Web Connectivity 0.5 has changed the way in which network_events are sorted. You no longer have the guarantee that all events related to the same transaction are together, which was a wrong assumption made in how the network events related to a particular TLS handshake was re-contructed in find_tls_handshake_network_events (https://github.com/ooni/data/blob/main/oonidata/transforms/nettests/measurement_transformer.py#L294).
It does however have a transaction_id which is actually a much cleaner way of figuring out what network events are related to a particular transaction.
We should refactor the find_tls_handshake_network_events to prefer using the transaction_id based lookup and failover to using the older method only when transaction_id is missing.
Web Connectivity 0.5 has changed the way in which network_events are sorted. You no longer have the guarantee that all events related to the same transaction are together, which was a wrong assumption made in how the network events related to a particular TLS handshake was re-contructed in
find_tls_handshake_network_events(https://github.com/ooni/data/blob/main/oonidata/transforms/nettests/measurement_transformer.py#L294).It does however have a
transaction_idwhich is actually a much cleaner way of figuring out what network events are related to a particular transaction.We should refactor the
find_tls_handshake_network_eventsto prefer using thetransaction_idbased lookup and failover to using the older method only when transaction_id is missing.Here are some sample measurements to use as test cases: Web Connectivity 0.5: https://explorer.ooni.org/m/20230817084207.519697_JO_webconnectivity_b62797f5aef7a7d6
Currently leads to erroneously flagging the last operation as
write_2, while it should instead bewrite_1.Sample measurement without
transaction_id: https://explorer.ooni.org/m/20230901130157.590570_US_webconnectivity_b5534b1c5b446826