2 ms·

I would love to have some understanding as to how this could happen.

by consumer451 4y ago

olitech 4y ago
I'm thinking it might be related to Bard.
- consumer451 4y ago
  Not sure about that, but it's interesting where the data is coming from. As you said, some of it is coming from old news. For example, I searched for one long phrase in quotes and it only appears in a one place on the English language web. (2003) [0] Another phrase appears in this news article. (2001) [1] None of these things are related to mathematics, and are not nonsense. Does this IPv6 prefix looking string mean anything? [2002:ae0:dd6e::]:3400 [0] https://www.yorkpress.co.uk/news/7943515.o-roy-you-left-me-cry-i-i-i-ing-over-you/ https://www.yorkpress.co.uk/news/7943515.o-roy-you-left-me-c... [1] https://www.theboltonnews.co.uk/news/5895388.no-parking-fee-plans-drivers-told/ https://www.theboltonnews.co.uk/news/5895388.no-parking-fee-...
andromeduck 4y ago
I'm pretty sure this is an example of 'Silent Data Corruption' on something like the record length logic/variable causing it to return the next record in addition to the first. I suspect if you look at the unexpected text, you'll find it'll be pretty close to a power of two or attributable to a single bit error in whatever processor logic that calculated it. https://support.google.com/cloud/answer/10759085?hl=en https://support.google.com/cloud/answer/10759085?hl=en
- consumer451 4y ago
  Thank you. I have been thinking about this since I posted, and the one thing I was sure of is that it was pretty far down the stack. Half halfheartedly, I was guessing it was something like a rare multi cosmic ray bitflip or something. From your link: > Today, many Google Cloud services contain such checks, thus preventing SDC errors from permanently corrupting customer or system data. It's interesting that it can still happen from one bit in a system like Google's storage for search. I guess that would probably be expensive?

I would love to have some understanding as to how this could happen.

olitech 4y ago

I'm thinking it might be related to Bard.

consumer451 4y ago
Not sure about that, but it's interesting where the data is coming from. As you said, some of it is coming from old news. For example, I searched for one long phrase in quotes and it only appears in a one place on the English language web. (2003) [0] Another phrase appears in this news article. (2001) [1] None of these things are related to mathematics, and are not nonsense. Does this IPv6 prefix looking string mean anything? [2002:ae0:dd6e::]:3400 [0] https://www.yorkpress.co.uk/news/7943515.o-roy-you-left-me-cry-i-i-i-ing-over-you/ https://www.yorkpress.co.uk/news/7943515.o-roy-you-left-me-c... [1] https://www.theboltonnews.co.uk/news/5895388.no-parking-fee-plans-drivers-told/ https://www.theboltonnews.co.uk/news/5895388.no-parking-fee-...

andromeduck 4y ago

I'm pretty sure this is an example of 'Silent Data Corruption' on something like the record length logic/variable causing it to return the next record in addition to the first. I suspect if you look at the unexpected text, you'll find it'll be pretty close to a power of two or attributable to a single bit error in whatever processor logic that calculated it. https://support.google.com/cloud/answer/10759085?hl=en https://support.google.com/cloud/answer/10759085?hl=en

consumer451 4y ago
Thank you. I have been thinking about this since I posted, and the one thing I was sure of is that it was pretty far down the stack. Half halfheartedly, I was guessing it was something like a rare multi cosmic ray bitflip or something. From your link: > Today, many Google Cloud services contain such checks, thus preventing SDC errors from permanently corrupting customer or system data. It's interesting that it can still happen from one bit in a system like Google's storage for search. I guess that would probably be expensive?