nx_tls13_loopback_test.nx

nx_tls13_loopback_test.nx -- SOVEREIGN end-to-end handshake proof. One main() runs BOTH halves of a TLS 1.3 1-RTT handshake against each other. No sockets, no external server, no Cloudflare, no anything outside our own bytes. This is the substrate's self-test that the entire client+server stack works END-TO-END with bit-identical keys derived from real ECDHE. Per user 2026-05-16: "i dont want anything but sovereign". The previous suggestion to test against cloudflare.com was sloppy and has been pulled back; THIS is the canonical proof. Inputs (RFC 8448 §3 for cross-validation against the spec): client_priv = 49af42ba7f9994852d713ef2784bcbcaa7911de26adc5642cb634540e7ea5005 server_priv = b1580eeadf6dd589b8ef4f2d5652578cc810e9980191ec8d058308cea216a21e Expected: ECDHE shared (both sides) = 8bd4054fb55b9d63fdfbacf9f04b9f0d35e6d63f537563efd46272900f89492d handshake_secret (both) = 1dc826e93606aa6fdc0aadc12f741b01046aa6b99f691ed221a9f0ca043fbeac client_hs_traffic_secret = b3eddb126e067f35a780b3abf45e2d8f3b1a950738f52e9600746a0e27a55a21 server_hs_traffic_secret = b67b7d690cc16c4e75e54213cb2d37b4e9c912bcded9105d42befd59d391ad38 (Note: chts/shts above match the RFC ONLY for the post-SH transcript hash 860c06ed...cad8. Our test uses a different SNI ("example.com") and a different client_random + server_random + cipher_suites order, so OUR transcript hash differs, so OUR chts/shts differ from the RFC values. What we CAN verify is that both sides derive the SAME chts/shts/keys, and that both reach CONNECTED with each other's MACs.) expect_exit: 0 license_tier: ORIGINAL