On Thu, Mar 20, 2014 at 3:55 PM, Brad Martin <[email protected]> wrote: > I completely get where you're coming from but your reply is a polemic.
... it read better at the time. in hindsight, i should have waited for a more coherent and concise response, ah well, .. long day. (for all sysadmins? ;) > Everyone in the party agrees to the following points: > * Distributed HWRNG is better than shared HWRNG > * Distributed trust systems are always worse than local trust systems i admit an unstated bias toward including per-host/instance entropy as part of OS installation; you are 100% correct that if network it up early on in init, this is a great (can't hurt!**) addition to entropy on host for any purpose. i should also have stated that i have specific use cases where no network activity is present or possible. funny how past experience limits and shapes future considerations at times. me culpa! in short, i agree that mixing network entropy into a host pool is only going to make your unpredictable numbers "less predictable" is a useful way, every time. > The original concept (shared HWRNG) is aimed at situations where entropy in > local nodes is insufficient - a data center hosting nodes having no local > HWRNG (or, whose local sources are insufficiently trustworthy). Grant us a > suspension of disbelief that such systems exist. indeed, this is far more common than not. (that was another aspect of my grief, we should have had good raw access entropy sources in all of our microprocessors for many years now!) Marsh Ray conveyed a beautiful picture of why this is not so for entirely reasonable reasons, but it is none the less frustrating. i will do a better job of orienting my diatribes toward the right targets in the future, thanks for taking the time to reply to my less than deserving ambiguities. i will write a better feedback with care, best regards, _______________________________________________ RNG mailing list [email protected] http://lists.bitrot.info/mailman/listinfo/rng
