Manifold Standard

Version: 0.1.12

Abstract

The Manifold Standard maps digital data to a space. It defines several types of JSON objects and how to reliably reconstruct the space from them. The reconstructed space contains a three-dimensional scene, and may specify how users can interact and leave traces in the scene.

Status of This Memo

This document is currently not open to the public.

Relation to the reference implementations

One of the reference implementations of the standard is the project Exterior Space. Exterior Space is an open-source browser for spaces stored on-chain, planned to launch alongside the standard.

Exterior Space implements the standard with its own profile for optional fields and modules. It also includes additional features optimized for browsing and editing spaces. Exterior Space supports the standard, and its major version aligns with the standard’s major version. The maintainers plan to extract the conformant core into a standalone open-source package as the reference implementation of the standard, which Exterior Space will then use. Until that is complete, this document will include footnotes about Exterior Space where relevant.

01 Table of Content

This table of contents is provided as a guide for navigating this document. It does not list all the subsections, but only lists sections to a manageable depth for navigation.

• 00-Manifold-Standard
  • 01-Table-of-Content
  • 02-Copyright-Notice
  • 03-Conventions
  • 04-Version-Control
  • 05-Conformance-Profile
• 10-Introduction
  • 11-Scope-and-Structure

02 Copyright Notice

Copyright (c) 2026 The Manifold Standard Editing Committee and the persons identified as the document authors. All rights reserved.

03 Conventions

03.01 Requirements Language

The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in BCP 14 RFC2119 RFC8174 when, and only when, they appear in all capitals, as shown here.

03.02 Conventions Used in This Document

Any JSON object defined in this document MUST conform to RFC8259.

Some examples use the combination of a TypeScript single-line comment (//) followed by an ellipsis (...) as placeholder notation for content deemed irrelevant by the authors. These placeholders must of course be deleted or otherwise replaced, before attempting to validate the corresponding JSON code example.

Whitespace is used in the examples inside this document to help illustrate the data structures, but it is not required. Unquoted whitespace is not significant in JSON.

03.03 Type Definitions

In this document, a JSON object type defines the required and optional name-value pairs of a JSON object. An object belongs to a type — or "is of that type" — if it contains all required name-value pairs and only name-value pairs that are either required or optional for that type. Each name-value pair in a type definition specifies the name of the pair and the constraint on its value. Required name-value pairs are indicated by the keywords MUST or REQUIRED; optional name-value pairs by MAY, OPTIONAL, SHOULD, or RECOMMENDED, as defined in Section 03.01.

A JSON type, or simply a type, is a constraint on a JSON value. A JSON object type, as defined above, is a type that constrains an object by specifying its required and optional name-value pairs. A type may also constrain a non-object value by specifying the set of allowed values, for example, a number within a particular range, or a string equal to one of a specified set.

03.04 Representation of Data

In this document, TypeScript code is used to represent data, such as JSON object and JSON object type. However, the TypeScript code SHOULD NOT be considered the only form in which the data can be realized in an implementation.

03.99 Notes

The section 03.02 significantly references the standard RFC 7946.

04 Version Control

In this section, the meaning and handling of the version of the standard are defined.

04.01 Version Number

The version number of the standard and each section of the standard use the format MAJOR.MINOR.PATCH. Notice this is not the standard semantic versioning.

1. MAJOR version number is the released stable version of the standard. Each MAJOR bump is a release of a stable version, which is a closure of a set of proposed changes to the standard. Each MAJOR bump resets the MINOR version number to zero and increases the MAJOR version number by one. One can refer to a stable version of the standard by the MAJOR version number. For example, Manifold Standard Version 1 refers to Version 1.0.0 of the standard.
2. MINOR version numbers are used for proposed changes for the next stable version of the standard. Each MINOR bump is a closure of a set of section changes to the standard, with each change labeled with a PATCH version number. Each MINOR bump resets the PATCH version number to zero and increases the MINOR version number by one.
3. PATCH version numbers are used for proposed changes for the next stable version of one section of the standard. A PATCH bump of the standard is performed when a section of the standard is changed. This section is also marked with the full version number it was last changed. This mark can be referred to as the version number of this section.

04.02 Breaking Changes

A breaking change is a change that either:

1. Makes the data format not backward compatible, or
2. Alters the relationship between the data and the reconstructed space in a non-compatible way. This includes changes of default behaviors.

It is RECOMMENDED that stable releases of the standard do not introduce breaking changes. The version number does not imply whether changes are breaking or non-breaking. If a breaking change is included in a stable release, an attachment describing migration from the previous version to the current version MUST be provided in Section 90 of the release.

04.03 Notes on Implementation

1. Closure is an event of obtaining and confirming the consensus of the Manifold Standard Editing Committee. The Manifold Standard Editing Committee is the maintenance team for the Manifold Standard before version 1.0.0. Each closure MUST ends with a version number bump and a tag of the version to the Git repository.
2. If a change in one section requires edits in another section, and if the changes have a clear logical order, each edit SHOULD be followed by a PATCH bump. If term changes, typo changes, or convention changes occur in multiple sections, changes to multiple sections MAY be merged into one PATCH bump. If there is a set of inseparable changes across multiple sections, these changes MAY be merged into one PATCH bump.
3. A section of the standard is defined as one markdown file.

04.04 Implementation of Referencing

1. Each release of the stable version contains an archive action, for which the released standard MUST be uploaded to durable storage, obtaining the URI for referencing the standard of the particular stable version.
2. Each proposed version can be referred to using the git tag.
3. MINOR/PATCH version of the standard MUST NOT be cited as normative in external integrations.

04.05 Errata for Stable Releases

Each post-release erratum of the stable release version is additional documentation called amendments to the stable release version, specifying fixes to the standard of the stable release version. The amended stable release version of the standard is referred to using the version number MAJOR.0.0.ERRATA. It is RECOMMENDED not to issue an erratum for the stable release version unless critical mistakes are made. The amendments MUST be uploaded to durable storage, obtaining the URI for referencing.

04.06 Expiration of This Section

The version numbering, meaning, and handling described in this section are for versions before version 1.0.0. At version 1.0.0, versioning and change control transfer to the governance model that is defined at that time. The governance model after version 1.0.0 may or may not follow this handling of version numbers.

05 Conformance Profile

TODO

10 Introduction

The Manifold Standard specifies a mapping between a defined set of data and a space. Its goal is to define a virtual space where users can store, arrange, decorate, present, generate, update, and interact with digital assets. It provides a canonical data format for serializing these spaces and reliably reconstructing a space from that data. Although similar implementations exist in products and games, they remain closed and therefore fragment users’ memories, creations, and belongings, while a standard provides the common ground for decentralized, user-owned spaces open to anyone, anytime, anywhere, with any digital assets.

The standard is organized into three parts that together define a framework enabling users to experience virtual spaces. The core specifications form the minimal kernel of the standard. They define the scene space and its artifacts, and hence how artifacts are placed within the scene. The extension specifications define how users are connected to the space, that is, how users present themselves and interact within and across spaces. Finally, modules are optional specifications built upon the core and extension specifications to enhance how users experience the space, providing additional layers of interaction, personalization, and growth. Together, the standard enables the space and its artifacts to encode users' live events, helps users share their spaces anywhere, assists users in discovering and engaging with other spaces, brings a sense of time or living creatures into the space, and much more.

11 Scope and Structure

[TODO]

20 Core Specification

The Core Specification defines objects that can be mapped to a scene and objects that represent assets which can be included in the scene.

21 Scene

21.01 Scene and Coordinate type

A scene is a vector space of $\mathbb{R}^3$. We fix the right-handed orientation and an orthonormal basis, labeled $\hat x$, $\hat y$, and $\hat z$. A point in the space can be expressed as a 3-tuple $(x, y, z) = x\,\hat x + y\,\hat y + z\,\hat z$, where $x$, $y$, and $z$ are real numbers.

A Coordinate type represents a point in the scene and is an array of three real numbers. It is represented as

type Coordinate = [number, number, number]

where the three numbers correspond to $x$, $y$, and $z$, in this order.

The ground plane is the subspace of the scene spanned by the basis $\hat x$ and $\hat y$, i.e., the set $\{(x, y, 0)\ |\ x \in \mathbb{R},\; y \in \mathbb{R}\}$. The up direction is defined as the positive direction of the basis vector $\hat z$.

21.02 Coordinate Transformation

A coordinate transformation with uniform scale in the scene defined above transforms a vector $\vec v = (v_x, v_y, v_z)$ to $s\, R \, \vec v + \vec t$, where $R$ is a rotation operator, $s$ is a real number for uniform scale, and $\vec t$ is the translation of the origin. Such a transformation can therefore be encoded in three pieces of information: translation $\vec t$, rotation $R$, and scale $s$.

21.02.01 Rotation and Euler Type

Rotation can be represented by the type Euler, specifying the Euler angles as a 3-tuple $(x, y, z)$ in radians. The rotations are applied in the order of the $x$-axis, then the $y$-axis, then the $z$-axis. It can be represented as

type Euler = [number, number, number]

where the three numbers correspond to rotation in radians around the $x$-axis, $y$-axis, and $z$-axis, in that order. For each rotation, a positive value means counterclockwise rotation. A zero rotation on all axes implies that the axes of the transformed coordinate align with those of the scene coordinate.

21.02.02 Scale and Scale Type

The type Scale represents the scaling of the transformed coordinate about its local origin and is a real number, where a value of 1 means no scaling. It can be represented as

type Scale = number

21.02.03 Local Coordinate

A coordinate distinct from the scene coordinate, also referred to as the global coordinate, is called a local coordinate. A local coordinate can be defined by the coordinate transformation above with respect to the scene coordinate, or can be specified by the origin of the local coordinate in the scene coordinate along with the transformed basis; see Section TODO: local scene.

21.03 Entity in the scene

Each entity in the scene comes with its own local coordinate, and its placement can therefore be described by the three pieces of information defined in Section 21.02. That is, one starts with the entity's local coordinate coinciding with the scene coordinate, then applies the transformation defined in Section 21.02 to the entity.

22 Form Type Object

A Form type object represents the layout of the scene.

Form type is defined as

interface Form {
    version: string // REQUIRED
    layout: Embedding[] // REQUIRED
    metadata?: FormMetadata // OPTIONAL
    properties?: FormProperties // OPTIONAL
}

where

• version is REQUIRED, and its value MUST be of type string equal to one of the standard’s version numbers (see Section 04-Version-Control).
• layout is REQUIRED, and its value MUST be an array of Embedding, representing placed artifacts in the space. An empty layout MUST have the value [].
• metadata is OPTIONAL, and its value MUST be of type FormMetadata. It is used to hold any additional descriptive data about the space. If not provided, an empty FormMetadata is used by default.
• properties is OPTIONAL, and its value MUST be of type FormProperties. It represents additional properties of the scene. If not provided, an empty FormProperties is used by default.

23 Embedding Type Object

An Embedding type object represents an artifact being placed in the scene.

Embedding type is define as

interface Embedding {
	artifact: Artifact // REQUIRED
    position: Coordinate // REQUIRED
    rotation: Euler // REQUIRED
    scale?: Scale //OPTIONAL
    properties?: EmbeddingProperties // OPTIONAL
}

where

• artifact is REQUIRED, and its value MUST be of type Artifact.
• position is REQUIRED, and its value MUST be of type Coordinate.
• rotation is REQUIRED, and its value MUST be of type Euler.
• scale is OPTIONAL, and its value MUST be of type Scale.
• properties is OPTIONAL, and its value MUST be of type EmbeddingProperties. It stores data that modifies or enhances the placed artifact.

An Embedding type object represent an artifact been placed at position with rotation and scale, as defined in 21-Scene#21.03 Entity in the scene.

24 Artifact Type Object

The Artifact type object represents an asset that can be loaded into the space.

Artifact type is defined as

interface Artifact {
	content: string // REQUIRED
	type: ArtifactType // REQUIRED
	metadata: Metadata // RECOMMENDED
}

where

• content is REQUIRED, and stores the content to be loaded into the scene.
• type is REQUIRED, and its value MUST be of type ArtifactType. It specifies the artifact type of content and therefore implies the possible loaders that an implementation can use to load the content.
• metadata is RECOMMENDED, and its value MUST be of type Metadata. It stores relevant information about the asset. If it is not provided, the default MUST be used.

24.01 ArtifactType data

The ArtifactType indicates the artifact type of the asset and therefore the loader that the implementation needs to use to correctly load it. Modules can define more ArtifactType values than those specified in this section.

type ArtifactType = "model"

• For model artifact type, the implementation need to use a GLTFLoader to load the content, following the glTF standard.

The artifact type, the type member of Artifact, and the ArtifactType type should not be confused with the JSON object type or JSON type. In this document, the single word "type" is reserved for JSON type.

24.02 Artifact Content

The artifact content, as specified in the content field of the Artifact object, MUST be a path or Uniform Resource Locator (URL) to the data that can be loaded by the loader and presented in the scene.

For artifact content of type "model", it is RECOMMENDED that the data be a model file following the GLTF standard. When loaded, such a model contains a local origin.

25 Metadata Type Object

The Metadata type object stores data relevant to an asset.

All members in Metadata are OPTIONAL and if they are not specified, the default value will be used. Modules and extension specifications can define additional members in type Metadata beyond those specified in this section. It is RECOMMENDED that modules make use of the attachment and propertiesmembers rather than introducing new members in the Metadata type.

Metadata type is define as

interface Metadata {
	id: string,
	name: string,
	description: string,
	preview: string,
	creators: string[],
	attachment: ArtifactAttachment
	properties: ArtifactProperties
	// ...
}

where

• id MUST be of type string representing the path or asset ID of the artifact. Its default value is the content string of the artifact. It is RECOMMENDED that all artifacts loaded in the scene have distinct id values.
• name MUST be of type string and is the display name of the artifact. Its default value is "Artifact".
• description MUST be of type string and is the display description of the artifact. Its default value is "", i.e., an empty string.
• preview MUST be of type string containing the URL to the preview image file, and it is RECOMMENDED that the preview image file be of type PNG.
• creators MUST be an array of string representing the creators of the artifact.
• attachment MUST be of type ArtifactAttachment, representing any associated files relevant to this artifact. The attachment member is mainly used by modules and extension specifications.
• properties MUST be of type ArtifactProperties. It stores data that modifies or enhances the artifact's behavior in the scene. The properties member is mainly used by modules and extension specifications.

30 Extension Specification

Extension Specifications define objects that represent users and their actions within the space, creating context for spaces and connections between users.

31 User and User Interface

/[TODO/]

A user is an being that must own or have access to an interface, and may have an inventory. Concretely, users include, but are not limited to, human beings, other intelligent beings, programs.

Interface

An interface is the input from a user to space and output to a user from space. The input from a user is a set of events that can trigger a set of corresponding functions known as interactions that changes the space. The output to a user is what user experienced from the space.

Camera

Each user may be associated with a camera. The camera can be moved in response to interface interactions and serves as the user’s primary visual viewport.

External interface

There are predefined sets of external interfaces commonly available on different platforms:

• Desktop:
  • Mouse click (left or right) (at a place)
  • Mouse drag (along a path)
  • Keyboard input (with some characters)
• Mobile:
  • Single tap and multi-tap (at places)
  • long tap (at a place)
  • Tap and drag (along a path)
  • On-screen keyboard input (with some characters)
• System / Null users:
  • System interface: may monitor user-defined events occurring in the space or inventory as interfaces.
  • A null user is a system user that is implemented in a non-explicit way.

Avatar interface

The avatar interface has two layers: the control layer and the avatar layer.

The control layer is the external interface. When triggered, it calls an interaction function to control the avatar and perform a behavior.

The avatar’s behavior then acts as an interface in the avatar layer. Common behaviors include:

• Moving (towards a direction or along a path)
• Standing (at a location)
• Engaging (with an object, self, or another avatar, may with options)

Inventory

An inventory is a list of items associated with a user. It represents the set of artifacts currently held by the user.

An inventory may include special slots that are monitored by system users. These are sometimes referred to as equipment slots or state slots.

In implementation, inventory elements may correspond to items or tokens stored in a user's wallet, or to contextual contents such as a backpack.

Item

Items are typically Artifact objects, but may be any object. They can also be monitored by system users.

32 Space

/[TODO/]

A space is the totality of what a user can experience. Concretely, a space can include the scene, its state, how it evolves and responds to user input, its recorded history, and any data stored within it along with how that data is presented.

In this section we define a few potential constituents of a space besides the scene. The constituents defined here should not be considered the only possible parts of a space beyond the scene.

State

A state of the scene is

UI Component

Interaction

Time

History

Background music and sounds effect