ipld

what's up
what's new
what's next

quick recap

IPLD stands for InterPlanetary Linked Data
IPLD grew out of IPFS
The goal is: specs (and libraries) that make developing interoperable, decentralized applications easier and better.

IPLD stands for InterPlanetary Linked Data
IPLD grew out of IPFS
The goal is: specs (and libraries) that make developing interoperable, decentralized applications easier and better.

quick recap

IPLD stands for InterPlanetary Linked Data
IPLD grew out of IPFS
The goal is: specs (and libraries) that make developing interoperable, decentralized applications easier and better.

a million years ago

IPFS is just[*] a thing that stores files.

IPLD started as being just dag-pb
(which stores info in a certain way that was really geared towards files... lots of blobs and links,

structure? not so much...)

[*] - never say "just" about a decentralized application; it's not easy

IPFS is just[*] a thing that stores files.

IPLD started as being just dag-pb
(which stores info in a certain way that was really geared towards files... lots of blobs and links,

structure? not so much...)

a million years ago

IPFS is just[*] a thing that stores files.

IPLD started as being just dag-pb
(which stores info in a certain way that was really geared towards files... lots of blobs and links,

structure? not so much...)

a Data Model was formed

IPLD is about structured data.

IPLD is about structured data.

a Data Model was formed

IPLD is about structured data.

IPLD is about structured data.

what can do with Data Model?

traversal

selectors

advanced layouts

schemas

....

your application

selectors

Selectors are machinery for traversing a tree and "selecting" data that matches.

Create them with just more IPLD!

Built for limited power; can be evaluated in low-trust environments.

(Powers new data transports like Graphsync!)

{
  "description": "Traverse selecting fields",
  "ipld": {
    "bar": false,
    "foo": true,
    "some_other": {}
  },
  "selector": {
    "f": {
      "f>": {
        "foo": {".": {}},
        "bar": {".": {}}
      }
    }
  },
  "expect_visit": [
    {
      "path": "",
      "node": {"map": null},
      "matched": false
    },
    {
      "path": "foo",
      "node": {"bool": true},
      "matched": true
    },
    {
      "path": "bar",
      "node": {"bool": false},
      "matched": true
    }
  ]
}

ADLs

Advanced Data Layouts

tl;dr: Sharding for everyone!

Small plugins that make transform some Data Model (including multiple nodes, or multiple blocks) into other Data Model.

Unixfsv1 had this via magic
(e.g. directories in the web gateway did this for you.)

Now we're bringing it to everyone,

ready to use wherever you need it in your application data structures!

Advanced Data Layouts

tl;dr: Sharding for everyone!

Small plugins that make transform some Data Model (including multiple nodes, or multiple blocks) into other Data Model.

Unixfsv1 had this via magic
(e.g. directories in the web gateway did this for you.)

Now we're bringing it to everyone,

ready to use wherever you need it in your application data structures!

🎉

Selectors, etc, work transparently over this.

🎉

ADLs

Generic Data Model trees are great...
But what happens when we want some more leverage for development?

Everybody who's written a data validator before knows there's a lot of work there. Type systems are great tools for getting the computer to do the work for you!

Generic Data Model trees are great...
But what happens when we want some more leverage for development?

Everybody who's written a data validator before knows there's a lot of work there. Type systems are great tools for getting the computer to do the work for you!

IPLD Schemas!

-> can have structs, unions, enums, etc here (in addition to the maps and lists and strings etc already in the DM)

- Type clauses describe the semantic behaviors

-> can have structs, unions, enums, etc here (in addition to the maps and lists and strings etc already in the DM)

- Representation clauses describe how to transcribe this information into DM (and thence, codecs!)

-> separately tunable!

IPLD Schemas!

- Types also have a Data Model interface.

This is independent of the representation used when wrangling serial data.

-> means: you can still run Selectors, traversals, and other generic code, also against data processed with Schemas!