I’m struggling to understand how a 1-bit hash error can get irreversibly incorporated into CT, while all the blockchains of the world hum along happily. groups.google.com/a/chromium.org…
The problem here is not that a hash can be corrupted, because that happens. The problem is that somehow the totally “breaks” the CT log? Seems like an avoidable design error. But it’s early and I’m still drinking my coffee.
Anyway, it seems to me that every cryptographic system should be built with the assumption that something (memory, network, 56K phone modem) will introduce errors, and the system will detect those errors — but not by exploding.
I wonder if anyone calculated the mean-time-to-hash-error before deploying CT. Using conservative assumptions about deployment of ECC RAM, etc. Are we way outside the expected range now?
In any case: someone down in comments points out that log failure is supposed to happen, and that there are lots of logs so this is NBD. I’m open to that perspective — but I’ll be honest it still makes me nervous.
• • •
Missing some Tweet in this thread? You can try to
force a refresh
