Skip to content

Latest commit

 

History

History
90 lines (59 loc) · 3.55 KB

iep-0008.md

File metadata and controls

90 lines (59 loc) · 3.55 KB
Raw

ISCC-UNIT Data-Code

IEP: 0008
Title: ISCC-UNIT Data-Code
Author: Titusz Pan tp@iscc.io
Comments: #13
Status: DRAFT
Type: Core
License: CC-BY-4.0
Created: {{ git_creation_date_localized }}
Updated: {{ git_revision_date_localized }}

!!! note

This document is a **DRAFT** contributed as input to 
[ISO TC 46/SC 9/WG 18](https://www.iso.org/committee/48836.html). The final version is 
published under [ISO 24138:2024](https://www.iso.org/standard/77899.html).

General

  1. The Data-Code shall be a similarity hash for any kind of data regardless of its media type.
  2. The Data-Code shall cluster digital assets that have near-identical data.
  3. Small differences (as a proportion of the whole) in referent data shall yield identical Data-Codes.
  4. More significant differences in referent data shall produce similar Data-Codes that can be compared against each other to estimate the data-similarity of the referents.
  5. The Data-Code shall be resistant to data shifting and reordering sequences of data within referent data.

!!! note "NOTE"

Changes of the Data-Code do not reflect semantic or syntactic changes of the content.

Format

The Data-Code shall have the data format illustrated in Figure 10:

![Figure 10 - Data format of the Data-Code](images/iscc-iep-0008-f10-data-code.png) Figure 10 - Data format of the Data-Code

!!! example "EXAMPLE 1: 64-bit Data-Code in its canonical form:"

ISCC:GAAWAIBQLNWP7X32

!!! example "EXAMPLE 2: 256-bit Data-Code in its canonical form:"

ISCC:GADWAIBQLNWP7X32J3INMAMDUJ4QMN67BBQKVTVZIWHXQ7QJIKHYTBY

Inputs

The input for calculating the Data-Code shall be the bytes of a file, without reference to their meaning or structure.

Outputs

Data-Code processing shall generate the following output elements:

  • iscc: the Data-Code in its canonical form (required).

Processing

An ISCC processor shall calculate the Data-Code as follows:

  1. Split the data into variable sized chunks with an average chunk size of 1024 bytes using the content defined chunking (CDC) algorithm.
  2. Calculate the 32-bit integer hash of each chunk using the XXH32 algorithm.
  3. Apply the minhash algorithm to the array of 32-bit integers to calculate the ISCC-BODY of the Data-Code with appropriate length.

!!! note "NOTE"

For further technical details see source-code in modules 
[code_data.py](https://github.com/iscc/iscc-core/blob/main/iscc_core/code_data.py)
and [minhash.py](https://github.com/iscc/iscc-core/blob/main/iscc_core/minhash.py) of the 
[reference implementation](https://github.com/iscc/iscc-core).

Conformance

An implementation of the Data-Code algorithm shall be regarded as conforming to the standard if it creates the same Data-Code as the reference implementation for the same data input.

!!! note "NOTE"

The ISCC reference implementation uses the open source XXHASH library [^1] for XXH32 chunk 
hashing and appropriate use of this software will generate the same codes as the 
reference implementation.

Bibliography