gotosocial/docs/federation/interaction_policy.md

Interaction Policy

GoToSocial uses the property interactionPolicy on posts, in order to indicate to remote instances what sort of interactions are (conditionally) permitted to be processed and stored by the origin server, for any given post.

The @context document for interactionPolicy is at https://gotosocial.org/ns.

!!! danger Interaction policy is an attempt to limit the harmful effects of unwanted replies and other interactions on a user's posts (eg., "reply guys").

However, it is far from being sufficient for this purpose, as there are still many "out-of-band" ways that posts can be distributed or replied to beyond a user's initial wishes or intentions.

For example, a user might create a post with a very strict interaction policy attached to it, only to find that other server softwares do not respect that policy, and users on other instances are having discussions and replying to the post *from their instance's perspective*. The original poster's instance will automatically drop these unwanted interactions from their view, but remote instances may still show them.

Another example: someone might see a post that specifies nobody can reply to it, but screenshot the post, post the screenshot in their own new post, and tag the original post author in as a mention. Alternatively, they might just link to the URL of the post and tag the author in as a mention. In this case, they effectively "reply" to the post by creating a new thread.

For better or worse, GoToSocial can offer only a best-effort, partial, technological solution to what is more or less an issue of social behavior and boundaries.

Overview

interactionPolicy is an object property attached to the post-like Objects Note, Article, Question, etc, with the following format:

{
  "@context": [
    "https://gotosocial.org/ns",
    [...]
  ],
  [...],
  "interactionPolicy": {
    "canLike": {
      "always": [ "zero_or_more_uris_that_can_always_do_this" ],
      "approvalRequired": [ "zero_or_more_uris_that_require_approval_to_do_this" ]
    },
    "canReply": {
      "always": [ "zero_or_more_uris_that_can_always_do_this" ],
      "approvalRequired": [ "zero_or_more_uris_that_require_approval_to_do_this" ]
    },
    "canAnnounce": {
      "always": [ "zero_or_more_uris_that_can_always_do_this" ],
      "approvalRequired": [ "zero_or_more_uris_that_require_approval_to_do_this" ]
    }
  },
  [...]
}

In the interactionPolicy object:

• canLike is a sub-policy which indicates who is permitted to create a Like with the post URI as the Object of the Like.
• canReply is a sub-policy which indicates who is permitted to create a post with inReplyTo set to the URI of the post.
• canAnnounce is a sub-policy which indicates who is permitted to create an Announce with the post URI as the Object of the Announce.

And:

• always denotes ActivityPub URIs/IDs of Actors or Collections of Actors who are permitted to create + distribute an interaction without manual approval from the interacted-with user.
• approvalRequired denotes ActivityPub URIs/IDs of Actors or Collections of Actors who are permitted create + distribute an interaction, but should wait for an Accept before distributing an interaction.

Valid URI entries in always and approvalRequired include:

• the magic ActivityStreams Public URI https://www.w3.org/ns/activitystreams#Public
• the URIs of the post creator's Following and/or Followers collections
• individual Actor URIs

For example:

[
    "https://www.w3.org/ns/activitystreams#Public",
    "https://example.org/users/someone/followers",
    "https://example.org/users/someone/following",
    "https://example.org/users/someone_else",
    "https://somewhere.else.example.org/users/someone_on_a_different_instance"
]

!!! info Be aware that according to JSON-LD the values of always and approvalRequired can be either single strings or arrays of strings. That is, the following are all valid:

- Single string: `"always": "https://example.org/users/someone"`
- Single-entry array: `"always": [ "https://example.org/users/someone" ]`
- Multiple-entry array: `"always": [ "https://example.org/users/someone", "https://example.org/users/someone_else" ]`

Specifying Nobody

To specify that nobody can perform an interaction on a post except for its author (who is always permitted), implementations should set the always array to just the URI of the post author, and approvalRequired can be unset, null, or empty.

For example, the following canLike value indicates that nobody can Like the post it is attached to except for the post author:

"canLike": {
  "always": "the_activitypub_uri_of_the_post_author"
},

Another example. The following interactionPolicy on a post by https://example.org/users/someone indicates that anyone can like the post, but nobody but the author can reply or announce:

{
  "@context": [
    "https://gotosocial.org/ns",
    [...]
  ],
  [...],
  "interactionPolicy": {
    "canLike": {
      "always": "https://www.w3.org/ns/activitystreams#Public"
    },
    "canReply": {
      "always": "https://example.org/users/someone"
    },
    "canAnnounce": {
      "always": "https://example.org/users/someone"
    }
  },
  [...]
}

!!! note GoToSocial makes implicit assumptions about who can/can't interact, even if a policy specifies nobody. See implicit assumptions.

Conflicting / Duplicate Values

In cases where a user is present in a Collection URI, and is also targeted explicitly by URI, the more specific value takes precedence.

For example:

[...],
"canReply": {
  "always": "https://example.org/users/someone",
  "approvalRequired": "https://www.w3.org/ns/activitystreams#Public"
},
[...]

Here, @someone@example.org is present in always, and is also implicitly present in the magic ActivityStreams Public collection in approvalRequired. In this case, they can always reply, as the always value is more explicit.

Another example:

[...],
"canReply": {
  "always": "https://www.w3.org/ns/activitystreams#Public",
  "approvalRequired": "https://example.org/users/someone"
},
[...]

Here, @someone@example.org is present in approvalRequired, but is also implicitly present in the magic ActivityStreams Public collection in always. In this case everyone can reply without approval, except for @someone@example.org, who requires approval.

In case the exact same URI is present in both always and approvalRequired, the highest level of permission takes precedence (ie., a URI in always takes precedence over the same URI in approvalRequired).

Default / fallback interactionPolicy

When the interactionPolicy property is not present at all on a post, or the interactionPolicy key is set but its value resolves to null or {}, implementations can assume the following implicit, default interactionPolicy for that post, which reflects the de facto interaction policy of all posts from pre-v0.17.0 GoToSocial, and other ActivityPub server softwares:

{
  "@context": [
    "https://gotosocial.org/ns",
    [...]
  ],
  [...],
  "interactionPolicy": {
    "canLike": {
      "always": "https://www.w3.org/ns/activitystreams#Public"
    },
    "canReply": {
      "always": "https://www.w3.org/ns/activitystreams#Public"
    },
    "canAnnounce": {
      "always": "https://www.w3.org/ns/activitystreams#Public"
    }
  },
  [...]
}

As implied in this default policy, the default value for approvalRequired is an empty array.

!!! info "Actors can only ever interact with a post they are permitted to see" Note that even when assuming a default interactionPolicy for a post, the visibility of a post must still be accounted for by looking at the to, cc, and/or audience properties, to ensure that actors who cannot see a post also cannot interact with it. Eg., if a post is addressed to followers-only, and the default interactionPolicy is assumed, then someone who does not follow the post creator should still not be able to see or interact with it.

!!! tip As is standard across AP implementations, implementations may still wish to limit Announce actities targeting the post to only the author themself if the post is addressed to followers-only.

Defaults per sub-policy

When an interaction policy is partially defined, implementations can make the following assumptions for each potential sub-policy in the interactionPolicy object.

!!! tip Note that the below list is not exhaustive, and extensions to interactionPolicy may wish to define different defaults for other types of interaction.

canLike

If canLike is missing on an interactionPolicy, or the value of canLike is null or {}, then implementations should assume:

"canLike": {
  "always": "https://www.w3.org/ns/activitystreams#Public"
}

In other words, the default is anyone who can see the post can like it.

canReply

If canReply is missing on an interactionPolicy, or the value of canReply is null or {}, then implementations should assume:

"canReply": {
  "always": "https://www.w3.org/ns/activitystreams#Public"
}

In other words, the default is anyone who can see the post can reply to it.

canAnnounce

If canAnnounce is missing on an interactionPolicy, or the value of canAnnounce is null or {}, then implementations should assume:

"canAnnounce": {
  "always": "https://www.w3.org/ns/activitystreams#Public"
}

In other words, the default is anyone who can see the post can announce it.

canQuote

!!! info canQuote is not implemented yet by GoToSocial, and is not yet in the https://gotosocial.org/ns context document, but other implementations may wish to add and use it.

If canQuote is missing on an interactionPolicy, or the value of canQuote is null or {}, then implementations should assume:

"canQuote": {
  "always": "the_activitypub_uri_of_the_post_author"
}

In other words, the default is nobody but the post author can quote the post.

!!! warning The default for canQuote differs from the above defaults for canLike, canReply, and canAnnounce in that it is explicitly opt-in rather than opt-out.

Indicating that verification is required / not required per sub-policy

Because not all servers have implemented interaction policies at the time of writing, it is necessary to provide a method by which implementing servers can indicate that they are aware of and will enforce interaction policies by a) waiting for permission before distributing an Activity (if necessary), b) sending out Accepts and Rejects as appropriate, and c) rejecting unpermitted interactions (see below).

This indication of interaction policy capability is done simply by a server setting interactionPolicy and its sub-policies on outgoing posts, instead of relying on the defaults described above.

That is, by setting interactionPolicy.* on a post, an instance indicates to other instance implementations that they enforce validation of interactions for each sub-policy that is explicitly set.

This means that implementations should always explicitly set all sub-policies on an interactionPolicy for which they have implemented interaction verification, and with which they expect other servers to comply.

For example, if a server understands and wishes to enforce the canLike, canReply, and canAnnounce sub-policies (as is the case with GoToSocial), then they should explicitly set those sub-policies on an outgoing post even when the values do not differ from the implicit defaults, so that other servers know that the origin server does enforcement and knows how to handle appropriate Reject / Accept messages for each sub-policy.

Another example: if a server only implements the canReply interaction sub-policy, and does not care about verification of canLike or canAnnounce, then they should always set canReply, and leave the other two sub-policies out of the interactionPolicy as they cannot understand/enforce them.

Implicit Assumptions

GoToSocial makes several implicit assumptions about interactionPolicys.

Firstly, users mentioned in, or replied to by, a post should ALWAYS be able to reply to that post without requiring approval, regardless of the post visiblity and the interactionPolicy, UNLESS the post that mentioned or replied to them is itself currently pending approval.

This is to prevent a would-be harasser from mentioning someone in an abusive post, and leaving no recourse to the mentioned user to reply.

As such, when sending out interaction policies, GoToSocial will ALWAYS add the URIs of mentioned users to the canReply.always array, unless they are already covered by the ActivityStreams magic public URI.

Likewise, when enforcing received interaction policies, GoToSocial will ALWAYS behave as though the URIs of mentioned users were present in the canReply.always array, even if they weren't.

Secondly, a user should ALWAYS be able to reply to their own post, like their own post, and boost their own post without requiring approval, UNLESS that post is itself currently pending approval.

As such, when sending out interaction policies, GoToSocial will ALWAYS add the URI of the post author to the canLike.always, canReply.always, and canAnnounce.always arrays, unless they are already covered by the ActivityStreams magic public URI.

Likewise, when enforcing received interaction policies, GoToSocial will ALWAYS behave as though the URI of the post author is present in these always arrays, even if it wasn't.

Examples

Example 1 - Limiting scope of a conversation

In this example, the user @the_mighty_zork wants to begin a conversation with the users @booblover6969 and @hodor.

To avoid the discussion being derailed by others, they want replies to their post by users other than the three participants to be permitted only if they're approved by @the_mighty_zork.

Furthermore, they want to limit the boosting / Announceing of their post to only their own followers, and to the three conversation participants.

However, anyone should be able to Like the post by @the_mighty_zork.

This can be achieved with the following interactionPolicy, which is attached to a post with visibility level public:

{
  "@context": [
    "https://gotosocial.org/ns",
    [...]
  ],
  [...],
  "interactionPolicy": {
    "canLike": {
      "always": "https://www.w3.org/ns/activitystreams#Public"
    },
    "canReply": {
      "always": [
        "https://example.org/users/the_mighty_zork",
        "https://example.org/users/booblover6969",
        "https://example.org/users/hodor"
      ],
      "approvalRequired": "https://www.w3.org/ns/activitystreams#Public"
    },
    "canAnnounce": {
      "always": [
        "https://example.org/users/the_mighty_zork",
        "https://example.org/users/the_mighty_zork/followers",
        "https://example.org/users/booblover6969",
        "https://example.org/users/hodor"
      ]
    }
  },
  [...]
}

Example 2 - Long solo thread

In this example, the user @the_mighty_zork wants to write a long solo thread.

They don't mind if people boost and like posts in the thread, but they don't want to get any replies because they don't have the energy to moderate the discussion; they just want to vent by throwing their thoughts out there.

This can be achieved by setting the following interactionPolicy on every post in the thread:

{
  "@context": [
    "https://gotosocial.org/ns",
    [...]
  ],
  [...],
  "interactionPolicy": {
    "canLike": {
      "always": "https://www.w3.org/ns/activitystreams#Public"
    },
    "canReply": {
      "always": "https://example.org/users/the_mighty_zork"
    },
    "canAnnounce": {
      "always": "https://www.w3.org/ns/activitystreams#Public"
    }
  },
  [...]
}

Here, anyone is allowed to like or boost, but nobody is permitted to reply (except @the_mighty_zork themself).

Example 3 - Completely open

In this example, @the_mighty_zork wants to write a completely open post that can be replied to, boosted, or liked by anyone who can see it:

{
  "@context": [
    "https://gotosocial.org/ns",
    [...]
  ],
  [...],
  "interactionPolicy": {
    "canLike": {
      "always": "https://www.w3.org/ns/activitystreams#Public"
    },
    "canReply": {
      "always": "https://www.w3.org/ns/activitystreams#Public"
    },
    "canAnnounce": {
      "always": "https://www.w3.org/ns/activitystreams#Public"
    }
  },
  [...]
}

Subsequent Replies / Scope Widening

Each subsequent reply in a conversation will have its own interaction policy, chosen by the user who created the reply. In other words, the entire conversation or thread is not controlled by one interactionPolicy, but the policy can differ for each subsequent post in a thread, as set by the post author.

Unfortunately, this means that even with interactionPolicy in place, the scope of a thread can inadvertently widen beyond the intention of the author of the first post in the thread.

For instance, in example 1 above, @the_mighty_zork specifies in the first post a canReply.always value of

[
  "https://example.org/users/the_mighty_zork",
  "https://example.org/users/booblover6969",
  "https://example.org/users/hodor"
]

In a subsequent reply, either accidentally or on purpose @booblover6969 sets the canReply.always value to:

[
  "https://www.w3.org/ns/activitystreams#Public"
]

This widens the scope of the conversation, as now anyone can reply to @booblover6969's post, and possibly also tag @the_mighty_zork in that reply.

To avoid this issue, it is recommended that remote instances prevent users from being able to widen scope (exact mechanism of doing this TBD).

It is also a good idea for instances to consider any interaction with a post-like Object that is itself currently pending approval, as also pending approval.

In other words, instances should mark all children interactions below a pending-approval parent as also pending approval, no matter what the interaction policy on the parent would ordinarily allow.

This avoids situations where someone could reply to a post, then, even if their reply is pending approval, they could reply to their own reply and have that marked as permitted (since as author, they would normally have implicit permission to reply).

Interaction Verification

The interaction policy section described the shape of interaction policies, assumed defaults, and explicit assumptions. This section describes the process of how servers that set interaction policies should send approval or rejection of a requested/pending interaction, and how remote servers can prove that permission to interact has been obtained.

Requesting, Obtaining, and Validating Approval

When an actor's URI is in the approvalRequired array for a type of interaction, or their presence in a collection needs to be validated (see below), implementaions wishing to interact with a policied post should do the following:

1. Compose the interaction Activity (ie., Like, Create (reply), or Announce), as normal.
2. Address the Activity to and cc the expected recipients for that Activity, as normal.
3. POST the Activity only to the Inbox (or sharedInbox) of the author of the post being interacted with.
4. DO NOT DISTRIBUTE THE ACTIVITY FURTHER THAN THIS AT THIS POINT.

At this point, the interaction can be considered as pending approval, and should not be shown in the replies or likes collections, etc., of the post interacted with.

It may be shown to the user who sent the interaction as a sort of "interaction pending" modal, but ideally it should not be shown to other users who share an instance with that user.

From here, one of three things may happen:

Rejection

In this scenario, the server of the author of the post being interacted with sends back a Reject Activity with the interaction Activity as the object property.

For example, the following json object Rejects the attempt of @someone@somewhere.else.example.org to reply to a post by @post_author@example.org:

{
  "@context": "https://www.w3.org/ns/activitystreams",
  "actor": "https://example.org/users/post_author",
  "to": "https://somewhere.else.example.org/users/someone",
  "id": "https://example.org/users/post_author/activities/reject/01J0K2YXP9QCT5BE1JWQSAM3B6",
  "object": "https://somewhere.else.example.org/users/someone/statuses/01J17XY2VXGMNNPH1XR7BG2524",
  "type": "Reject"
}

If this happens, @someone@somewhere.else.example.org (and their instance) should consider the interaction as having been rejected. The instance should delete the activity from its internal storage (ie., database), or otherwise indicate that it's been rejected, and it should not distribute the Activity further, or retry the interaction. The server may wish to indicate to the interacter that their interaction was rejected.

Nothing

In this scenario, the author of the post being interacted with never sends back a Reject or an Accept Activity. In such a case, the interaction is considered "pending" in perpetuity. Instances may wish to implement some kind of cleanup feature, where sent and pending interactions that reach a certain age should be considered expired, and Rejected and then removed in the manner gestured towards above.

Acceptance

In this scenario, the author of the post being interacted with sends back an Accept Activity with the interaction Activity as the object property.

For example, the following json object Accepts the attempt of @someone@somewhere.else.example.org to reply to a post by @post_author@example.org:

{
  "@context": "https://www.w3.org/ns/activitystreams",
  "actor": "https://example.org/users/post_author",
  "cc": [
    "https://www.w3.org/ns/activitystreams#Public",
    "https://example.org/users/post_author/followers"
  ],
  "to": "https://somewhere.else.example.org/users/someone",
  "id": "https://example.org/users/post_author/activities/reject/01J0K2YXP9QCT5BE1JWQSAM3B6",
  "object": "https://somewhere.else.example.org/users/someone/statuses/01J17XY2VXGMNNPH1XR7BG2524",
  "type": "Accept"
}

If this happens, @someone@somewhere.else.example.org (and their instance) should consider the interaction as having been approved / accepted. The instance can then feel free to distribute the interaction Activity to all of the recipients targed by to, cc, etc, with the additional property approvedBy (see below).

!!! Note In the above example, actor https://example.org/users/post_author addresses the Accept activity not just to the interacting actor https://somewhere.else.example.org/users/someone, but to their followers collection as well (and, implicitly, to the public). This allows followers of https://example.org/users/post_author on other servers to also mark the interaction as accepted, and to show the interaction alongside the interacted-with post.

approvedBy

approvedBy is an additional property added to the Like, and Announce activities, and any Objects considered to be "posts" (Note, Article, etc).

The presence of approvedBy signals that the author of the post targeted by the Activity or replied-to by the Object has approved/accepted the interaction, and it can now be distributed to its intended audience.

The value of approvedBy should be the URI of the Accept Activity created by the author of the post being interacted with.

For example, the following Announce Activity indicates, by the presence of approvedBy, that it has been Accepted by @post_author@example.org:

{
  "actor": "https://somewhere.else.example.org/users/someone",
  "to": [
    "https://somewhere.else.example.org/users/someone/followers"
  ],
  "cc": [
    "https://example.org/users/post_author"
  ],
  "id": "https://somewhere.else.example.org/users/someone/activities/announce/01J0K2YXP9QCT5BE1JWQSAM3B6",
  "object": "https://example.org/users/post_author/statuses/01J17ZZFK6W82K9MJ9SYQ33Y3D",
  "approvedBy": "https://example.org/users/post_author/activities/accept/01J18043HGECBDZQPT09CP6F2X",
  "type": "Announce"
}

When receiving an Activity with an approvedBy value attached to it, remote instances should dereference the URI value of the field to get the Accept Activity.

They should then validate that the Accept Activity has an object value equal to the id of the interaction Activity or Object, and an actor value equal to the author of the post being interacted with.

Moreover, they should ensure that the URL host/domain of the dereferenced Accept is equal to the URL host/domain of the author of the post being interacted with.

If the Accept cannot be dereferenced, or does not pass validity checks, the interaction should be considered invalid and dropped.

!!! info "Accepts should be dereferenceable" As a consequence of the validation mechanism, instances should make sure that they serve a valid ActivityPub response to dereferences of Accept URIs that pertain to an interactionPolicy. If they do not, they inadvertently risk restricting the ability of remote instances to distribute their posts.

Validating presence in a Followers / Following collection

If an Actor interacting with an object (via Like, inReplyTo, or Announce) is permitted to do that interaction based on their presence in a Followers or Following collection in the always field of an interaction policy, then their server should still wait for an Accept to be received from the server of the target actor, before distributing the interaction more widely with the approvedBy property set to the URI of the Accept.

This is to prevent scenarios where third servers have to somehow verify the presence of the interacting Actor in the Followers or Following collection of the Actor being interacted with. It is simpler to allow the target server to do that verification, and to trust that their Accept implicitly agrees that the interacting Actor is present in the relevant collection.

Likewise, when receiving an interaction from an Actor whose permission to interact matches with one of the Following or Followers collections in the always property, the server of the interacted-with Actor should ensure that they always send out an Accept as soon as possible, so that the interacting Actor server can send out the Activity with the proper proof of acceptance.

This process should bypass the normal "pending approval" stage whereby the server of the Actor being interacted with notifies them of the pending interaction, and waits for them to accept or reject, since there is no point notifying an Actor of a pending approval that they have already explicitly agreed to. In the GoToSocial codebase in particular, this is called "preapproval".

Optional behaviors

This section describes optional behaviors that implementers may use when sending Accept and Reject messages, and should account for when receiving Accept and Reject messages.

Always send out Accepts

Implementers may wish to always send out an Accept to remote interacters, even when the interaction is implicitly or explicitly permitted by their presence in the always array. When receiving such an Accept, implementations may still want to validate the Accept, and, if valid, update their representation of the interaction to include an approvedBy URI pointing at that Accept. This may be useful later on when handling revocations.

Type hinting: inline object property on Accept and Reject

If desired, implementers may partially expand/inline the object property of an Accept or Reject to hint to remote servers about the type of interaction being Accepted or Rejected. When inlining in this way, the object's type and id must be defined at a minimum. For example:

{
  "@context": "https://www.w3.org/ns/activitystreams",
  "actor": "https://example.org/users/post_author",
  "cc": [
    "https://www.w3.org/ns/activitystreams#Public",
    "https://example.org/users/post_author/followers"
  ],
  "to": "https://somewhere.else.example.org/users/someone",
  "id": "https://example.org/users/post_author/activities/reject/01J0K2YXP9QCT5BE1JWQSAM3B6",
  "object": {
    "type": "Note",
    "id": "https://somewhere.else.example.org/users/someone/statuses/01J17XY2VXGMNNPH1XR7BG2524",
    [...]
  },
  "type": "Accept"
}

Set target property on Accept and Reject

If desired, implementers may set the target property on outgoing Accept or Reject activities to the id of the post being interacted with, to make it easier for remote servers to understand the shape and relevance of the interaction that's being Accepted or Rejected.

For example, the following json object Accepts the attempt of @someone@somewhere.else.example.org to reply to a post by @post_author@example.org that has the id https://example.org/users/post_author/statuses/01JJYV141Y5M4S65SC1XCP65NT:

{
  "@context": "https://www.w3.org/ns/activitystreams",
  "actor": "https://example.org/users/post_author",
  "cc": [
    "https://www.w3.org/ns/activitystreams#Public",
    "https://example.org/users/post_author/followers"
  ],
  "to": "https://somewhere.else.example.org/users/someone",
  "id": "https://example.org/users/post_author/activities/reject/01J0K2YXP9QCT5BE1JWQSAM3B6",
  "object": "https://somewhere.else.example.org/users/someone/statuses/01J17XY2VXGMNNPH1XR7BG2524",
  "target": "https://example.org/users/post_author/statuses/01JJYV141Y5M4S65SC1XCP65NT",
  "type": "Accept"
}

If desired, the target property can also be partially expanded/inlined to type hint about the post that was interacted with. When inlining in this way, the target's type and id must be defined at a minimum. For example:

{
  "@context": "https://www.w3.org/ns/activitystreams",
  "actor": "https://example.org/users/post_author",
  "cc": [
    "https://www.w3.org/ns/activitystreams#Public",
    "https://example.org/users/post_author/followers"
  ],
  "to": "https://somewhere.else.example.org/users/someone",
  "id": "https://example.org/users/post_author/activities/reject/01J0K2YXP9QCT5BE1JWQSAM3B6",
  "object": "https://somewhere.else.example.org/users/someone/statuses/01J17XY2VXGMNNPH1XR7BG2524",
  "target": {
    "type": "Note",
    "id": "https://example.org/users/post_author/statuses/01JJYV141Y5M4S65SC1XCP65NT"
    [ ... ]
  },
  "type": "Accept"
}