{"format":"vess/1","version":1,"threeRecords":{"jsonBytes":314,"jsonPlusTagBytes":330,"vessColdBytes":197,"vessWarmBytes":155,"savingVsJsonPct":37.3,"savingWarmVsJsonPct":50.6,"shapeReusedOnSecondFrame":true},"stream1000":{"jsonBytes":98623,"vessWarmBytes":38026,"jsonBytesPerRecord":98.6,"vessBytesPerRecord":38,"savingPct":61.4},"throughput1000":{"vessEncodeMs":6.95,"vessDecodeMs":3.21,"vessEncodeUsPerRecord":6.95,"vessDecodeUsPerRecord":3.21,"jsonStringifyMs":0.37,"jsonParseMs":0.51,"jsonEncodeUsPerRecord":0.37,"jsonDecodeUsPerRecord":0.51,"note":"CPU is NOT the win: JSON's native stringify is faster per record than this implementation. The win is bytes on the wire."},"security":{"tamperedFrameRejected":true,"wrongKeyRejected":true,"replayedCounterRejected":true,"frameWithUnknownShapeRejected":true},"opacity":{"longestPrintableAsciiRunMac":25,"longestPrintableAsciiRunAead":4,"note":"MAC mode still shows string VALUES (they are data). AEAD mode shows almost nothing, because it is encrypted. Either way, opacity — not security."},"roundTrip":{"exactValues":true,"records":3,"sampleDecoded":{"delivered":true,"handle":"l33t0bot","kind":3,"note":"postage settled","score":0.8125,"seq":1042}},"honest":["This is a binary serialization grammar. Binary serialization is a known family; the specific decisions here (structure/content split, booleans as structure, session-registered shapes, one tag per stream) are the new part — not the existence of bytes.","Unreadability is not security. The printable-run number shows the wire carries no readable label; it says nothing about whether the wire can be decoded.","Security is the HMAC tag (authenticity, tamper and wrong-key rejection) plus the monotonic counter (replay), both tested above. Confidentiality mode uses AES-256-GCM, a standard cipher, deliberately.","It is stateful: shapes live in the session, which is what makes it cheap and also why it is impractical for one-shot document exchange.","CPU is not the win: measured JSON.stringify is faster per record than this implementation. The win is 50-70% fewer bytes on the wire, which is what actually costs time between agents."],"cachedAgoMs":0}