Clarify hash-addressed vs schema-addressed storage distinction

[dimitri-yatsenko]

Summary

Clarifies the technical distinction between hash-addressed and schema-addressed storage based on their capability to handle different object types.

Changes

Key distinction added:

• Hash-addressed storage (<blob@>, <attach@>, <hash@>): Handles individual/atomic objects only - single files or serialized blobs
• Schema-addressed storage (<npy@>, <object@>): Can handle complex multi-part objects - Zarr arrays, HDF5 datasets, directory structures

Files Modified

src/how-to/use-object-storage.md

• Updated OAS addressing schemes table with new "Object Type" column
• Added explicit key distinction callout
• Fixed terminology: inline → in-table storage

src/reference/specs/type-system.md

• Enhanced <object@> description: "complex, multi-part objects (files, folders, Zarr arrays, HDF5)"
• Enhanced <hash@> description: "individual, atomic objects only"
• Clarified that hash-addressed cannot handle Zarr/HDF5

src/llms-full.txt

• Regenerated to include updated documentation

Technical Background

This distinction is important because:

1. Zarr arrays are directory structures with multiple chunks/metadata files
2. Hash-addressed storage stores single blobs identified by content hash
3. Schema-addressed storage uses hierarchical paths that can contain multiple files
4. Users choosing between <blob@> and <object@> need to understand this limitation

Related

Addresses feedback about clarifying storage capabilities for complex data structures.

🤖 Generated with Claude Code

[dimitri-yatsenko]

Consolidated into #119 - Documentation Cohesion Review: Comprehensive Improvements for DataJoint 2.0