5 ms·

The script linked in the gist (https://gist.github.com/bmastenbrook/14c0e22fc02b95d4a48f82d3ec3123db https://gist.github.com/bmastenbrook/14c0e22fc02b95d4a48f82

The script linked in the gist (https://gist.github.com/bmastenbrook/14c0e22fc02b95d4a48f82d3ec3123db https://gist.github.com/bmastenbrook/14c0e22fc02b95d4a48f82d...) is reporting diffs for me on Sonic fiber as well, which is leased from AT&T in much of the Bay Area. So seems to be a network layer or below issue. Interestingly, the incorrect ASCII chars are always 8 chars higher or lower than their counterpart, indicating that bit 3 is consistently being flipped: flip: f|0b1100110 orig: n|0b1101110 --- flip: i|0b1101001 orig: a|0b1100001 --- flip: 6|0b110110 orig: >|0b111110 --- flip: (|0b101000 orig: |0b100000 --- flip: p|0b1110000 orig: x|0b1111000 --- flip: m|0b1101101 orig: e|0b1100101 Edit: Can confirm connecting to VPN (ProtonVPN in this case) solves the issue.

heleninboodler 6y ago
I love this. In 1992, I bought a Kenwood stereo receiver, which I still have in my living room to this day, which you could program your radio station call letters into for your presets, and I noticed that after setting one of the presets to "KNBR" it switched to "KNBB". I assumed I made a mistake and corrected it, but a few days later I noticed it happened again. Assuming ASCII encoding, B is 0x42 and R is 0x52, so the difference is the single bit (and even assuming non-ASCII, the delta of 16 is likely still the same), so I experimented and determined that that particular memory location always eventually zeroed the 4th least significant bit, so all characters from P-Z would be shifted down to the corresponding character 16 positions less than it. Nowadays I don't really use the radio presets anymore, but I still like knowing the flaw is there nearly 30 years later.