IPLD Specifications

The goal of IPLD is to enable decentralized data-structures that are universally addressable and linkable which in turn will enable more decentralized applications. These data-structures allow us to do for data what URLs and links did for HTML web pages. Read more about the principles that are guiding the ongoing development of IPLD in IPLD Foundational Principles.

IPLD is not a single specification, it is a set of specifications. Many of the specifications in IPLD are inter-dependent.

IPLD Blocks

The block layer encompasses all content addressed block formats and specifies how blocks are addressed, how they self-describe their codec for encoding/decoding, and how blocks link between each other.

IPLD blocks alone do not define data structures or types, although many codecs may convert these formats into native types, there are no type requirements or assurances about types at the block layer.

Documents:

IPLD Codecs

Codecs serve as an intermediary between raw bytes and the IPLD Data Model. They determine how data is converted to and from the Data Model.

Codecs vary in the completeness in which they can represent the IPLD Data Model. DAG-CBOR and DAG-JSON are native IPLD codecs that currently enable the most complete form of the Data Model. Their base codecs, CBOR and JSON, are also valid IPLD codecs, are unable to represent some Data Model kinds on their own, in particular the Link (CID) kind (and Bytes for JSON), so DAG-JSON and DAG-CBOR provide mechanisms to represent these kinds.

IPLD can operate across a broad range of content-addressable codecs, including Git, Ethereum, Bitcoin, and more. DAG-PB is a legacy IPLD format that is still actively used for representing file data for IPFS.

The IPLD Data Model

The Data Model describes a set of base required types to be implemented by a subset of IPLD codecs.

With these basic types authors can create various single-block data structures which can be read with predictable paths and selectors.

With just the data model, several data structures can be authored and put into a single block. These data structures can also link to one another, but a single collection (Map or List) cannot be spread across many blocks with only the Data Model.

Since different systems and transports may impose block size limits (often 2mb or more) in order to control memory usage, larger collections need to be sharded over many blocks at the Schema Layer.

Documents:

Schemas and Advanced Data Layouts

IPLD Schemas define a mapping from the Data Model to instantiated data structures comprising complex layouts. Schemas add the ability to extend the IPLD Data Model to the wide variety of types required for typical programatic interaction with a data source without the need to implement custom translation abstractions.

Schemas will also serve as an enabling layer for complex multi-block data structures via Advanced Data Layouts by providing stability and consistency of data model use within individual blocks and defined interaction points for the logic required for building and interacting with advanced data layouts, such as multi-block Maps, Lists and Sets.

Documents:

Specification document status

Specification documents in this repository fit into one of two categories and have one of three possible statuses:

• Prescriptive
  • Exploratory
  • Draft
  • Final
• Descriptive
  • Draft
  • Final

Prescriptive specifications are intended to describe future implementations or, in some cases, changes to existing implementations.

Descriptive specifications describe existing behavior. In many cases these specifications are not intended to drive new implementations and are only written to understand existing behaviors.

Documents labelled "Specification" in this repository will also be labelled with a descriptor that indicates the category and status. e.g. "Status: Prescriptive - Draft" or "Status: Descriptive - Final".

Design documentation & Library recommendations

Included in this repository are some documents which chronicle our process in developing these specs, as well as some documents which are advisory to library authors (but not specifications, per se):

• design/... -- gathers all such documents
• design/history/... -- gathers research work and pre-spec content and notes
• design/libraries/... -- gathers recommendations for library authors

These documents may be useful to read for those who want to participate more deeply in the design and specification processes (as well as implementation processes!) of IPLD.

Contributing & Discussion

Suggestions, contributions, criticisms are welcome.

Discussion of specifications happens in this repository's issues or via pull request. Discussion of IPLD more generally happens in the IPLD repository.

Check out our contributing document for more information on how we work, and about contributing in general. Please be aware that all interactions related to IPLD are subject to the IPFS Code of Conduct.

Governance

All changes to documents must take place via pull request.

Pull requests are governed by different rules depending on the document type and status of that document:

Specifications:

• Exploratory Stage
  • Authors can merge changes into exploratory specifications at their own discretion
  • Exploratory specifications live in the design directory until they reach the draft stage. Specs names should include an alternative friendly naming convention (-A, -B, etc) while in this stage.
• Draft Stage
  • Authors must attempt to reach a consensus between all active participants before a merge
  • If no objections are raised in a 48 hours period changes can be merged
  • If objections cannot be resolved the change can be voted on by the IPLD Team
• Final Stage
  • Improvements that have a consensus can be merged
  • Changes to behavior should not be merged unless absolutely necessary and agreed upon by a vote of the IPLD Team

Concepts and other documents (including README.md):

• Authors must attempt to reach a consensus between all active participants before a merge
• If no objections are raised in a 48 hours period changes can be merged
• If objections cannot be resolved the change can be voted on by the IPLD Team

Glossary

• DAG: Short for "Directed Acyclic Graph." It's a tree where two branches can point to the same sub-branch, but only in one direction so there's no possibility of recursion.

IPLD Team

The IPLD Team consists of currently active IPLD developers.

• @mikeal
• @vmx
• @warpfork
• @rvagg
• @ribasushi
• @mvdan

License

This repository is only for documents. All of these are licensed under the CC-BY-SA 3.0 license, © 2016 Protocol Labs Inc.