Are Zk-oracles impossible due to TLS / HTTPS lack of a “non-repudiation” property?
TLS was designed for secure communication, and is not conducive to retroactively proving authentic receipt of messages to a third party. If the web server were configured to sign all HTTP responses, it could work, however that would be non-standard.
Also relevant:
DECO basically is trying to solve that exact problem. They developed a ZK oracle that works with the symetric key of TLS. Sadly their 3 way handshake either requires interaction from the prover with the provee or requires a trusted (decentralized possible) setup. They are enabling oracles for private data tho. Surely an very important paper when researching trustless oracles, maybe the 3 party handshake can be improved.