227400+ entries in 0.128s
davout: but i also get
the other point, a lot of
that complexity becomes apparent once one actually goes ahead and pops
the hood
davout: yet another obvious benefit of amputating
the wallet, miner and everything
that can pretty obviously done without
a111: Logged on 2017-01-03 21:29 asciilifeform:
though what i pictured is
that
trb can finally produce
the motherfucking ~book~ and it will be possible
to start rewrite...
davout:
http://btcbase.org/log/2017-01-03#1595836 <<< my image is more like: "trb is
this
thing from which more and more is removed, until only
the radioactive code consisting in ball of
tightly packed hot wires which we proceed
to put in a little box in which epoxy is poured, and is only interacted with as some black box"
☝︎ lobbes: BingoBoingo: aha. Yeah, I guess more
towards high school. Luckily I was just old enough
to still have been
taught how
to research using actual b00ks and/or libraries.
mircea_popescu: and what i pictured were 72 cubits high,
translucent, ageless, nonmenstruating and deliver pregnancy
to
term within
the day.
mircea_popescu: still,
the "frozen
trb because networking" or "because badlt done block check" etc can't go on forever.
mircea_popescu: still, due
to
the fact
that v allows attribution,
the change can be digested over
time.
mircea_popescu: i suspect graph
theory may have a solution for us, but it is not clear
to me how.
mircea_popescu: davout
the gordian knot is how
to make it both unmagical and self-summarizing.
lobbes: Re: wikipedia. I remember early on in grade school I was actually
taught
that wikipedia was shit. By
the
time I entered college
this
tune had changed. Now it all makes sense
davout: but if it really needs a magical difftron, can it still be said
the operator can see everything with naked eye?
lobbes: That was a great read,
thank you. Lined up exactly with my own anecdotal experiences. If I learned one
thing in 'business school', it was how
to properly bullshit. Only cost some 20k! Real learning didn't happen until after graduating (funnily enough, I also learned
to drive -after- getting my license)
mircea_popescu: lobbes
this incidentally explaisn why wikipedia is such shit - it's ~only function is a sort of open-sourced cliffnotes, and people would much prefer it
to be bland and stupidly written so
the
teacher in class doesn't feel
too inclined
to
think
the kids' lifted material isn't his. after all he added all
the flavour words in
there!
mircea_popescu: asciilifeform
this is a problem yeah. will however have
to be bit, wtf else
to do.
☟︎ lobbes:
http://btcbase.org/log/2017-01-03#1595456 << "First, and obviously, since
the majority of
the students are going
to get an A, he just has
to do just as well/horrifically as
the average student, and if
they're all writing about slavery with
the enthusiasm of a photocopier
then if he wants an A he better buckle down and learn
the
truly useful skill of masking
the words of a Wikipedia page. "
☝︎ mircea_popescu: anyway,
the great gain is
that no
two elements need/have write access
to
the same
thing by
this scheme. in point of fact one way
to look at current
trb/prb is
to say
that
they have "write locks" on all
the fucking
time and deadlock.
davout is getting lost in
the variable names
mircea_popescu: (this knob is
then in practice equivalent
to
the "checkpoints" discussed previously)
mircea_popescu: in particular N.B should be "older overwrites newer" style ring buffer. of particular concern are situations where
the buffer is set shorter
than
the longest reorg, in which case
the node will wedge.
TRB.N not accepting blocks with index lower
than highest of B.B is for sure not feasible. "how many behind" should be an operator knob.
☟︎ mircea_popescu: "this is my B.T1 of all
txn with no fee,
this is my B.T2 of all
the payments
to my X,
this is my B.T3 of all
txn over 5kb relayed by ip X"etc.
mircea_popescu: but it is not required for B.T
to be used only in
this way or for
this purpose. in principle
there could be a whole pile of
these, readily extended into whatever operator wants
to do.
mircea_popescu: this scheme among other
things cheaply allows
the "add arbitrary new address
to wallet", just have utility
that (separately) processes B.B and produces new set of B.T.
☟︎ mircea_popescu: in any case :
TRB.N needs write access
to N.B and M.T and read access
to B.B ;
TRB.B needs read access
to N.B and write access
to B.B and B.T. it may be a good idea
to also give
TRB.N read access
to B.T but
this should be operator-knob
mircea_popescu: M.T deliberately left unspecified, it is
the equivalent of
today's "mempool". perhaps should also be a ring buffer like
the other 3.
mircea_popescu: all
three queues
to be implemented as ring buffers of user specified size.
☟︎ mircea_popescu: otherwise it is discarded. B.T may be pruned (according
to arbitrary address list, for instance). Rate limiting in
TRB.N may be constructed
to observe N.B items
that fail
to propagate
to B.B and ban
the originating peers.
☟︎ mircea_popescu: TRB
to be split into
two parts :
TRB.B and
TRB.N. Queues B.B B.T N.B
to be created.
TRB.N inherits
the code
to connect
to peers.
TRB.N reads blocks from peers, and puts
them in N.B.
TRB.N reads
txn from peers and puts
them in M.T.
TRB.N does nothing else (with
the possible exception of rate limiting for peers).
TRB.B reads N.B and verifies
the blocks. if
the block is verified it is added
to B.B and its component
txn
to B.T ;
mircea_popescu: yes,
they're both "transactions" in
the
terms of
the eventual datastruct
they'll occupy.
they aren't for
that reason
the same
thing.
mircea_popescu: "i have heard
this
transaction" is of interest
to b ; not of a. "this is a
transaction from a block" is of interest
to a.
a111: Logged on 2016-12-20 19:42 mircea_popescu: nope. blockchain part will get
to it when it gets
to it, and
tell you. until
then, peer part builds queues.
mircea_popescu: you're not at liberty
to discuss something else, in different
terms.
mircea_popescu: no such
third was contemplated ; when discussing a proposal you are stuck, willy nilly, first understanding it and
then referring it
mircea_popescu: it ~may~ be
the case
that some arbitrary level of cleanliness requires an entirely new universe.
this, however, can never inhibit
the brushing of
toilets.
a111: Logged on 2017-01-03 19:59 asciilifeform:
trinque: it isn't , currently, clear
to me
that you can make
this cut cleanly without hard-breaking with
the
traditional protocol.
a111: Logged on 2017-01-03 19:56 ben_vulpes: because "map-reduce" does not reduce
to "here's how i'm going
to solve specifically
the case where
thread n finds an unspent out and
thread n+1 finds its spent and in
the reducing phase i collate everything proper-like"
a111: Logged on 2017-01-03 19:53 davout:
trinque: but "must RI be a single binary?"
trinque: "don't
talk
to idiots" is a far broader problem
than bitcoin
trinque: I was going
to say
the same, does require gossipd