What the authors propose here is much better. The methodology and protocol for transforming the seed digits is agreed to ("committed") ahead of time, before the digits are known, and only later are the numbers generated.
What the authors propose here is much better. The methodology and protocol for transforming the seed digits is agreed to ("committed") ahead of time, before the digits are known, and only later are the numbers generated.
You don't need lotteries to break this logjam. You just demand the minimal parameters that satisfy the security a/o performance considerations for your problem. LaMacchia and Costello did this for NUMS without even needing recourse to constants: they just start with the target security level, find the minimal prime for that security level, and then the minimal suitable (given, for instance, SafeCurves) d for an Edwards curve.
† My friend Peter found it: http://pi.nersc.gov/cgi-bin/pi.cgi?word=BADA55&format=hex --- so: justify the binary offset "2733181964". :)
For example they could choose Pi and say that the following are the official NUMSN for each order of magnitude: 1, 41, 592, 6535, 89793, 238462, 6433832, ... and if a number is needed with a specific property (e.g. 40 digits and prime) than the nearest number to a NUMSN is the official NUMSN for that purpose.
Then any future standard that used a constant they claimed was a NUMSN but wasn't complaint with RFC 31415: Procedures of Generating Constants would be suspicious.
Pi has not been proven to be normal, so you can't be completely sure that you could always find any particularly long (backdoored) sequence.