[jbarrs]

I find it amusing that the table published a conservative cutoff year and an optimistic cutoff year. Based on the trends I've seen, most non-critical software would probably have made the switch in time for the conservative year, whereas anything security-critical like a bank would probably use the optimistic year.

[upofadown]

The author of the paper had a problem. His elliptic curve method seemed like it might overtake the best known algorithm at the time for factoring. So the conservative estimate takes that into account. The elliptic curve method never managed supremacy so the optimistic estimate is actually more relevant. That means that the actual prediction is 2040 but it seems the various national cybersecurity entities might of missed that point.