The cr.yp.to microblog: 2016.05.05 02:09:34

2016.05.05 02:09:34 (728013734858346497) from Daniel J. Bernstein, replying to "You and 52 others (@bahstgwamt)" (728005415062151168):

"Busy-loop to const time T" is an auditor's nightmare: you'll underestimate T _and_ screw up the loop _and_ miss most time channels. @kragen

Context

2016.05.05 01:31:41 (728004199657689090) from "You and 52 others (@bahstgwamt)", replying to "Dan Kaminsky (@dakami)" (728003954588573696):

are you honestly suggesting that @isislovecruft isn't an operational engineer and she wouldn't spinlock because it isn't her style?

2016.05.05 01:33:06 (728004554613243904) from "You and 52 others (@bahstgwamt)", replying to "You and 52 others (@bahstgwamt)" (728004199657689090):

Or the same suggestion about @agl__? I could maaybe see it if you're talking about @hashbreaker but not really.

2016.05.05 01:35:47 (728005232676872192) from "Dan Kaminsky (@dakami)", replying to "You and 52 others (@bahstgwamt)" (728004554613243904):

a fair question for the primary sources. @agl__ @hashbreaker is establishing a timing floor above variation viable?

2016.05.05 01:36:31 (728005415062151168) from "You and 52 others (@bahstgwamt)", replying to "Dan Kaminsky (@dakami)" (728005232676872192):

I've often wondered why we don't just NOP our way to constant-time, but I assume there's a why.