asciilifeform: by making the machine draw same current regardless of what keys are pressed, or if any are
asciilifeform: where they countered differential power analysis
asciilifeform: mircea_popescu: do you perchance have the link to the sov typewriter thread ?
asciilifeform: mircea_popescu: running safety-critical code on pc ~is~ a hack, and there is no way fully around it except to burn the pc.
asciilifeform: the job of the algo is to perform same count of 'ideal cpu, no cache, all instr take the cycle count printed in the book' instructions per sig verif.
asciilifeform: (e.g., i cannot know the rate at which mircea_popescu's reactor control card throws interrupts, or how many cpu core he has to handle them with, etc.)
asciilifeform: mircea_popescu: the ceiling is necessarily machine-specific
asciilifeform: if enemy can make it happen every second, programmer is a tard and ought to be fed to pigs.
asciilifeform: if this happens, interval is bumped up.
asciilifeform: and calculation ~never~ returns a result prior to interval elapsing.
asciilifeform: mircea_popescu: which is why you determine a ceiling (longest conceivable interval, assuming every single ram access is a cache miss, etc.)
asciilifeform: cache, incidentally, is controllable at least on amd.
asciilifeform: but even on pc you can make a hard ceiling.
asciilifeform: mircea_popescu: my proposal included a device for gateway 'line rate crypto', recall.
asciilifeform: mircea_popescu: on embedded cpu, with cache switched off, interrupts off - quite certainly, yes, known cycle count.
asciilifeform: my understanding is based in this exercise having been my profession.
asciilifeform: http://btcbase.org/log/2016-08-26#1529877 << timing can be 'ceilinged' and it solves problem. but i am still chewing on the problem of enemy being able to determine who is speaking to whom by deriving the public keys. (this is trivial with rsa, and i've been working on answering the q of whether is is also true for c-s)☝︎☟︎
asciilifeform: why am i reading a chronicle of some dude cleaning his fridge, BingoBoingo ?
asciilifeform: it is general in the sense where, if you include it, 'sybil' is no longer meaningful concept
asciilifeform: ~everyone here, phf, asciilifeform, mircea_popescu, et al, has his own chamber of unreleased horrors. this is right and proper.
asciilifeform: but they ought not to complain when 'my tcp connections are blackholing' or 'someone derived my rsa privkey using known-ciphertext attacks' etc.
asciilifeform: a great many x11 proggies i use run on remote machines. and i have NO intention of partaking in the 'own one computer' horseshit.
asciilifeform: and regard anyone who advocates it as a procrustes.
asciilifeform: i.e. i have 0 interest in 'like x11 but without network pipe'
asciilifeform: phf: my published positions re x11 concern the machines ~i presently use~
asciilifeform: i dun even recall a debate. there was gabriel_laddel proposing to build hybrid of bulldog and rhino, 'lisp machine with linux', a mengelian atrocity with no future, because a foundation cannot be corrected any more than one can un-drop a baby.
asciilifeform: likewise a gossip node ought never to rely on a single entry point. and certainly not on a single ~published~ entry point.
asciilifeform: and there are no sybils, even as a theoretical item, in a correct gossiptron - every receiver knows exactly who (pubkey-wise) has any business transmitting to it, and rejects packet that is malformed, replayed, or signed with ANY other key, in constant time.
asciilifeform: the cost of validation in a single-packet-authenticating protocol where you crunch the numbers at line speed is effectively 0.
asciilifeform: but it was ~always~ possible, from day1 of tcp, and this is evident to anyone with a copy of, e.g, richard stevens's 'tcp/ip illustrated'.
