On 2020-04-26 23:32:38 +0200, Gero Treuner wrote: > Hi Vincent, > > On Sun, Apr 26, 2020 at 10:51:51PM +0200, Vincent Lefevre wrote: > > On 2020-04-26 02:33:00 +0200, Gero Treuner wrote: > > > The MessageId still starts with the time, but is now included in the > > > base64 part, joined with the random section before encoding. > > > > Why is the time needed? Since you use a CSPRNG, you can just use > > random data for the full local part of the Message-Id. > > As no acceptance criterias are defined, "need" still is under > discussion. > > With which option do you feel more comfortable (so obviously I made the > choice for myself): > > (1) Probability that the same MessageId will be generated is practically > zero, but a very very small amount above zero. > > (2) Probability that the same MessageId will be generated is practically > zero, but a very very small amount above zero, and starting with the > next second really is zero (unless somebody intentionally > replicates a MessageId, which is easy as we know).
I prefer (3): In any case, the probability is even smaller than the one in (1). The probability would be small enough that one would not care whether this is this probability or "really 0". This would also protect better against attacks where the attacker knows the time at which some mail will be sent, e.g. concerning mail sent everyday at a fixed time. > Is this the time to start a poll, and also about the length of the > random part? With only random data, 144 bits would give 24 characters in base64, which is the equivalent of what Mutt can currently generate. I think that this would be completely safe in practice. -- Vincent Lefèvre <vinc...@vinc17.net> - Web: <https://www.vinc17.net/> 100% accessible validated (X)HTML - Blog: <https://www.vinc17.net/blog/> Work: CR INRIA - computer arithmetic / AriC project (LIP, ENS-Lyon)
