Packet Encoding & Compression

ELI5

After you know what goes into a packet (STRATT-001), you need to know how to squish it down for transmission. Encoding is the process of turning the structured data into a compact byte stream. Compression makes it even smaller.

Technical Deep Dive

Encoding Pipeline

flowchart LR
    A["Unit\n(JSON)"] -->|"~50% smaller"| B["CBOR\nbinary"]
    B -->|"30–70% smaller"| C["Deflate\ncompressed"]
    C -->|"integrity hash"| D["Blake3\nfingerprint"]
    D -->|"wrap"| E["STRATT\nEnvelope"]

Each stage has a distinct responsibility — serialisation, compression, integrity, and framing — so they can be swapped independently. The Blake3 step hashes the compressed payload rather than the original JSON; this locks the fingerprint to the exact bytes on the wire, making tampering detectable even if the decompressed content looks intact.

1. CBOR (Concise Binary Object Representation): ~50% smaller than JSON
2. Deflate (zlib): Additional 30-70% compression for text-heavy payloads
3. Blake3: Integrity verification of the compressed payload

Compression Strategy

Error Handling

Encoding failures surface as STRATT_ENCODE_ERROR with subcodes:

• E1001: Schema validation failure
• E1002: Compression buffer overflow
• E1003: Fingerprint mismatch (tampering detected)

Key Terms

• CBOR: Binary JSON alternative. Smaller, faster, type-preserving.
• Deflate: LZ77 + Huffman coding compression.
• Streaming: Processing large payloads in chunks without loading entirely into memory.

Q&A

Q: Why not just use Protocol Buffers? A: CBOR is self-describing and doesn’t require a schema at decode time. STRATT values flexibility over absolute minimal size.

Examples

Before encoding:

{ "src": "council-1", "dst": "council-2", "body": "Hello world" }

After CBOR + Deflate:

0x1f 0x8b 0x08 0x00 ... (14 bytes)