This is an issue I noticed while using @ainghazal openvpn experiment. Basically, the way in which we serialize options so far only accounts for when they're provided with -O key=value from the command line. However, the proper thing to do would be to always serialize richer input options directly from the richer input target into the measurement. Incidentally, this enables performing additional transformations while loading richer input, which is a nice-to-have property, as suggested by @ainghazal in a private conversation. When going back and looking at the richer input code, I noticed that I had already flagged this as a TODO for later, but it seems important to do now to submit a meaningful measurement for openvpn.
This is an issue I noticed while using @ainghazal
openvpnexperiment. Basically, the way in which we serialize options so far only accounts for when they're provided with-O key=valuefrom the command line. However, the proper thing to do would be to always serialize richer input options directly from the richer input target into the measurement. Incidentally, this enables performing additional transformations while loading richer input, which is a nice-to-have property, as suggested by @ainghazal in a private conversation. When going back and looking at the richer input code, I noticed that I had already flagged this as a TODO for later, but it seems important to do now to submit a meaningful measurement foropenvpn.