Thou shalt have no other gods before the ANSI C standard 1659

Randy Howard

That argument has obvious holes. Suppose DJB had suggested using int256; you could have made the same response, and it would have been obviously bogus.

As for int64, I'm not convinced the danger is as near as you suggest. The difference between 2^16 and 2^32 is much smaller than the difference between 2^32 and 2^64.

I understand all those reasons, and those are fine and understandable arguments. But the argument you originally raised is that int64 may be insufficient soon. I find the latter argument a little questionable. I suspect you're on safer ground if you stick to the former arguments.

Try the math. I'm serious.

Given current memory latency figures, how long would it take to write 2^64 bytes to memory today? What's the exponential growth rate for memory latency (the "Moore's law" for memory)? Given that growth rate, how many decades until you're able to write 2^64 bytes to memory in a reasonable amount of time?

I'm giving you a hard time because I have a suspicion that you have not done the math, and I have a suspicion that if you were to do the math you might be surprised enough to retract the argument about int64 being inadequate for memory offsets for the foreseeable future.