Implement proper DAG-CBOR (DRISL-CBOR) encoding/decoding layer #20

New issue

Closed

opened 2026-04-12 17:28:44 +00:00 by Grandiras · 1 comment

Grandiras commented

2026-04-12 17:28:44 +00:00

Owner

Summary

ATProto.NET has a CAR file reader, but lacks proper DAG-CBOR (DRISL-CBOR) encoding/decoding for the AT Protocol data model. The spec requires deterministic CBOR encoding for signing/verification and proper round-tripping between JSON and CBOR representations.

Spec Reference

From the Data Model specification:

Required capabilities

DRISL-CBOR Encoding - Deterministic encoding following DRISL rules:
- Map keys must be sorted by byte value
- No floating point numbers allowed
- CIDs encoded with CBOR tag 42
- Specific integer encoding rules (canonical lengths)
DRISL-CBOR Decoding - Parse and validate CBOR data against AT Protocol rules
JSON ↔ CBOR Round-tripping - Lossless conversion between representations:
- $link objects ↔ CID tag 42
- $bytes objects ↔ CBOR byte strings
- $type discriminator preservation
CID Computation - Generate CIDs from CBOR-encoded data:
- CID v1 with SHA-256 hash
- DRISL codec (0x71) for data nodes
- Raw codec (0x55) for blobs
Data Model Validation - Validate data against abstract AT Protocol model:
- No floats
- Integer 64-bit signed range
- Valid $bytes base64 encoding
- Valid $link CID format
- Reserved $ prefix field handling

Current state

The project uses System.Formats.Cbor for basic CAR reading, but doesn't have a proper AT Protocol data model layer that handles deterministic encoding, CID computation, or JSON↔CBOR conversion.

Why this matters

Required for repository commit signing and verification
Required for MST implementation
Required for firehose data processing/verification
Essential for PDS implementation

## Summary ATProto.NET has a CAR file reader, but lacks proper DAG-CBOR (DRISL-CBOR) encoding/decoding for the AT Protocol data model. The spec requires deterministic CBOR encoding for signing/verification and proper round-tripping between JSON and CBOR representations. ## Spec Reference From the [Data Model specification](https://atproto.com/specs/data-model): ### Required capabilities 1. **DRISL-CBOR Encoding** - Deterministic encoding following DRISL rules: - Map keys must be sorted by byte value - No floating point numbers allowed - CIDs encoded with CBOR tag 42 - Specific integer encoding rules (canonical lengths) 2. **DRISL-CBOR Decoding** - Parse and validate CBOR data against AT Protocol rules 3. **JSON ↔ CBOR Round-tripping** - Lossless conversion between representations: - `$link` objects ↔ CID tag 42 - `$bytes` objects ↔ CBOR byte strings - `$type` discriminator preservation 4. **CID Computation** - Generate CIDs from CBOR-encoded data: - CID v1 with SHA-256 hash - DRISL codec (0x71) for data nodes - Raw codec (0x55) for blobs 5. **Data Model Validation** - Validate data against abstract AT Protocol model: - No floats - Integer 64-bit signed range - Valid `$bytes` base64 encoding - Valid `$link` CID format - Reserved `$` prefix field handling ### Current state The project uses `System.Formats.Cbor` for basic CAR reading, but doesn't have a proper AT Protocol data model layer that handles deterministic encoding, CID computation, or JSON↔CBOR conversion. ### Why this matters - Required for repository commit signing and verification - Required for MST implementation - Required for firehose data processing/verification - Essential for PDS implementation

Grandiras referenced this issue from a commit

2026-04-12 17:54:56 +00:00

feat: DAG-CBOR encoding/decoding and CID computation

Grandiras closed this issue

2026-04-12 17:54:56 +00:00

Grandiras referenced this issue from a commit

2026-04-12 17:54:56 +00:00

docs: add changelog entries for issues #20, #23, #30

Grandiras commented

2026-04-12 17:55:17 +00:00

Author

Owner

Implemented in commit bc2c651. Added DagCborEncoder, DagCborDecoder, and CidComputation with full DRISL-CBOR spec compliance: deterministic encoding, CID tag 42, sorted keys, no floats, Base32Lower CID strings.

Implemented in commit bc2c651. Added DagCborEncoder, DagCborDecoder, and CidComputation with full DRISL-CBOR spec compliance: deterministic encoding, CID tag 42, sorted keys, no floats, Base32Lower CID strings.