the targets by the target URLs, using a "urls" trigger spec object or some aggregating spec such as the "url-regex-match". If content URLs are transformed by a transit CDN in a cascade, that transit CDN MUST similarly transform URLs in triggers it passes to its dCDNs. When processing a trigger, CDNs MUST ignore the URL scheme (HTTP or HTTPS) in comparing URLs. For example, for a CI/T "invalidate" or "purge" action, content MUST be invalidated or purged regardless of the protocol clients used to request it.

A trigger resource, as defined in , includes an array of trigger spec objects. Each trigger spec object contains properties that are used as trigger target selection directives for the dCDN when processing the trigger, e.g., content URLs or metadata URL patterns. Each such trigger spec is a specialization of a CDNI GenericTriggerSpec object. The GenericTriggerSpec object abstracts the basic information required for trigger distribution from the specifics of any given property (i.e., property semantics, enforcement options, etc.). The semantics of the trigger specs list is additive, i.e., the trigger applies to any object matching one of the listed specs. A GenericSpecObject object is a wrapper for managing individual CDNI trigger specs in an opaque manner. It is encoded as a JSON object containing the following attributes: Description: Case-insensitive CDNI trigger subject. Value: String containing the type of the subject matching the "cit-spec-value" property, such as "content" or "metadata" as defined in . Mandatory-to-Specify: Yes. Description: Case-insensitive CDNI trigger spec type. Value: String containing the spec type of the object contained in the cit-spec-value property (see table in ). Mandatory-to-Specify: Yes. Description: A CDNI trigger spec object. Value: Defined by the value of the cit-spec-type property. Mandatory-to-Specify: Yes. The structure of a JSON-serialized GenericTriggerSpec object, containing a specific trigger spec is illustrated below: , "cit-spec-value": { }, "trigger-subject": } ]]>

Because the scope of the trigger may relate to metadata and/or content, the "trigger spec object" also specifies the trigger's target subject (i.e., "metadata" or "content") against which to match. All trigger subjects MUST be registered in the IANA "CDNI CI/T Trigger Subjects" registry (see ). A dCDN that receives a trigger creation request containing a trigger subject it does not recognize or support MUST reject the request with a 400 ("Bad Request") HTTP status code. The response body SHOULD contain an Error.v2 Description with the error code "esubject", identifying the unrecognized or unsupported subject. Such a request does not create a trigger resource. This document also defines an initial set of trigger subject values and registers them in the IANA "CDNI CI/T Trigger Subjects" registry: JSON string Description metadata Indicating the trigger target specification refers to Metadata object(s), as defined at . content Indicating the trigger target specification refers to client-facing content objects. The dCDN MUST support at least one of the trigger subjects. The dCDN MAY advertise separate endpoints for each one of the two trigger subjects, using the CI/T Endpoint Capability Object.

There are certain constraints on the way the trigger specs can be combined with trigger subject and trigger actions: Trigger spec types Trigger subject ("content" and "metadata") Trigger action ("preposition", "purge", and "invalidate") urlsAnyAny ccids"content" only"purge" or "invalidate" uri-pattern-matchAny"purge" or "invalidate" uri-regex-matchAny"purge" or "invalidate" content-objectlistAnyAny The trigger specification types "ccids", "uri-pattern-match", and "uri-regex-match" operate by matching attributes of objects already known to the dCDN. As a result, these specification types can only identify objects that are present in, or otherwise known to, the dCDN at the time the trigger is processed, and therefore cannot be used for "preposition" actions. CCIDs, as defined in , apply to groupings of content objects and are not defined for metadata objects. Consequently, the "ccids" trigger specification type is limited to triggers whose subject is "content".

The "urls" spec type allows the uCDN to manage uCDN content or metadata objects held by the dCDN based on the objects' URLs. Full URLs SHOULD be used to ensure unambiguous identification of the referenced objects. The URLs spec is encoded as a JSON object containing the following attributes: Description: An array of URLs over which the trigger MUST be executed. Value: A JSON array of URLs, each represented as a JSON string. Mandatory-to-Specify: Yes. Description: Type of URL used. Value: URL Type as defined in . Mandatory-to-Specify: No. When omitted or empty, the "published" URL type is assumed. Below is an example of a JSON-serialized URLs spec object, matching the metadata at metadata.example.com/a/b/c.

The "ccids" spec type allows the uCDN to specify the Content Collection IDentifier (CCID) of content to which the trigger applies. The CCID is a grouping of content as defined by . The "ccids" spec type is valid only for the "content" spec subject (see ). CCIDs spec is encoded as a JSON object containing the following attributes: Description: An array of Content Collection IDentifiers over which the trigger MUST be executed. Value: A JSON array of strings, where each string is a Content Collection IDentifier. Mandatory-to-Specify: Yes.

The "uri-pattern-match" spec type allows the uCDN to manage its content or metadata objects held by the dCDN based on the objects' URI pattern. The value is a UriPatternMatch object, as defined in .

A UriPatternMatch consists of a string pattern to match against a URI, and flags describing the type of match. It is encoded as a JSON object containing the following attributes: Description: A pattern for URI matching. Value: The pattern represented as a JSON string. The pattern can contain the wildcards "*" and "?", where "*" matches any sequence of pchar or "/" characters (including the empty string) and "?" matches exactly one pchar character. The three literals "$", "*", and "?" MUST be escaped as "$$", "$*" and "$?" (where "$" is the designated escape character). All other characters are treated as literals. Mandatory-to-Specify: Yes. Description: Flag indicating whether or not case-sensitive matching SHOULD be used. Value: A JSON boolean. When set to "true", matching is case-sensitive; when set to "false", matching is case-insensitive. Mandatory-to-Specify: No; default is "false", i.e., a case-insensitive match. Description: Flag indicating whether to include the query part of the URI when compared against the pattern. Value: One of the JSON values "true" (the full URI including the query part SHOULD be compared against the given pattern) or "false" (the query part of the URI SHOULD be dropped before comparison with the given pattern). Mandatory-to-Specify: No; default is "false". The query part of the URI SHOULD be dropped before comparison with the given pattern. Description: Type of URLs to match. Value: URL Type as defined in . Mandatory-to-Specify: No. When omitted or empty, "published" URL type is assumed. Example of case-sensitive prefix match against "https://www.example.com/trailers/":

The "uri-regex-match" spec type allows the uCDN to manage content or metadata objects held by the dCDN based on the objects' URI regex.

A RegexMatch consists of a regular expression string against which a URI is matched, and flags describing the type of match. It is encoded as a JSON object with the following properties: Description: A JSON string containing the regular expression for URI matching. Value: A regular expression to match against the URI, i.e., against the path-absolute and the query string parameters . The regular expression string MUST be compatible with POSIX Section 9 Extended Regular Expressions. This regular expression MUST be evaluated in the POSIX locale (POSIX Section 7.2). Note: Because '\' has a special meaning in JSON as the escape character within JSON strings, the regular expression character '\' MUST be escaped as '\\'. Mandatory-to-Specify: Yes. Description: Flag indicating whether or not case-sensitive matching should be used. Value: JSON boolean. Either "true" (the matching is case-sensitive) or "false" (the matching is case insensitive). Mandatory-to-Specify: No; default is "false", i.e., a case-insensitive match. Description: Flag indicating whether to include the query part of the URI when compared against the regex. Value: JSON Boolean. Either "true" (the full URI, including the query part, should be compared against the regex) or "false" (the query part of the URI should be dropped before comparison with the given regex). Mandatory-to-Specify: No; default is "false". The query part of the URI MUST be dropped before comparison with the given regex. This makes the regular expression simpler and safer for cases in which the query parameters are not relevant to the match. Description: Type of URLs to match against. Value: URL Type as defined in . Mandatory-to-Specify: No. When omitted or empty, "published" URL type is assumed. Example of a case-sensitive, no query parameters, regex match against is below. Please note that some lines in the example are wrapped for clarity. This regex matches URLs of the domain "video.example.com" where the path structure is /(single lower case letter)/(name-of-title)/(single digit between 1 to 7)/(index.m3u8 or a 3 digit number with ts extension). For example: or

The "objectlist" spec type allows the uCDN to manage content or metadata held by the dCDN based on structured object lists. The ObjectList spec type is valid only for the "content" spec subject (see ). An object list is encoded as a JSON object with the following properties: Description: An array of objects to be used in the trigger Value: Array of ContentObject objects Mandatory-to-Specify: Yes.

A "trigger" object, as defined in includes an optional array of trigger extension objects. A trigger extension contains properties that are used as directives for the dCDN when executing the trigger command, e.g., location policies, time policies, and so on. Each such CDNI trigger extension is a specialization of a CDNI GenericTriggerExtension object. The GenericTriggerExtension object abstracts the basic information required for trigger distribution from the specifics of any given property (i.e., property semantics, enforcement options, etc.). All trigger extensions are optional, and it is thus the responsibility of the extension specification to define a consistent default behavior for extensions supported by the dCDN when not specified by the uCDN.

The trigger enforcement options concept is in accordance with the metadata enforcement options as defined in Section 3.2 of . The GenericTriggerExtension object defines the properties contained within it as well as whether or not the properties are "mandatory-to-enforce". If the dCDN does not understand or support a mandatory-to-enforce property, the dCDN MUST NOT execute the trigger command. If the extension is not mandatory-to-enforce, then that GenericTriggerExtension object can be safely ignored and the trigger command can be processed in accordance with the rest of the CDNI trigger spec. Although a CDN MUST NOT execute a trigger command if a mandatory-to-enforce extension cannot be enforced, it could still be safe for a transit CDN (tCDN) to redistribute that trigger (the "safe-to-redistribute" property) to another CDN without modification, provided the tCDN does not need to do trigger processing of its own and only pass the trigger to one or more dCDNs. For example, in the cascaded CDN case, a transit CDN (tCDN) could convey mandatory-to-enforce trigger extension to the dCDN. For a trigger extension that does not require customization or translation (i.e., trigger extension that is safe-to-redistribute), the data representation received off the wire MAY be stored and redistributed without being understood or supported by tCDN. However, for a trigger extension that requires translation, transparent redistribution of the uCDN trigger values might not be appropriate. Certain trigger extensions can be safely, though perhaps not optimally, redistributed unmodified. For example, a preposition command might be executed in suboptimal times for some geographies if transparently redistributed, but it might still work. Redistribution safety MUST be specified for each GenericTriggerExtension listed. If a CDN does not understand or support a given GenericTriggerExtension object that is not safe-to-redistribute, the CDN MUST set the "incomprehensible" flag to true for that GenericTriggerExtension object before redistributing it. The "incomprehensible" flag signals to the dCDN that trigger metadata was not properly transformed by the tCDN. A CDN MUST NOT attempt to execute a trigger with an extension that has been marked as "incomprehensible" by the uCDN. tCDNs MUST NOT change the value of mandatory-to-enforce or safe-to-redistribute when propagating a trigger to the dCDN. Although a tCDN can set the value of "incomprehensible" to true, a tCDN MUST NOT change the value of "incomprehensible" from true to false. describes the action to be taken by a tCDN for the different combinations of mandatory-to-enforce ("MtE") and safe-to-redistribute ("StR") properties when the tCDN either does or does not understand the trigger extension object in question: MtE StR Extension object understood by tCDN Trigger action False True True Can execute and redistribute. False True False Can execute and redistribute. False False False Can execute. MUST set "incomprehensible" to true when redistributing. False False True Can execute. Can redistribute after transforming the trigger extension (if the CDN knows how to do so safely); otherwise, MUST set "incomprehensible" to true when redistributing. True True True Can execute and redistribute. True True False MUST NOT execute but can redistribute, provided own processing is not required. True False True Can execute. Can redistribute after transforming the trigger extension (if the CDN knows how to do so safely); otherwise, MUST set "incomprehensible" to true when redistributing. True False False MUST NOT execute. May redistribute, provided own processing is not required. MUST set "incomprehensible" to true when redistributing. describes the action to be taken by the dCDN for the different combinations of mandatory-to-enforce and "incomprehensible" ("Incomp") properties, when the dCDN either does or does not understand the trigger extension object in question: MtE Incomp Extension object understood by the dCDN Trigger action False False True Can execute. False True True Can execute but MUST NOT interpret/apply any trigger extension marked as "incomprehensible". False False False Can execute. False True False Can execute but MUST NOT interpret/apply any trigger extension marked as "incomprehensible". True False True Can execute. True True True MUST NOT execute. True False False MUST NOT execute. True True False MUST NOT execute.

A GenericTriggerExtension object is a wrapper for managing individual CDNI Trigger extensions in an opaque manner. It is encoded as a JSON object containing the following attributes: Description: Case-insensitive CDNI trigger extension object type. Value: String containing the CDNI Extension Type of the object contained in the "cit-extension-value" property (see table in ). Mandatory-to-Specify: Yes. Description: CDNI trigger extension object. Value: Defined by the value of the "cit-extension-type" property above. Mandatory-to-Specify: Yes. Description: Flag identifying whether or not the enforcement of this trigger extension is mandatory. Value: Boolean. Mandatory-to-Specify: No. The default is to treat the trigger extension as mandatory to enforce (i.e., a value of True). Description: Flag identifying whether or not this trigger extension can be safely redistributed without modification, even if the CDN fails to understand the extension. Value: Boolean. Mandatory-to-Specify: No. The default is to allow transparent redistribution (i.e., a value of True). Description: Flag identifying whether or not any CDN in the chain of delegation has failed to understand and/or failed to properly transform this trigger extension object. Note: This flag only applies to trigger extension objects whose "safe-to-redistribute" property has a value of False. Value: Boolean. Mandatory-to-Specify: No. The default is comprehensible (i.e., a value of False). The structure of a JSON-serialized GenericTriggerExtension object containing a specific trigger extension object is illustrated below: , "cit-extension-value": { }, "mandatory-to-enforce": , "safe-to-redistribute": , "incomprehensible": } ]]>

The objects defined below are intended to be used in the GenericTriggerExtension object's cit-extension-value field as defined in , and their cit-extension-type property MUST be set to the appropriate Extension Type as defined in .

A content operation may be relevant for a specific geographical region or need to be excluded from a specific region. In this case, the trigger should be applied only to parts of the network that are either "included" or "not excluded" by the location policy. Note that the restrictions here are on the cache location rather than the client location. The LocationPolicy object defines the cache or CDN locations in which the trigger is to be executed, thereby constraining the trigger's scope to those locations. Although users are typically proximate to the corresponding cache nodes, the policy applies to the caches themselves, not to the users' locations. Example use cases: Preposition: Because modern streaming content often includes numerous renditions - across resolutions, bitrates, and languages - prepositioning all variants can be untenable. The uCDN MAY perform location-aware prepositioning by selecting which content renditions to preposition based on the languages prevalent in a given region. For example, only Danish, Norwegian and Swedish audio or subtitle renditions might be prepositioned in southern Scandinavia. Purge: In certain cases, content may have been located on servers in regions where the content must not reside. In such cases, a purge operation to remove content specifically from that region is required. Object specification: Description: An Access List that allows or denies (blocks) the trigger execution per cache location. Value: Array of LocationRule objects (see Section 4.2.2.1 of ). The LocationRule utilizes Footprint objects to define footprints in which the rule is to be applied, as defined in Section 4.2.2.2 of and extended by to support the "subdivisioncode" footprint type. Mandatory-to-Specify: Yes. If a location policy object is not listed within the trigger command, the default behavior is to execute the trigger in all available caches and locations of the dCDN. The trigger command is allowed or denied for a specific cache location according to the action of the first location whose footprint matches that cache's location. The evaluation order is implicit and follows the sequence of LocationRule objects as they appear in the extension. If two or more footprints overlap, the first footprint that matches against the cache's location determines the action a CDN MUST take. If the "locations" property is an empty list, or if none of the listed footprints match the location of a given cache, the trigger MUST be treated as not applicable to that cache - that is, the result is equivalent to a "deny" action. The following is an example of a JSON-serialized generic trigger extension object containing a location policy object that allows the trigger execution in the US but blocks its execution in Massachussetts. The execution of the trigger outside of the US is blocked implicitly.

The uCDN MAY wish to perform content management operations on the dCDN on a specific schedule. The TimePolicy extension allows the uCDN to instruct the dCDN to execute the trigger command in a desired time window. For example, a video content provider may wish to pre-populate a new episode during off-peak hours so that it is ready on caches at prime time when the episode is released for viewing. A scheduled operation enables the uCDN to direct the dCDN in what time frame to execute the trigger. This specification supports region-by-region time scheduling when used in conjunction with the Location Policy defined in . The uCDN can trigger separate commands for different geographical regions using a different schedule for each region. This allows the uCDN to control the execution time per region. Object specification: Description: A UNIX epoch time window in which the trigger SHOULD be executed. Value: TimeWindow object using UNIX epoch timestamps (see Section 4.2.3.2 of ). Mandatory-to-Specify: No, but exactly one of either "unixEpochWindow" or "utcWindow" MUST be present. Description: A UTC time window in which the trigger SHOULD be executed. Value: UTCWindow object as defined in . Mandatory-to-Specify: No, but exactly one of either "unixEpochWindow" or "utcWindow" MUST be present. If a time policy object is not listed within the trigger command, the default behavior is to execute the trigger in a time frame most suitable to the dCDN taking under consideration other constraints and / or obligations. If trigger processing cannot be completed within the limits specified by the TimePolicy extension, the dCDN SHOULD set the trigger state to "failed" and include an Error.v2 Description with error "eextension" (see ). Example of a JSON-serialized generic trigger extension object containing a time policy object that schedules the trigger execution to a window between 09:00 01/01/2000 UTC and 17:00 01/01/2000 UTC, using the "unix-time-window" property:

A UTCWindow object describes a time range in UTC or UTC and a zone offset that can be applied by a TimePolicy. It is encoded as a JSON object containing the following attributes: Description: The start time of the window. Value: Internet date and time as defined in . Mandatory-to-Specify: No. but at least one of "start" or "end" MUST be present and non-empty. If "start" is empty or not specified, the time range is considered to begin at an arbitrary (unspecified) time. Description: The end time of the window. Value: Internet date and time as defined in . Mandatory-to-Specify: No. but at least one of "start" or "end" MUST be present and non-empty. If "end" is empty or not specified, the time range is considered to end at an arbitrary (unspecified) time. Example JSON-serialized UTCWindow object that describes a time window from 02:30 01/01/2000 UTC to 04:30 01/01/2000 UTC: Example JSON-serialized UTCWindow object that describes a time window in New York time zone offset UTC-05:00 from 02:30 01/01/2000 to 04:30 01/01/2000:

Unless specified otherwise, the dCDN is at liberty to decide how to choose trigger commands for execution from all pending commands, whether to process trigger commands sequentially or in parallel, immediately upon acceptance, or with a delay in batches. The uCDN MAY wish to control trigger processing in more detail, including the order of execution, dependencies, and concurrency. Please note that the uCDN MAY request immediate processing of a trigger either by setting its state to "active" at creation time or by modifying a pending trigger to set its state to "active". Example use cases: Priority: The uCDN MAY have multiple trigger commands in "pending" and/or "active" mode. For example, trigger commands with policy constraints, a large number of content objects affected, or other dCDN business logic may take a long time to execute. The uCDN MAY wish to prescribe the order in which the dCDN picks up its trigger commands for execution from the "pending" queue, by indicating a relative priority of each trigger. The priority would affect the selection of trigger commands specific to the requesting uCDN. The dCDN MAY separately prioritize triggers from multiple uCDNs subject to its business logic. Additionally, the uCDN MAY wish to prescribe the order in which parts of "active" triggers are processed. The priority may affect the order within the same trigger and/or multiple triggers that are in "active" state at the same time. Multiple priority-related use cases exist: The uCDN needs to introduce an urgent "purge" or "invalidate" trigger into an existing queue of trigger commands to correct wrong versions of content objects published by it The uCDN needs to indicate which content objects should be prepositioned, purged, or invalidated first, for example prepositioning newer released content before prepositioning updates to an existing catalog Prerequisite: In some cases, the uCDN MAY wish to indicate what trigger commands should be processed and completed before another trigger command is processed. For example, the uCDN MAY want to rectify incorrectly published content by purging content objects and then prepositioning them again. In this case, the uCDN MAY want the preposition trigger command to be processed only after the purge trigger command has been processed because the concurrent processing of these triggers MAY cause the new version of these content objects to be purged. Alternatively, the uCDN MAY wish to condition the execution of purge or invalidation triggers upon the completion or cancellation of long-running preposition triggers to avoid race conditions that would result from processing these in parallel. The prerequisite requirement implies that a previous trigger reaches one of the following states: "complete" for successful completion "failed" for failed processing "cancelled" for completion of cancellation The ExtensionPolicy extension is encoded as a JSON object containing the following attributes: Description: Relative weight of the trigger. When picking a trigger for execution from all pending triggers posted by each uCDN, the dCDN MUST choose the trigger with the highest priority first. Value: An integer in the range -100 to 100, inclusive. Mandatory-to-Specify: No. The value defaults to zero if omitted. Description: Links to trigger resources whose processing should fully finish before starting execution of the current trigger. The triggers SHOULD be in one of the following states to be considered finished: "complete", "failed" or "cancelled". Value: A JSON array of zero or more URLs represented as JSON strings. Mandatory-to-Specify: No. In case of a missing or an empty list, no dependencies are assumed. If the dCDN receives a new trigger with the ExecutionPolicy extension that does not reference an existing trigger URL in the "prerequisites" extension attribute, the dCDN MUST set the trigger state to "failed", set the error to "eextension", and MAY include an optional error description. A trigger modification request that would result in the "prerequisites" attribute containing invalid trigger URLs MUST be rejected by the dCDN with status code 409 ("Conflict"). URL values that reference existing finished triggers are allowed but ignored. A dCDN SHOULD reject a trigger whose ExecutionPolicy contains a "prerequisites" dependency referencing a trigger in the "processed" state, since completion of such a trigger cannot be confirmed. The rejection MUST be indicated by returning HTTP status code 400 ("Bad Request"). The following is an example of a JSON-serialized generic extension trigger object containing an execution policy object that specifies trigger priority of 100, and makes its execution dependent on the completion of the previously created triggers:

The uCDN MAY combine multiple options in the same trigger command. The dCDN determines how to handle an incoming trigger according to the following rules: When the dCDN receives a request to create a trigger, it MUST reject the request with status code 400 ("Bad Request") if the trigger depends on other pending triggers with lower priority in order to prevent deadlocks. Otherwise, if the request for new trigger resource creation sets its state to "active": If the trigger has a TimePolicy extension that has a "start" attribute set in the future, the dCDN MUST reject the request. Otherwise, if the trigger has an ExecutionPolicy extension, with prerequisite triggers that are not finished at the time of the request, the dCDN MUST reject the request. Otherwise, if the trigger has a lower priority than pre-existing triggers in "pending" state, the dCDN MUST reject the request. Please note that the trigger has an implicit priority of zero if no priority is explicitly specified. Otherwise, if the trigger has a TimeWindow that ends before the start of the TimeWindow of any trigger on which it depends, or before the start of the TimeWindow of any pending trigger with higher priority, the dCDN MUST reject the request. Otherwise, the dCDN MAY still reject the trigger due to its business logic (e.g,, if the dCDN processes triggers at specific times during the day). Otherwise, the dCDN SHOULD create the trigger in "active" state and begin its processing. In all of the above failure cases, the dCDN MUST reject the request by returning a 4xx or 5xx HTTP status code. The response body for a rejected request SHOULD contain an Error.v2 Description object identifying the reason for the rejection. Otherwise, the dCDN SHOULD create the new trigger resource as requested and set its state to "pending". The dCDN MAY start the trigger processing at any time after the trigger creation, as long as the prerequisites are met, i.e., the start of processing is within the TimePolicy window, there are no other pending triggers with higher priority, and there are no incomplete prerequisite triggers. The dCDN SHOULD periodically re-evaluate the pending trigger queue for triggers that have a TimePolicy set, to ensure that processing of such triggers completes before the corresponding TimePolicy window expires. If during such evaluation a pending trigger is deemed to have expired, the dCDN MUST set the trigger state to "failed". Whenever a trigger reaches a terminal state, the dCDN SHOULD re-evaluate the queue of pending triggers. If during such evaluation a pending trigger is deemed to have expired, the dCDN MUST set the trigger state to "failed" and the error to "ereject". Otherwise, the dCDN MAY start processing triggers that were previously dependent on the completed trigger. Whenever a pending trigger becomes eligible for processing, the dCDN SHOULD re-evaluate the pending trigger queue. If during such evaluation a pending trigger is deemed to have expired, the dCDN MUST set the trigger state to "failed" and the error to "ereject". Otherwise, the dCDN MAY start processing additional triggers that could not be processed earlier due to their lower priority relative to the active trigger. When the dCDN receives a request to modify the priority and/or dependencies and/or TimeWindow of a pending trigger, it MUST reject the request with status 409 ("Conflict") if the change would result in that trigger depending on other triggers with lower priority, or its TimeWindow ending before the start of the TimeWindow of any trigger it depends on, or before the start of the TimeWindow of any trigger with higher priority. Otherwise, the dCDN MUST reject the request with status code 409 ("Conflict") if the change would result in the trigger depending on other triggers that, directly or indirectly, depend on the modified trigger. When the dCDN receives a request to modify the TimeWindow extension, it MUST reject the request with status code 409 ("Conflict") if the modification would set the "end" time to a value in the past. Otherwise, the dCDN SHOULD accept the request and evaluate whether the modified trigger becomes eligible for processing as a result of the change. If at this time, the modified trigger has a TimePolicy "end" time set in the past, the dCDN MUST set the trigger state to "failed" and the error to "ereject". When the dCDN receives a request to change the state of a pending trigger to "active", it MUST reject the request with status code 409 ("Conflict") if, as a result of the change, the trigger would depend on incomplete triggers, have lower priority than other pending triggers, have a TimeWindow start time in the future, or set a new TimeWindow end time in the past. Whenever a 409 ("Conflict") status code is returned, the dCDN SHOULD include an Error.v2 Description in the response body, with the error code set to "eextension" and a human-readable error description. In all of the above cases involving asynchronous error reporting within accepted triggers, the dCDN SHOULD also provide a human-readable error description in the Error.v2 Description returned in the trigger resource. Otherwise, the dCDN MAY still reject the request with status code 409 ("Conflict") if the processing of the trigger could not be started immediately due to the dCDN business logic. Otherwise, the dCDN MUST set the trigger state to "active" and start its processing.

Trigger labels provide a framework for the uCDN to associate a set of key-value pairs with trigger resources. The labels may be used to simplify the management of a large number of triggers by grouping related triggers and tracking their status using the trigger collection resource associated with the label value, when published by the dCDN (see for more details). In this case, the label values remain fully opaque to the dCDN and are evaluated for trigger grouping purposes only.

The trigger state describes the current state of the triggered activity. It MUST be one of the JSON strings in the following table: JSON string Description pending The trigger processing has not begun yet. active The trigger is currently being executed. complete The trigger processing completed successfully. processed The trigger has been created, and no further status update will be made (can be used in cases where completion cannot be confirmed). failed The trigger processing could not be completed. cancelling The trigger cancellation has been requested by the uCDN. cancelled The trigger has been cancelled in response to a uCDN request. Along with the trigger state, the trigger resource has a state reason property, allowing the dCDN to provide additional information for the trigger state. For example, the dCDN may indicate that the trigger state is "pending" due to one of the execution prerequisites not being fulfilled. Such a prerequisite may be specified via one of the extensions.

An Error.v2 Description reports an error associated with a trigger. The same representation is used in two contexts: within an accepted trigger, to report failures that occur during trigger processing; and within the body of a 4xx or 5xx response, to report the reason a trigger creation request was rejected. A request rejected with a 4xx or 5xx status code (see ) does not create a trigger resource; the Error.v2 Description in the response body identifies the cause of the rejection. Once a request has been accepted, failures in the resulting trigger processing are reported only through Error.v2 Descriptions in the trigger resource, not through HTTP status codes. In a 4xx or 5xx response body, the Error.v2 Description is carried as a standalone JSON object with the payload type "ci-trigger-error.v2". Within a trigger resource, it is carried as a member of the "errors" array with the payload type of the enclosing resource. The Error.v2 Description is encoded as a JSON object with the following attributes: Description: Specifies the known error code for the condition that caused the trigger to fail. Value: Error Code, as defined in . Mandatory-to-Specify: Yes. Description: A human-readable description of the error. Value: A JSON string, the human-readable description. Mandatory-to-Specify: No. Description: Array of trigger spec objects from the corresponding "specs" array in the trigger resource. Only those specs to which the error applies are listed. Value: Array of trigger specs, as defined in , where each spec object MUST be exactly as it appears in the trigger resource. Mandatory-to-Specify: Yes. Description: Array of trigger extension objects copied from the corresponding "extensions" array in the trigger resource. Only those extensions to which the error applies are included, but those extensions MUST be exactly as they appear in the trigger resource. Value: Array of GenericTriggerExtension objects, where each extension object is copied from the "extensions" array values in the trigger resource. Mandatory-to-Specify: No. The "extensions" array SHOULD be used only if the error relates to extension objects. Property omission should be interpreted as "the error is not related to any extension". Description: The CDN PID of the CDN where the error occurred. The "cdn-id" property is used by the originating uCDN or by the propagating dCDN to distinguish in which CDN the error occurred. Value: A non-empty JSON string, where the string is a CDN PID as defined in Mandatory-to-Specify: Yes. The dCDN may use its own CDN PID if it does not want to expose the PIDs of its dCDNs. Description: List of objects that failed to be processed during trigger execution. Value: An array of ContentObject objects. The dCDN SHOULD provide the list of objects that it failed to process during trigger execution with , provided that the dCDN advertised support for extended status. Mandatory-to-Specify: No. An empty array is allowed and is equivalent to omitting "objects" from the Error.v2 Description. In a cascade of CDNs, a downstream CDN MAY reject a propagated trigger creation request with a 4xx or 5xx status code, returning an Error.v2 Description in the response body. Because the transit CDN has already accepted the corresponding request from its own upstream, it cannot relay this rejection as an HTTP status code. Instead, the transit CDN includes the received Error.v2 Description in the "errors" array of its own trigger status resource, reporting it to the upstream CDN as a processing error of the accepted trigger. Example of a JSON-serialized Error.v2 Description object reporting a malformed HLS playlist: Example of a JSON-serialized Error.v2 Description object reporting an unsupported extension object:

This type is used by the dCDN to report an error associated with a trigger (see ), whether it arises during asynchronous processing or is reported in a synchronous 4xx or 5xx response to a rejected request. All Error Codes MUST be registered in the IANA "CDNI CI/T Error Codes" registry (see ). Unknown Error Codes MUST be treated as fatal errors, and the request MUST NOT be automatically retried without modification. The following Error Codes are defined by this document and MUST be supported by an implementation of the CI/T v2 interface. Error Code Description emeta The dCDN was unable to acquire and/or does not have metadata required to fulfill the request. econtent The dCDN was unable to acquire content (CI/T "preposition" commands only). eperm The uCDN does not have permission to create the trigger as requested (e.g., the data is owned by another CDN). ereject The dCDN is not willing to process the trigger (for example, a "preposition" request for content whose TimePolicy extension time limit has expired). ecdn An internal error in the dCDN or one of its dCDNs. ecancelled The uCDN cancelled the request. eunsupported The trigger resource used an "action type" that is not supported by the dCDN. espec An error occurred while parsing a trigger spec, or that the specific trigger spec is not supported by the CDN. esubject An error occurred while parsing a trigger subject, or that the specific trigger subject is not supported by the CDN. eextension An error occurred while parsing or applying a trigger extension, or that the specific extension is not supported by the CDN. A dCDN MUST use these Error Codes only to communicate errors that arise during asynchronous trigger processing after the trigger has been created, or when a transit CDN receives a synchronous rejection from one or more of its downstream CDNs after it has already created the trigger resource itself. Any error condition that the dCDN can determine at trigger creation time MUST be communicated then, using a 4xx HTTP status code, and MUST NOT be deferred to an asynchronous Error Code. Such conditions include, but are not limited to, the dCDN's own lack of support for a trigger type, spec, subject, or extension, and a lack of permission for the uCDN to create the trigger as requested.

As described in , the dCDN maintains RESTful trigger resources that represent actions ("triggers") requested by the uCDN for execution by the dCDN. The trigger index resource maintains references to all trigger collection resources that can be used to retrieve triggers. The dCDN MUST create the trigger index resource when it first instantiates the CI/T interface for a uCDN, before any trigger resources are created. The dCDN MUST also create the unfiltered trigger collection containing all triggers, as well as trigger collections for each trigger state, and include references to these collections in the top-level trigger index resource. The dCDN MUST NOT remove the trigger index resource, the unfiltered trigger collection, or the state-based trigger collections once they have been created, as long as it continues to offer CI/T services to the uCDN. The dCDN MUST update the unfiltered trigger collection, the state-based trigger collections, and any additional filtered trigger collections it publishes, when triggers are created or deleted or when their state changes. The dCDN MAY create additional trigger collection resources with other filters, such as per-label collections or collections combining state and label criteria. The dCDN is not required to publish a collection for every label or filter combination in use; the set of filtered collections beyond those required above is chosen by the dCDN. For every filtered collection it publishes, the dCDN MUST keep the collection contents and its trigger index reference current, and SHOULD remove the collection when no triggers remain that match its filter. As a top-level resource, the trigger index also includes global dCDN attributes, such as the "staleresourcetime" attribute, which regulates the expiration of completed triggers, and the "cdn-id" attribute, which indicates the CDN PID of the dCDN. The trigger index resource representation MUST use a MIME media type of "application/cdni; ptype=ci-trigger-index.v2". A trigger index is encoded as a JSON object containing the following attributes: Description: Array of Trigger Collection View objects. Value: An array of JSON-encoded Trigger Collection View objects, one for each existing trigger collection resource. This includes the unfiltered trigger collection, per-state trigger collections (one for each trigger state as specified in ), and any additional filtered trigger collections the dCDN publishes. Mandatory-to-Specify: Yes. Description: The length of time for which the dCDN guarantees to keep a completed trigger resource. After this time, the dCDN SHOULD delete the trigger resource and all references to it from the collection. Value: A JSON number, which must be a positive integer, representing time in seconds. Mandatory-to-Specify: Yes. Description: The dCDN PID. Value: A JSON string, dCDN's PID, as defined in . Mandatory-to-Specify: No.

The collection of trigger resources represents triggers created by the dCDN, optionally filtered by a parameter. Trigger resources in a collection are usually represented using their unique URIs. Note that the collection may refer to CI/T Resources from several versions of CI/T objects, i.e., a subsequent call for the retrieval of the relevant trigger resource may provide objects of various MIME media types: ci-trigger-status as defined in , ci-trigger.v2 defined in this document, or objects of future CI/T objects versions, based on the version of the JSON object used to create the trigger. To allow the uCDN to check the status of multiple triggers in a single request, the dCDN maintains filtered representations of the trigger collection that contain a subset of all triggers. Per-state trigger collections are mandatory, as specified in Section 4.2. Trigger collections with any other filter are optional-to-implement. For every filtered collection it implements, the dCDN MUST include a reference to it in the trigger index resource. All trigger collection representations MUST use a MIME media type of "application/cdni; ptype=ci-trigger-collection.v2". A trigger collection is encoded as a JSON object containing the following attributes: Description: Links to trigger resources in the collection. Value: A JSON array of zero or more URLs represented as JSON strings. Mandatory-to-Specify: Yes. Description: Array of all triggers in the collection. SHOULD be returned only when an extended trigger collection view is requested as described in . Value: An array of JSON-encoded trigger resources. Mandatory-to-Specify: No. The "trigger-objects" attribute SHOULD only be used by the dCDN that supports and advertises the appropriate extended status for trigger collections (see for details). Description: Describes the filter applied to select the triggers in the collection. Value: A Trigger Filter object, as defined in . Mandatory-to-Specify: No. If the "filter" attribute is omitted or is an empty object, the trigger collection MUST include all existing triggers.

The Trigger Collection View provides a brief description of a trigger collection resource within the trigger index. It is encoded as a JSON object containing the following attributes: The "filter" attribute describes the selection criteria used for the referenced trigger collection. If "filter" is omitted, the collection represents all triggers. Description: Describes the filter applied to select the triggers included in the described trigger collection. Value: A Trigger Filter object, as defined in . Mandatory-to-Specify: No. If the "filter" attribute is omitted or is an empty object, the described trigger collection includes all existing triggers. Description: URI of the trigger collection. Value: A URI represented as a JSON string. Mandatory-to-Specify: Yes.

This section describes common CI/T objects, which are used as part of the specification of several other CI/T objects, and their encodings.

This type is used by the uCDN to indicate how URLs referenced by trigger specifications are to be interpreted. URL types apply to trigger specs that reference URLs, such as , , , and . The following URL types are defined by this document: URL Type Description published Published URLs used by end users to access content. private Private URLs used by the dCDN to look up content objects in cache. Implementations of the CI/T interface MUST support the "published" URL type. Support for the "private" URL type is OPTIONAL. If both URL types are supported by the dCDN, the uCDN MUST use only one URL type within a given trigger.

Published URLs are the URLs used by end users. When processing a trigger that uses this URL type, the dCDN MUST be able to match the URLs with metadata objects provided by the uCDN. When this is not the case, the dCDN MUST return the error code "emeta". When processing published URLs in a "preposition" trigger action, the dCDN MUST invoke the metadata processing it would normally perform during content acquisition to satisfy an end-user request, for example SourceMetadata (see Section 4.2.1 of ).

Private URLs are based on cache keys dynamically constructed from properties of HTTP requests and/or responses. For example, an origin might specify a cache key using a value returned in a specific HTTP response header. The uCDN MAY use private URLs in "purge" or "invalidate" trigger actions to simplify processing. A dCDN supporting the private URL type SHOULD advertise this capability via FCI using . If the private URL type is not supported, the dCDN MUST reject the request with a 400 ("Bad Request") HTTP status code. The response body SHOULD contain an Error.v2 Description with the error code "eunsupported", identifying the unsupported action. Such a request does not create a trigger resource.

A ContentObject represents a content object referenced by a trigger and is used to construct lists of objects both in trigger specifications and in trigger status reporting. ContentObject objects convey metadata associated with objects, including labels, object type, and object size. The object type determines how the referenced object is processed by the dCDN. A ContentObject MAY represent either a single content object or an object that, when processed, yields references to additional objects. Such expansion allows structured object descriptions to be resolved into individual objects affected by a trigger. ContentObject objects are used in two contexts: to identify objects that are the subject of a trigger action; and to report objects derived or processed by the dCDN during trigger execution. Please note that when the uCDN accesses ContentObject resources published by the dCDN, the same interface authentication and authorization requirements would apply, as when accessing the interface itself. Unless otherwise specified, the ordering of ContentObject elements does not imply processing order. It is RECOMMENDED that ContentObject lists be flattened, when used by the dCDN to return trigger status, avoiding the use of recursion, in order to simplify processing. The ContentObject properties are defined as follows: Description: URI identifying the referenced object. Value: A URI represented as a JSON string. Mandatory-to-Specify: Yes. Description: Indicates how the referenced object is interpreted during trigger processing. Value: ContentObjectType (see ). Mandatory-to-Specify: No. If omitted, the object is treated as a single content object. Description: Object size in bytes. This attribute MAY be used by the dCDN to make processing decisions, such as ignoring objects that are too small or too large. Value: Integer. Mandatory-to-Specify: No. Description: Labels associated with the ContentObject. The semantics are identical to trigger labels. Value: A JSON object, whose members are the object labels. Each member name is a label key and each member value is the corresponding label value; both are JSON strings. A label key and a label value MUST each be no more than 63 characters in length, MUST begin with a letter or a number, and MAY contain letters, numbers, hyphens, dots, and underscores. Mandatory-to-Specify: No. The default is no labels. The following is an example of JSON-serialized ContentObjects:

The "type" attribute specifies how a ContentObject is interpreted by the dCDN. The following ContentObject types are defined by this document: Content Object Type Description object A single object identified by a URI. hls An object containing references to additional objects using the playlist format defined by HTTP Live Streaming (HLS) (). dash An object containing references to additional objects using the description format defined by MPEG-DASH (). mss An object containing references to additional objects using the format defined by Microsoft Smooth Streaming (). The referenced object MAY be used to derive individual content objects. json List of objects encoded using JSON as defined in . text List of objects encoded as plain text, as defined in .

When encoded using JSON, the representation consists of an array of ContentObject JSON objects. Depending on the specified ContentObject type, an object MAY expand into additional content objects, enabling recursive object lists that MAY combine ContentObjects of different types. The content objects encoded using JSON should follow the JSON grammar specification , including explicit newline encoding and absolute URLs MUST be used at all times.

When encoded using the text representation, each line contains a single object URI. Lines are separated by a line-feed character. Empty lines MAY be ignored. Unlike the JSON-encoded Content Object type, the text-based ContentObject will not support a recursive object list structure, and every object specified in it SHOULD be acted upon without additional processing.

Extended Status Type identifies optional representations that provide additional information about trigger processing beyond the base resource representation. Support for Extended Status is advertised by the dCDN using the FCI capability defined in . When supported and requested, Extended Status representations MAY be returned in trigger resources or trigger collections to expose additional status information or processing details. The following Extended Status Types are defined by this document: Extended Status Type Description trigger-state Provides additional information using "objects" attribute in the Trigger object. error-v2-description Provides additional information using the "objects" attribute in the Error.v2 description object. trigger-collection Provides an extended representation of triggers using the "trigger-objects" attribute in the Trigger Collection object. A dCDN MAY support one or more Extended Status Types. A uCDN MAY request an Extended Status representation. If the uCDN requests an Extended Status Type that the dCDN has not advertised as a capability, the dCDN MUST reject the request with an appropriate error response.

The CDN PID consists of the two characters "AS" followed by the CDN provider's Autonomous System number , then a colon (":") and an additional qualifier that is used to guarantee uniqueness in case a particular AS has multiple independent CDNs deployed -- for example, "AS64496:0". If the CDN provider has multiple ASes, the same AS number SHOULD be used in all messages from that CDN provider, unless there are multiple distinct CDNs. If the CDNI Request Routing Redirection interface (RI) described in is implemented by the dCDN, the CI/T interface and the RI SHOULD use the same CDN PID. The use of an Autonomous System (AS) number as part of the CDN PID provides a convenient identifier for many deployments; however, not all CDN providers operate their own AS, and some providers may operate multiple CDNs spanning several ASes or within networks owned by third parties. Consequently, the AS-based format described above SHOULD be considered one possible identifier construction rather than a mandatory or universally applicable scheme. A CDN provider MAY use an alternative identifier construction, provided that the resulting CDN PID remains globally unique within the scope of CDNI interactions and is used consistently across CDNI interfaces. Future specifications may define additional or more flexible mechanisms for CDN provider identification.

The Trigger Filter object describes the criteria used to select the triggers in a trigger collection. It is used by the Trigger Collection resource () and the Trigger Collection View (). A Trigger Filter is encoded as a JSON object containing the following attributes, each of which is optional: Description: The trigger state shared by all triggers in the collection. Value: Trigger state, as defined in . Mandatory-to-Specify: No. Description: Labels shared by all triggers in the collection. Value: A JSON object containing one or more members, each representing a label. Each member name is a label key and each member value is the corresponding label value; both are JSON strings, subject to the same syntax constraints as the "labels" trigger attribute (). The "labels" attribute, if present, MUST NOT be an empty object. Mandatory-to-Specify: No. A trigger belongs to the collection if and only if it satisfies every attribute of the Trigger Filter object: the trigger's state MUST equal the value of the "state" attribute, if present, and the trigger MUST carry every label listed in the "labels" attribute, if present. Filter criteria are combined as a logical AND; a filter offers no means of expressing alternatives.

This section covers the FCI objects required for the advertisement of the specs, extensions, and properties introduced in this document.

The CI/T trigger endpoint capability object is used to advertise one or more CI/T interface endpoints along with CI/T interface versions supported by these endpoints. The capability type is "FCI.CITEndpoint". Version 1, as originally defined in , is the default if this capability is not explicitly declared. A CI/T Endpoint capability object is encoded as an array of JSON objects containing the following attributes: Description: CI/T endpoint URI Value: A URL represented as a JSON string. Mandatory-to-Specify: Yes. Description: A list of CI/T versions supported by the trigger endpoint. Value: An array of JSON strings. Valid values include "v2", corresponding to this version of the interface, and "v1", corresponding to . Mandatory-to-Specify: Yes. Description: Array of trigger subjects supported by the trigger endpoint. Value: An array of Strings containing the type of the subject matching the cit-spec-value property, such as "content" or "metadata" as defined in . Mandatory-to-Specify: No. A missing or empty "trigger-subjects" list means that all trigger subjects are supported by the endpoint. The dCDN SHOULD advertise only one endpoint for every trigger subject and CI/T interface version pair. If more than one interface endpoint supports the same trigger subject and CI/T interface version (e.g., CI/T version 2 interface for content objects), the uCDN SHOULD be able to use any of the advertised CI/T interface endpoints interchangeably.

The following example shows the serialization of a CI/T Endpoint Capability object for a dCDN that supports version 2 of the CI/T interface for content trigger subjects and version 1 for metadata trigger subjects, with one metadata endpoint for both the US and Brazil, and two separate content endpoints for the two countries.

The CI/T supports several trigger actions for different trigger subjects as defined in and . Additional actions, as well as subjects, may be defined in the future. The trigger scope capability object is used to indicate support for a trigger action for a subject. It further specifies the trigger generic spec types that may be used for selecting the targets to which the triggers apply, along with the supported trigger generic extension types. All supported combinations of trigger actions and trigger subjects MUST be explicitly advertised, with one FCI.CITScope object provided for each combination. The "trigger-scope-capability" object matches the "FCI.CITScope" capability type and is encoded as a JSON object containing the following attributes: Description: The supported CDNI CI/T trigger action. Value: A string corresponding to an entry from the "CDNI CI/T Trigger Types" registry , which corresponds to a CDNI CI/T trigger action. Mandatory-to-Specify: Yes. Description: The supported CDNI CI/T trigger subject. Value: A string corresponding to an entry from the "CDNI CI/T Trigger Subjects" registry , which corresponds to a CDNI CI/T trigger subject. Mandatory-to-Specify: Yes. Description: A list of supported CDNI CI/T GenericSpecObject types for trigger action and subject. Value: List of JSON strings corresponding to entries from the "CDNI CI/T Trigger Specs" registry , which correspond to CDNI CI/T GenericSpecObject objects. Mandatory-to-Specify: No. The default in case of a missing or an empty list MUST be interpreted as "no GenericSpecObject types supported". A non-empty list MUST be interpreted as containing "the only GenericSpecObject types that are supported". Description: A list of supported CDNI CI/T GenericExtensionObject types for trigger action and subject. Value: List of JSON strings corresponding to entries from the "CDNI CI/T Trigger Extension Types" registry , which corresponds to a CDNI CI/T GenericExtensionObject object. Mandatory-to-Specify: No. The default in case of a missing or an empty list MUST be interpreted as "no GenericExtensionObject types are supported". A non-empty list MUST be interpreted as containing "the only GenericExtensionObject types that are supported".

The following shows an example of a JSON-serialized CI/T Trigger Scope Capability objects serialization for the dCDN that supports the prepositioning and invalidation of content, using "urls" and "ccids" generic spec types, with "time-policy" but only for the "preposition" action. Note that in this example, purge is not supported, and no actions involving metadata are supported either.

Given a dCDN supports "content-objectlist" trigger spec, the CI/T Content Object Type capability object is used to indicate support for one or more Content Object types listed in by the type property of the "ContentObject" object. The capability type is "FCI.CITContentObjectType". Description: A list of supported ContentObject types. Value: An array of JSON strings representing ContentObjectTypes. Mandatory-to-Specify: No. A missing or an empty list MUST be interpreted as no ContentObject types are supported.

The following shows an example of a JSON-serialized CI/T Content Object Type Capability object serialization for the dCDN that supports "hls", "dash", and "json", in the US only.

The CI/T URL Type capability object is used to indicate support for URL types. The capability type is "FCI.CITUrlType". Description: Array of supported URL types. Value: An array of JSON strings representing URL types. Mandatory-to-Specify: No. A missing or an empty attribute MUST be interpreted as support for "published" URL types.

The following shows an example of a JSON-serialized CI/T URL Type Capability object serialization for the dCDN that supports the private URL type in URL-based trigger spec types.

The CI/T extended trigger status capability object is used to indicate support for extended trigger status, as specified in . The capability type is "FCI.CITExtendedStatus". Description: List of CI/T objects that support extended attributes. Value: An array of JSON strings representing Extended Status types. Mandatory-to-Specify: No. By default, in case of a missing or an empty list, no extended status objects are supported.

The following shows an example of a JSON-serialized CI/T Extended Status Capability object serialization for the dCDN that supports extended status in trigger, Error.v2 description, and trigger collections objects.

This section provides examples of using the CI/T interface and its features. The discovery of the CI/T interface is out of the scope of this document. In an implementation, all CI/T URLs are under the control of the dCDN. The uCDN MUST NOT attempt to ascribe any meaning to individual elements of the path. In examples in this section, the root URI "https://dcdn.example/cit/" is used as the location of the trigger collection resource, and the PID of the uCDN is "AS64496:1".

Below is an example of a "preposition" trigger creation. The uCDN sends HTTP POST request to the trigger collection URI with the trigger representation in the request body.

Below is an example of a CI/T "invalidate" trigger creation. This trigger instructs the dCDN to revalidate: the metadata objects with URLs prefixed by "https://metadata.example.com/a/b/" using case-insensitive matching a single content object identified by the URL "https://www.example.com/a/index.html" the content objects with URLs prefixed by "https://www.example.com/a/b/" using case-sensitive matching

In the following example, a CI/T "invalidate" trigger uses the Regex property to specify the range of content objects for invalidation. The dCDN accepts the trigger creation request. A failure occurs asynchronously during trigger processing, when the dCDN cannot process the regex due to its complexity, and the resulting error is reflected in the trigger resource. Please note that some lines in the example are wrapped for clarity.

In the following example, a CI/T "preposition" trigger uses the ObjectList property to specify the full media library of a specific content. The command fails due to object list parse error and an appropriate error is reflected in the response.

The uCDN can modify triggers while they are in a "pending" state. One example of this might be to adjust a trigger's "specs" and/or "labels" attributes. In the below example, the uCDN updates a trigger created earlier by removing the metadata portion of the trigger spec and adding trigger labels. The dCDN responds with a 200 ("OK") response containing the updated trigger representation.

The uCDN can cancel triggers that are not in a terminal state by requesting to update the trigger state to "cancelled". In case of asynchronous processing, the dCDN will respond by setting the trigger state to "cancelling" and update it "cancelled" when the cancellation is complete.

The uCDN can delete completed and failed triggers to reduce the size of the collections, as described in . For example, to delete the "preposition" trigger from earlier examples:

Once triggers have been created, the uCDN can check their status as shown in the following examples.

The uCDN can fetch all active trigger collections, representing all existing trigger resources.

Before the dCDN starts processing the remaining trigger shown above, it will appear in the collection of pending triggers. For example:

At this point, if no other triggers had been created, the trigger collection for failed triggers would hold the two failed triggers shown above while other trigger collections would be empty. For example:

The uCDN can also examine individual triggers:

The uCDN SHOULD use the ETags and/or Last-Modified headers when polling for changes in trigger collections or the status of individual triggers, as shown in the following examples:

When the trigger processing is complete, the contents of the filtered collections will be updated. The dCDN SHOULD also update the "ETag" and/or "Last-Modified" response headers - whichever was previously sent - when delivering the updated collection representations. The dCDN SHOULD also use cache control headers, such as "Expires" and "Cache-Control", to indicate how caching of the resource representation should happen by the uCDN and intermediate proxies. For example, when the two example triggers are complete, the collections of pending and complete triggers look as follows:

This subsection illustrates the uses of the Execution Policy extension. The uCDN can create a dependency between triggers. For example, a preposition trigger should only be processed by the dCDN after a previous purge trigger has been completed.

The uCDN can also stagger long-running triggers to control processing order. In the following example, the uCDN creates a preposition trigger with higher priority, which dCDN should pick up for execution before the earlier triggers.

The following example shows error propagation in a cascade. An upstream CDN, uCDN-A ("AS64496:0"), sends a CI/T "preposition" command that uses two extensions to control the way the trigger is executed. The receiving CDN, tCDN-B ("AS64500:0"), supports both extensions, accepts the request, and creates the trigger resource. tCDN-B then propagates the trigger to its own downstream CDN, dCDN-C ("AS64501:0"), which does not support the "time-policy" extension and rejects the propagated request with a 400 ("Bad Request") status code and an Error.v2 Description in the response body. tCDN-B has already accepted the request from uCDN-A and cannot relay the rejection as an HTTP status code. It sets the state of its own trigger resource to "failed" and records the Error.v2 Description received from dCDN-C in the "errors" array, retaining dCDN-C's CDN PID in the "cdn-id" attribute. uCDN-A retrieves the error when it queries the trigger resource.

All references to RFC 8007 in the IANA registries should be replaced with references to this document, apart from references associated with the following registrations: Payload Type Specification ci-trigger-command RFC 8007 ci-trigger-status RFC 8007 ci-trigger-collection RFC 8007 The IANA is requested to register the following new Payload Types in the "CDNI Payload Types" registry defined by , for use with the "application/cdni" MIME media type. Payload Type Specification ci-trigger.v2RFCthis ci-trigger-index.v2RFCthis ci-trigger-collection.v2RFCthis ci-trigger-error.v2RFCthis FCI.CITScopeRFCthis FCI.CITContentObjectTypeRFCthis FCI.CITEndpointRFCthis FCI.CITExtendedStatusRFCthis FCI.CITUrlTypeRFCthis [RFC Editor: Please replace RFCthis with the published RFC number for this document.]

Purpose: The purpose of this payload type is to define a new CI/T trigger object (and any associated capability advertisement) Interface: CI/T Encoding: see

Purpose: The purpose of this payload type is to define a new CI/T trigger index object (and any associated capability advertisement) Interface: CI/T Encoding: see

Purpose: The purpose of this payload type is to define a new CI/T trigger collection object (and any associated capability advertisement) Interface: CI/T Encoding: see

Purpose: The purpose of this payload type is to identify a standalone Error.v2 Description object returned in the body of a rejected CI/T request Interface: CI/T Encoding: see

Purpose: The purpose of this payload type is to distinguish FCI advertisement objects for CI/T Endpoint objects Interface: FCI Value: "FCI.CITEndpoint" Encoding: see

Purpose: The purpose of this payload type is to distinguish FCI advertisement objects for CI/T trigger scope Interface: FCI Value: "FCI.CITScope" Encoding: see

Purpose: The purpose of this payload type is to distinguish FCI advertisement objects for CI/T Content Object Type objects Interface: FCI Value: "FCI.CITContentObjectType" Encoding: see

Purpose: The purpose of this payload type is to distinguish FCI advertisement objects for CI/T URL Type objects Interface: FCI Value: "FCI.CITUrlType" Encoding: see

Purpose: The purpose of this payload type is to distinguish FCI advertisement objects for CI/T Extended Status objects Interface: FCI Value: "FCI.CITExtendedStatus" Encoding: see

In the IANA was requested to create a new "CDNI CI/T Trigger Types" registry under the "Content Delivery Network Interconnection (CDNI) Parameters" registry group. Additions to the "CDNI CI/T Trigger Types" registry are made via the RFC Required policy as defined in . In this second edition of the interface, trigger types are referred to as "trigger actions". The "Trigger Types" registry is used for action definitions. Furthermore, this document, and specifically , reuses the definition of "trigger types" as defined in as trigger actions, and provide their specifications, with no modification compared to .

The IANA is requested to create a new "CDNI CI/T Trigger Specs" registry in the "Content Delivery Networks Interconnection (CDNI) Parameters" registry group. The "CDNI CI/T Trigger Specs" namespace defines the valid trigger targets' spec values in , used by the trigger spec object. Additions to the "CDNI CI/T Trigger Specs" registry are made via the RFC Required policy as defined in . The initial contents of the "CDNI CI/T Trigger Specs" registry consist of the names and descriptions listed in , with this document serving as their specification.

The IANA is requested to create a new "CDNI CI/T Trigger Subjects" registry in the "Content Delivery Networks Interconnection (CDNI) Parameters" registry group. The "CDNI CI/T Trigger Subjects" namespace defines the valid trigger targets' subject values in , used by the trigger spec object. Additions to the "CDNI CI/T Trigger Subjects" registry are made via the RFC Required policy as defined in . The initial contents of the "CDNI CI/T Trigger Subjects" registry consist of the names and descriptions listed in , with this document serving as their specification.

The IANA is requested to create a new "CDNI CI/T Content Object Types" registry in the "Content Delivery Networks Interconnection (CDNI) Parameters" registry group. The "CDNI CI/T Content Object Types" namespace defines the valid object list type values in , used by the ContentObject object. Additions to the "CDNI CI/T Content Object Types" registry are made via the Specification Required policy as defined in . The initial contents of the "CDNI CI/T Content Object Types" registry consist of the names and descriptions listed in , with this document serving as their specification.

The IANA is requested to create a new "CDNI CI/T Trigger Extension Types" registry in the "Content Delivery Networks Interconnection (CDNI) Parameters" registry group. The "CDNI CI/T Trigger Extension Types" namespace defines the valid trigger extension type values in , used by the trigger spec object. Additions to the "CDNI CI/T Trigger Extension Types" registry are made via the Specification Required policy as defined in . The initial contents of the "CDNI CI/T Trigger Extension Types" registry consist of the type names and descriptions listed in , with this document serving as their specification.

In the IANA was requested to create a new "CDNI CI/T Error Codes" registry under the "Content Delivery Network Interconnection (CDNI) Parameters" registry group. Additions to the "CDNI CI/T Error Codes" registry are made via the Specification Required policy as defined in . The Designated Expert will verify that new Error Code registrations do not duplicate existing Error Code definitions (in name or functionality), prevent gratuitous additions to the namespace, and prevent any additions to the namespace that would impair the interoperability of CDNI implementations. In this second edition of the interface, the definitions of the Error Codes from are without change. Additionally, the IANA is requested to register three additional error codes, "espec", "esubject", and "eextension", with the specification as defined in .

The IANA is requested to create a new "CDNI CI/T URL types" registry in the "Content Delivery Networks Interconnection (CDNI) Parameters" registry group. The "CDNI CI/T URL Types" namespace defines the valid URL type values in , used by , , , and . The initial contents of the "CDNI CI/T URL Types" registry consist of the names and descriptions listed in , with this document serving as their specification.

The CI/T interface provides a mechanism to allow the uCDN to generate requests into the dCDN and to inspect its own CI/T requests and their current states. The CI/T interface does not allow access to, or modification of, the uCDN or the dCDN metadata relating to content delivery or to the content itself. It can only control the presence of that metadata in the dCDN and the processing work and network utilization involved in ensuring that presence. By examining "preposition" requests to the dCDN, and correctly interpreting content and metadata URLs, an attacker could learn the uCDN's or content owner's predictions for future content popularity. By examining "invalidate" or "purge" requests, an attacker could learn about changes in the content owner's catalog. An attacker or misbehaving uCDN could inject CI/T triggers to generate processing workload in both the dCDN and uCDN. Similarly, a man-in-the-middle attacker could modify valid trigger requests from the uCDN to achieve the same effect. In both cases, that would decrease the dCDN's caching efficiency by causing it to unnecessarily acquire or reacquire content metadata and/or content. The dCDN implementation of CI/T MUST restrict the actions of the uCDN to the data corresponding to that uCDN. Failure to do so would allow the uCDNs to detrimentally affect each other's efficiency by generating unnecessary acquisition or reacquisition load. An origin that chooses to delegate its delivery to a CDN is trusting that CDN to deliver content on its behalf; the interconnection of CDNs is an extension of that trust to the dCDNs. That trust relationship is a commercial arrangement, outside the scope of the CDNI protocols. So, while a malicious CDN could deliberately generate load on the dCDN using the CI/T interface, the protocol does not otherwise attempt to address malicious behavior between interconnected CDNs.

A CI/T implementation MUST support Transport Layer Security (TLS) transport for HTTP (HTTPS) as per . TLS MUST be used by the server side (dCDN) and the client side (uCDN) of the CI/T interface, including the authentication of the remote end, unless alternate methods are used to ensure the security of the information in the CI/T interface requests and responses (such as setting up an IPsec tunnel between the two CDNs or using a physically secured internal network between two CDNs that are owned by the same corporate entity). The use of TLS for transport of the CI/T interface allows the dCDN and the uCDN to authenticate each other using TLS client authentication and TLS server authentication. Once the dCDN and the uCDN have mutually authenticated each other, TLS allows: The dCDN and the uCDN to authorize each other (to ensure that they are receiving trigger requests from, or responding to, an authorized CDN). CDNI commands and responses to be transmitted with confidentiality. Protection of the integrity of CDNI commands and responses. When TLS is used, the general TLS usage guidance in MUST be followed. The mechanisms for access control are dCDN-specific and are not standardized as part of this CI/T specification. HTTP requests that attempt to access or operate on CI/T data belonging to another CDN MUST be rejected using, for example, HTTP 403 ("Forbidden") or 404 ("Not Found"). This is intended to prevent unauthorized users from generating unnecessary load in the dCDNs or the uCDNs due to revalidation, reacquisition, or unnecessary acquisition.

This document does not define a specific mechanism to protect against Denial-of-Service (DoS) attacks on the CI/T interface. However, CI/T endpoints can be protected against DoS attacks through the use of TLS transport and/or via mechanisms outside the scope of the CI/T interface, such as firewalling or the use of Virtual Private Networks (VPNs). Depending on the implementation, triggered activity may consume significant processing and bandwidth in the dCDN. A malicious or faulty uCDN could use this to generate unnecessary load in the dCDN. The dCDN should consider mechanisms to avoid overload -- for example, by rate-limiting acceptance or processing of triggers, or by performing batch processing.

The CI/T protocol does not carry any information about individual end users of a CDN; there are no privacy concerns for end users. The CI/T protocol does carry information that could be considered commercially sensitive by CDN operators and content owners. The use of mutually authenticated TLS to establish a secure session for the transport of CI/T data, as discussed in , provides confidentiality while the CI/T data is in transit and prevents parties other than the authorized dCDN from gaining access to that data. The dCDN MUST ensure that it only exposes CI/T data related to the uCDN to clients it has authenticated as belonging to that uCDN.

&RFC1930; &RFC2119; &RFC3339; &RFC3986; &RFC8126; &RFC8259; &RFC8006; &RFC8007; &RFC8174; &RFC9110; &RFC9112; &RFC9325; &RFC9388; &RFC9562; &RFC6707; &RFC7336; &RFC7337; &RFC7736; &RFC7975; &RFC8216; ISO Microsoft ECMA International

The authors thank Kevin Ma for his input, and Carsten Bormann for his review and formalization of the JSON data. Initial work on parts of this document was undertaken while Alan Arolovitch was affiliated with Viasat.