Skip to content

TAXONOMY.md

Latest commit

 

History

History
122 lines (95 loc) · 3.44 KB

TAXONOMY.md

File metadata and controls

122 lines (95 loc) · 3.44 KB
Raw

Taxonomy v3

OpenDevIndex separates a module's stable address, semantic classification, domains, and relationships so the catalog can grow into a technology knowledge graph without breaking published links.

Stable address

Every module has a durable <category>/<slug> address, for example:

tool/git
framework/pytorch
ai/tensorflow
protocol/mcp
hardware/ryzen-9-9950x
architecture/x86-64
algorithm/dijkstra
model/llama
format/parquet
dataset/imagenet

Existing addresses are not renamed merely because classification improves. Legacy namespaces such as ai/, security/, cloud/, and opensource/ remain valid.

Canonical kind

kind answers what the subject is. Taxonomy v3 keeps the software kinds from v2 and expands them with technology-wide kinds including:

  • hardware
  • architecture
  • algorithm
  • model
  • format
  • device
  • dataset
  • technique
  • operating-system
  • kernel
  • specification

This lets one stable address remain compatible while semantic classification becomes more precise. For example, opensource/linux-kernel can remain a stable address while being classified as kind: kernel.

Domains

domains answer where the technology belongs or is commonly used. They are multi-valued facets. Taxonomy v3 expands coverage beyond software into areas such as:

domains:
  - computer-architecture
  - hardware
  - operating-systems
  - cryptography
  - identity
  - internet
  - wireless
  - embedded
  - iot
  - robotics
  - quantum-computing
  - supply-chain
  - performance
  - reliability

A subject may belong to several domains without being duplicated. The controlled vocabulary lives in taxonomy/v3.yaml.

Relationships

Schema v3 can connect modules using typed relationships:

relationships:
  - type: implements
    target: architecture/x86-64
  - type: depends-on
    target: protocol/http
  - type: alternative-to
    target: tool/example
  - type: successor-of
    target: technology/example-v1

The controlled relationship vocabulary includes:

  • depends-on
  • implements
  • built-with
  • uses
  • used-by
  • integrates-with
  • compatible-with
  • alternative-to
  • extends
  • part-of
  • based-on
  • successor-of
  • predecessor-of
  • replaces
  • secures
  • related-to

A relationship may include a short note when the edge is not obvious. Relationship targets use stable module references so graph edges remain durable.

Deployment and license metadata

Taxonomy-aware entries may also include deployment and licensing metadata:

license: Apache-2.0
deployment_types:
  - self-hosted
  - service

Taxonomy v3 also supports physical and low-level deployment forms such as hardware, device, firmware, and embedded.

Compatibility

The validator continues to accept schema versions 1 and 2. New modules should use schema version 3 so they can participate fully in the knowledge graph. Taxonomy v2 remains in the repository as a historical compatibility artifact; current tooling defaults to v3.

Why this model

A single hierarchy cannot represent modern technology accurately. A technology can simultaneously be a protocol implementation, security boundary, hardware dependency, learning prerequisite, alternative, historical successor, and part of several domains. Stable addresses plus canonical kinds, multi-valued domains, and typed graph relationships provide compatibility without sacrificing depth or discoverability.