285800+ entries in 0.174s
phf: mircea_popescu: it's actually pretty immaterial, public key crypto doesn't buy you anything in
this case
phf: each slave vm gets a random int id and a symmetric key, master store int<->symmetric key in a binary
tree. packet comes in <int><sig>, you get
the symmetric key for int and use it
to verify sig
mircea_popescu: hard
to do when
there's nobody with a clue in "the
team", obv.
mircea_popescu: of course, if you read
the java-esque nightmare
that "solidity" is, reimplementing eth as a small emu
thing would benefit it immensely.
mircea_popescu: reimplementing bitcoin is not
the end of
the world anyway.
the extant version is crap.
phf: otherwise you have
to read
the whole
thing before you can make a decision whether or not whole
thing is bogus
phf: in order
to do
that you need crypto where each individual byte in a stream carries enough information
to make a decision
there and
then. so you send a byte packet and have a 1/256 change of going
to next byte
mircea_popescu: you want contact
to my uci interface, we agree in advance, and i whitelist your key.
phf: but
that's unimplementable constraint. anon rando can send you a well crafted pubkey
that is bogus, and you will only know it's bogus after you finished your read
phf: well,
that's going back
to your "not a single byte" unless crypto
phf: asciilifeform: same way as you do with wot folks. you state who you are,
then prove it with a sig
phf: "but own key for each vm
to vm pairing"
phf: you will have
that problem with any form of crypto. if your binary is compromised it's compromised (ignoring fancy
techniques of avoiding compromise)
phf: but own key for each vm
to vm pairing
phf: asciilifeform:
that solution doesn't preclude fancy metering
techniques, wot members own control hubs,
talk
to each other gossip style, and route
tasks accordingly
mircea_popescu: fundamentally uci is a lateral pill for
the hardware problem, oddly enough.
mircea_popescu: main reason hardware is so centralizing
today is exactly lack of
this sort of marketplace,
mircea_popescu: "is your dec alpha/toaster/bitcoin miner pluggable into uci ? cool,
then it makes money. no ? it's broken fix it"
phf: then write your lisps or c compilers
that would
target
the arch
phf: so a virtual machine with constraints, "jump
to here, run at most 100
ticks, claim at most 100 heap"
that's generated for
the
target with
target platform and control key baked in during generation
mircea_popescu: the whole point of uci is for
there
to be no more platform in
that sense. course,
this seems an unlikely goal.
mircea_popescu: i'm not saying
that we are, but WHY NOT must be stated, and in such a way 107 iq guy in 2049 reading
this understands exactly wtf.
mircea_popescu: ben_vulpes
that
there's no specific check or
test other
than
top | grep "ram" sorta
thing.
mircea_popescu: if we do a "trust but verify" no agreement all code, we end up strongly decentralizing, but exposed
to you know,
thedao.
mircea_popescu: anyway,
this is a political choice mostly. if we do a "gentleman's agreement" and no code sorta
thing, we end up strongly centralizing, army of lovers.
ben_vulpes: asciilifeform: aok.
then not 'any key in wot' and bill afterwards
mircea_popescu: at
the other end of
the spectrum, it can just be open bash and well, don't
take code from people you don't
trust.
ben_vulpes: i imagined
that each installed worker process would listen for orders from only one key, which'd handle billing.
ben_vulpes: so
then worker process would have
to get involved in
the billing loop?
mircea_popescu: not
the end of
the world, 1700 business orders carried expiration.
mircea_popescu: asciilifeform if we do it like
that will have
to have expiration.
ben_vulpes: if
the lubified
turd came from
the supervisor process C could come along with
the 'script'
ben_vulpes: oh a buildroot
that compiles a 'linux'
to
the nintendo emu arch?
mircea_popescu: no i know, it's not on you ; just,
the matter has
to be discuss't.
mircea_popescu: silk road hurt
the shit out of itself with "their own
tools"
mircea_popescu: they know more holes in "their own
tools"
than you do.
ben_vulpes: but
the fun of hurting
them with
their own
tools
mircea_popescu: bear in mind, we're not selling investors on
this. cheap, broken, angry sample
to iterate on is jus' fine.
mircea_popescu: even if you just run
top and count ram/cpu it's close enough
mircea_popescu: anyway. uci is not gossipd.
there is a good reason
to implement gossipd OVER uci, but on
the uci layer
the
thing shouldn't be concerned with
this kind of security i don't
thinl
mircea_popescu: asciilifeform
the
tor model is, other
than its many implementation warts, fundamentally broken i
think.
mircea_popescu: asciilifeform i'm
thinking a "independent nodes" sort of botnet, as opposed
to
the zombie windowze
thing usually run
mircea_popescu: not entirely sure bash it is
the right way here ; but it does cheaply (in cognitive
terms) expose
the system for usage. which is why ssh runs into a bash shell.
mircea_popescu: ben_vulpes ideally it'd be more like "run
this bash script"