326100+ entries in 0.197s
mircea_popescu: which is why bitcoin protocol discussions are so complex, and
the whole
thing so hard : because omfg,
the complexity. you don't even rightly know where
the problem comes from really, in
the rat's nest.
mircea_popescu: the problem ~MAY~ be
that bitcoin
transactions are dupe-able in
the first place, for instance.\
☟︎ mircea_popescu: anyway, back
to
the issue. having nodes discard dupes is in principle stupid ; having nodes discard dupes on a first-seen basis is entirely warantless ;
trying
to run an implementation of
this design will run into problems.
mircea_popescu: war is
the
thing
that happens when idiots fucked up
the design.
mircea_popescu: <asciilifeform>
the net-as-publictoilet era will e << and
this is exactly
the situation we have now, except we don't like
the miners.
BingoBoingo: <mircea_popescu> BingoBoingo what's
the cannonical article re pres. bahamas, witgh
the pic ?
the asean one ? << yes
mircea_popescu: how
this is supposed
to be an argument against sane design i have no idea, but whatevs, "kludge works for me" is how we get kludges ever since kludge one.
mircea_popescu: asciilifeform i suppose his idea is, in plainer english,
that in principle fee is no deterrent, because if he makes 1mn
txn
that are all valid but spend
the same inputs, only one can be eventgually mined and so he can create as many
txn as he wants for you
to relay and only pay
the fee once, eventually. whereas only keeping
the first seen one protects from
this.
jurov: yes
this needs some examples and i don't have it
mircea_popescu: but
the important part
there is making
the complete list, like with any other scientific approach.
jurov: in many cases.
then alf or someone other explains again and you get it
mircea_popescu: in all cases when what you're
trying
to explain makes no sense or generally ?
jurov: overall, i feel i can't explain myself well
to you no matter what i
try
mircea_popescu: BingoBoingo what's
the cannonical article re pres. bahamas, witgh
the pic ?
the asean one ?
mircea_popescu: but
this aside. so you said nat routers are stupid and i said i'm happy with
them and
this evnetually means i came
to
terms
to what, using
the jurov
two way router ? or what ?
mircea_popescu: i
think i had moneys-something in
there for a while also, or whatever.
mircea_popescu: i have no fucking idea, it's a random string. it certainly didn't have some sort of intention, and i was certainly not
thinking of you.
jurov: then we had dicsussion where I was unable
to explain why addresses aren't spent, but outputs are, same reult
jurov: i explained why NAT is stupid, like one way
telephone
that only accepts calls, you laughed at me and
told you wouldn't want any other
jurov: eh why do i strive here, like with NAT, inputs vs. addresses and maybe several other
things, you will eventually come
to
terms with what i meant
jurov: we don't have anything better
than cartesian joins?
mircea_popescu: EVEN if you build a
tree - which you do not, chiefly because don't know how - you STILL lose out on
the rebalancings
jurov: mircea_popescu of course. please reread and
think
jurov: oh and actually, how do you check if enemy sends
the same
tx many
times?
jurov: and also loses
txfee when one of
these
tx is mined
jurov: they must make all
these zillion conflicting signatures
mircea_popescu: producing
the equivalent of penises with a clitoral hood and cunts with a bit of penile foreskin and on and on.
mircea_popescu: dumbass doing
the original ~prototype~ had nfi of design, ended up shoehorning everything in everything else
jurov: either
there's a state machine
that considers other
txen in mempool, or no conflict resolution,
there's no
third way
mircea_popescu: asciilifeform> but 'seen first is
The One' is asinine. << srsly.
jurov: but
then you have
to check for conflicts!
jurov: and you reacted by putting forward no less retarded notion
to keep everythign
that has inputs in block
jurov: i don't argue
to keep current algo!
jurov: but ALL
the conflicts have same $maxint fee
jurov: your memory
together for all
the conficted
txs falls below
the specified fee/byte mark, you can admit at least
this is
true
jurov: letting memory
to fill with
turds isn't lame?
jurov: that works great,
too
mircea_popescu: and
the idiots
that made a state machine out of
the fucking relays are so stupid it's basically inhuman.
mircea_popescu: the overarching design principle being
that ALL
the information deciding
the
treatment of a
txn MUST BE ~COMPLETELY~ in
the
txn in question.
jurov: so, i'll spend 1 btc million
times with 0.1 fee, feed it
to you and you'll be like "not my problem"?
mircea_popescu: certainly ANY relay
that currently drops on
the floor a VALIDLY SIGNED
txn is broken. no possible argument here.
mircea_popescu: there is absolutely nothing wrong relaying multiple
txn
that spend
the same inputs ;
there is EVERYTHING wrong with
the notion you shouldn't do
this.
jurov: the realy
that allows
to fillt is mempool by conflicting
transactions is indeed defective
mircea_popescu: "is it signed validly ? are
the inputs in blocks ?" and
that is IT.
mircea_popescu: but it is absolutely not
the relay's fucking job
to attempt
to otherwise "check"
txns
jurov: if
there same and lower, jsut drop it
mircea_popescu: jurov
that's
the end result, once proper ring buffer sorts by fee yes.
jurov: and
the rule is simple -
the 'double' is
the one with lower
txfee.
mircea_popescu: "close and wipe
this. how
the fuck did it end up in ?" x6.
mircea_popescu: the notion
that it works any other way is so fucktarfded i couldn't begin
to explain it
jurov: so i can generate N doublespends with high
tx fee and you're happily gonna carryy
them in mempool and propagate?
jurov: you have
to "walk" mempool for every
tx regardless
assbot: Logged on 04-03-2016 19:28:45; jurov:
to compare, implementing mircea's ring buffer with both random insertion and low memory overhead, now
that's some real algorithmic complication on C machine