211 entries in 0.649s
a111: Logged on 2017-09-20 02:01 asciilifeform: in other noose, asciilifeform found that 1) knuthian division can be sped up by ANOTHER factor of 2, by walking the bits of the quotient instead of shifting'em 2) barrett reduction in
constant time is almost certainly possible
a111: Logged on 2017-09-30 18:26 asciilifeform: the continuing lulzy part is how everywhere on the net you will find '
constant time crypto libs are available', 'it's a solved problem', 'this reduction routine is
constant time', and all of it is liquishit and doesn't stand up to 30 seconds of examination with naked eye
a111: Logged on 2017-09-13 17:17 asciilifeform: apeloyee: anything that beats multiply-then-divide is an improvement, so long as it meets the basic demands (
constant time,
constant space, NO branches that depend on input bits, no use of approximations, no massively heavy - 100 loc is a good approx max - code )
a111: Logged on 2017-09-12 23:12 mircea_popescu: and finally 3. the item there described is not exactly a function. it rather something i'd call a mechanism, a discrete item that does a fully defined thing. as we're looking more and more through ada eyes and
constant time things and so on, a study of these mechanisms as an distinct category will prolly be useful. somewhere between conway's cells and commandline utils, they are.
mircea_popescu: and finally 3. the item there described is not exactly a function. it rather something i'd call a mechanism, a discrete item that does a fully defined thing. as we're looking more and more through ada eyes and
constant time things and so on, a study of these mechanisms as an distinct category will prolly be useful. somewhere between conway's cells and commandline utils, they are.
☟︎ mircea_popescu: if you want
constant time, you feed the list 9, 0,0,0,0,8,0,0,1. it will do 18, 1, 18, 1, 18, 1, 18, 1 etc.
mircea_popescu: will necessarily have the modulus of the sum. this entire procedure is
constant time.
mircea_popescu: you write by hand a function which takes a list with a promise none of the items on it exceed a mod, and returns the mod of the sum of the sum of the elements, in
constant time.
mircea_popescu: that small cost can be slightly higher and
constant time.
a111: Logged on 2017-08-15 22:52 mircea_popescu:
constant-
time MPFHF is now an open question for teh interested.
mircea_popescu:
constant-
time MPFHF is now an open question for teh interested.
☟︎ mircea_popescu: the ONE way to
constant-mpfhf is to calculate ALL the tree of possibilities, 2^message length items EVERY
TIME mircea_popescu: asciilifeform amusingly, the guy complains about the modular exponentiation not being
constant time. maybe write to him ask where he ever saw a sane algo ?
mircea_popescu: sina you mean, is there a side channel for
constant time ops ? or for rsa as commonly implemented atm ?
mircea_popescu: this would be automatic-
constant-
time cpu, run any software you want.
sina: if you impl as
constant time you don't need ctgrind
sina: asciilifeform: btw I was going to ask you about your feelings on this
https://github.com/cforler/Ada-Crypto-Library ...obviously hasn't been impl for
constant time/space but regardless. may be possible to ctgrind it using that valgrind patch I linked in the logs
mircea_popescu: basically the other side of the "
constant time" coin is that... YOU GET CONSTANTTIME, BITCH.
a111: Logged on 2017-06-27 00:57 asciilifeform: sina: one of the things gossipd needs is a
constant-
time-
constant-space rsa. if you don't have one, enemy can derive your privkeys remotely based on timing.
sina: ok fair point, I get the general need for
constant time constant space algo regardless of gossipd stuff anyway
mircea_popescu: this then reduces to the case where
constant, space,
time etc
ben_vulpes: do forgive my ignorance, but why must rsa ops run in
constant, worst-case
time?
sina: (Checking that functions are
constant time with Valgrind)
Framedragger: i mean, 'compared to what'. certainly not compared to asciilifeform's actually-fucking-
constant-
time crypto architecture
Framedragger: (okay apparently if your cpu supports 'constant_tsc' (as seen in cpuinfo) then this timestamp counter actually counts
time and not processor ticks which is a *good thing* given freq scaling etc.; this is available in all new intel processors; what a rabbit hole, man)
pete_dushenski: ben_vulpes: beeping and flashing orange dash light every
time traction control activates, which in rwd car with 275 rear section tires on freshly snowed roads is...
constant.
Framedragger: quite sure the diff'ing / updates were thought out thoroughly, i.e.
time complexity is
constant.
Framedragger: asciilifeform: re. "Enemy can spam the channel but each of his packets can be rejected in ~
constant time~" - ahh! that clarifies matters for me. will comment on blog later by PC. ttyl
ben_vulpes: phf: i tried for a very long
time with erc in a seperate emacs instance running on a server as a bouncer, and the *
constant* muscle memory confusion over the subtle differences between emacs-in-terminal and emacs-with-gui drove me up a wall
assbot: Logged on 03-01-2016 03:56:57; BingoBoingo: <pete_dushenski> last
time i tried it was >1gb of ram, so the footprint's coming down. << Ironically my bastart build went from consistent ~300 MB to
constant 1 GB of RAM after BDB fix
BingoBoingo: <pete_dushenski> last
time i tried it was >1gb of ram, so the footprint's coming down. << Ironically my bastart build went from consistent ~300 MB to
constant 1 GB of RAM after BDB fix
☟︎ mircea_popescu: neither my e nor my planck
constant have "declined" over
time.
nubbins`: "The need to assemble is as
constant among humans as the necessity of making decisions is rare. Assembling corresponds to the joy of feeling a common power. Decisions are vital only in emergency situations, where the exercise of democracy is already compromised. The rest of the
time, “the democratic character of decision making” is only a problem for the fanatics of process. It’s not a matter of critiquing assemblies or abandoning them,
Adlai: (this seems to be djb's real reason behind discounting curves without a *convenient*
constant-
time algorithm)
Adlai: re: montgomery ladder, you can implement
constant-
time secp256k1, but it's a pita
adlai: well anyone can factor a number of arbitrary length in arbitrary
time... consistently being able to factor arbitrary length numbers in
constant time, now that is a 'supernatural' algorithm, relative to current knowledge
BingoBoingo: spring operated wrist mounted
time standard is only 1.5 minutes off of ntp reported
time in 8 years of
constant operation.
BingoBoingo: Let
time stay
constant. If planetary rotation slows acknowledge that as the problem.
mircea_popescu: What if the terrorists hear about fast secure crypto? Yikes! Similar to
constant-
time story. Dont standardize good crypto. Discourage use of good crypto. If the good crypto persists, try to bury it behind a huge menu of bad options. Advertise cryptographic agility; actually cryptographic fragility. Pretend that this agility justifies using breakable crypto.
pete_dushenski: asciilifeform: '
time was flowing at a
constant rate of one second per second!'