437900+ entries in 0.302s
decimation: yeah
that's kinda what I'm hinting at I guess
phf: mucking around with
the code
that desides what blocks
to send
to a requesting client
phf: asciilifeform: it's not a permanent stuck, but a slowdown. i haven't sent out
that orphanage graph
that i posted some
time ago, because i'm still kicking shit around, but
the beahior
that you can see from it, is
that blocks are sent out as multiple subchains. when a subchain arrives
that's missing a parent subchain, it gets rejected many
times over and over, until parent subchain is filled in. i
think
the behavior can be improved by
decimation: jurov, what is 'N'?
the length of
the cache?
jurov: only needs wallet
that creates N copies of
the
tx with N addresses
jurov: it's not hard for every node
to advertise addy for caching fees
☟︎ decimation: yes, but you would agree
that building such a network with an eye
toward minimizingn latency everywhere would be expensive
☟︎ decimation: I suppose my point (central clearinghouse) is
that I foresee some entity like akamai
that will 'magic away'
this problem for miners, who will all pay for
the service
phf: re: orphanage, i'm still investigating, but
there's no reason why we can't have a better initial sync block handoff strategy,
that doesn't get stock, because some parent in an orphanage subchain failed
to get sent out
decimation: mempool in
this case being
the gigantic c++ 'map'
that holds
transactions in memory?
decimation: but someone still needs
to hold
transactions until
they 'clear'
decimation: umm, didn't you just complain about
the orphan problem?
decimation: satoshi was being dumb when he failed
to program payments for caching
decimation: asciilifeform: I predict
the bitcoin version will be resolved with some kind of central
transaction clearhouse monopoly, sadly
decimation: similarly,
the 'ddos attackers' end up driving everyone
to akamai, etc
decimation: the pre-digital version was expecting some kind of handout/human attenion being paid
to a stranger
mircea_popescu: the process ensures everyone has
to swim in
the same pisspool.
mircea_popescu: decimation riddle me
this : you don't like reading dumb "newspapers", but how many good ones were sunk by exactly
this before you heard of
them ?
decimation: 'web of
trust' in
this case being
the poorly done implmentations of smtp routing
decimation: for instance, spam drives everyone
to big isp mail hosts, gmail, etc - not only for spam protection but also for 'web of
trust'
decimation: no, I would argue
that
the benefits accrue
to specific corporations
decimation: asciilifeform: in
this way you can see how
the internet 'as is' is doomed
to some degree
assbot: Logged on 19-07-2015 18:34:53; mod6: but now I'm scared
that even if i /do/ remove
them by hand,
they might get accidentially pruned by a downstream patch (later in
time) causing
the makefile
to puke.
mircea_popescu: apparently either
there's none or everyone else in bitcoin is just watching
tv.
assbot: Logged on 19-07-2015 18:26:59; asciilifeform: i've been waiting
to hear somebody describe ~some~, even very
theoretical, down side for
those
decimation: you did force
the attacker
to lay down one card
mircea_popescu: anyway, wanted
to see i can first
take down random sites with it
decimation: I see ip has here
to stay forever, because of
the $$ bil it would cost
to do otherwise
decimation: my earlier proposal of 'charging per packet routed' is a jungle way of implementing
this
decimation: yeah I was more interested in
the argument
that generated it
decimation: yeah I didn't recall
the encrypted part being mentioned before
phf: and you want it
to be fully stateless?
decimation: as I recall his objection is
that he didn't want
third parties
to be able
to easily glean identites from packets
decimation: back in
the day, ham radio folks used
to run something much like fidonet except completely over shortwave/vhf radio
decimation: I
think ascii is proposing something like
tor except everyone knows everyone else, and routes accordingly
jurov: ah
that i wanted
to know
jurov: how does
the gate look? like
tor enter node?
jurov: but if you get gigawolves via ip arriginv, how are you getting sheep
through?
jurov: what if
there are million wolves and you must check
their sigs?
decimation: jurov: why does qntra need more
than 1 gigabit
decimation: which is roughly
the position we are all in, failing
to have $$ mil networks
phf: build FIDO on
top of gossipd ;)
jurov: dear decimation
these ddoses have
tens or hundreds gigabits... and not with 1500 byt packets
☟︎ jurov: we can do asics
that easily do 1e6*1e5 operations per second ?
decimation: probably could go
to opencores right now and find a crypto block
jurov: any asic
that comes even close
to
that? ecdsa/2048RSA?
jurov: they need
to be checked at places where do 1e6 pkt/sec flow
jurov: but if you check sigs at destination,
then you haven't solved ddos at all
decimation: asciilifeform: how could you route if
the destination were encrypted?
decimation: jurov:
to be fair, cisco does
the same
thing (custom logic in router)
decimation: jurov: yeah it would need
to be done in custom logic
jurov: asciilifeform: how would
that cheap 1e6/sec sig verificating router look?