248200+ entries in 0.162s
mircea_popescu: o check it out,
the supply of internet boat has made it
to
the french malvines!
mircea_popescu: certainly
the discussion of gossipd needs some updating.
mircea_popescu: n exist.
there's a lot of similarly broken stuff in
there, and
then
there's strange notions i can't readily evaluate. it is also much
too long, yet it evidently is
too sparsely written in most nodes. do yourself &
teh world a favour and rewrite
this
thing.
mircea_popescu: the concept you're describing is sound,
the example you use is utterly idiotic. for one
thing "validating"
that a video was watched is necessarily a void concept ; for another who even fucking cares if
they did. in
the path between a and b you've introduced a "simplifying detour" of
the format a->moon, moon->b, which is pluriously dysfunctional -
there's not even any reason
to believe moon exists in any space where a and b ca
jhvh1: mircea_popescu:
The operation succeeded.
mircea_popescu: !~later
tell gabriel_laddel so
the
thing is - you have
to rewrite
this.
there's a lot of very powerful ideas struggling in
the mire of very poor (and often rushed) writing.
there's no way out, i can't rewrite it for you. while it's starting
to emerge into sense it's not yet at
the level of development where it can be productively engaged.
take but one simple example : "$computation against $uri validate videos were watched".
mircea_popescu: lol
the newman quote found its way into g_l's arsstep
thingee
mircea_popescu: hence looking into what ext2 does if you do a lot of simlinks. not like
this can't be dry run with bogodata.
mircea_popescu: the one concern is
that 5mn simlinks in you'll discover
that "oops, we decided X number of simlinks should be enough for everyone fuck you"
mircea_popescu: a block only comes in every x minutes, only has a hundred or so
txn if
that.
mircea_popescu: what
time ? a
txid doesn't change what block it's in does it ?
mircea_popescu: (speaking of which : maxsimlinks used
to be hardcoded
to 40 via param.h for reasons. dunno if still
the case)
mircea_popescu: whenever you process a new block, you add
the needed simlinks in /txid/etc
mircea_popescu: how about
this scheme : you do /bitcoin/blockid/aaaa/bbbb/cccc/dddd/ a
total of 16 levels, off
the blockid, wherein at
the end a block.dat 1mb file is found. you also do /bitcoin/txid/aaaa/bbbb/ etc also 16 levels,
the last of which you SIMLINK
to
the correct block.dat dir
mircea_popescu: the only correct way
to do
this is
to make a binary
tree out of
txids and save
that somehow isn't it
trinque: I will be interesting
to hear what ben_vulpes has
to say about
this, given recent block-parsing adventures
☟︎ mircea_popescu: very much so. also happens
to be
the correct approach.
mircea_popescu: it's clear
that it needs a message queue / interface with
the proggy, but
that's about all
that's clear.
trinque: for
that matter neither
the "utxos and other lookups"
mircea_popescu: sadly.
the main
thing
that holds back sanevelopment on
trb is
the fact
that whatever you want
to discuss everyone starts muttering curses and mashing kbd about
the 10bn other
things connected.
mircea_popescu: because lo and behold,
the mempool of
their presence and
the blockchain of
their past deeds are not
the same exact fucking item.
mircea_popescu: the
three girls presently in
the room have
three girls names, but if i have
them fuck each other and film
the proceedings
the filename will be something with a
time in it. apparently
this doesn't make
the walls collapse on my head.
mircea_popescu: they're entirely separate concepts, mashed
together because idiots
think
that
the way
the world appears
to
them in
the moment is
the way
the world is.
mircea_popescu: asciilifeform no,
the
txid as used in mempool and
the
txid as used in block is not required
to match.
mircea_popescu: definitely
trb fork will have a sane
tx-to-block coding scheme
mircea_popescu: reverse ptr for
the love of fuck. it's FOUR BYTES OMFG
trinque: doing something non-idiotic with
the set of utxos seems unavoidable if you want
to use
the
thing
mircea_popescu: asciilifeform sure. i suppose you save lookup
tables huh.
mircea_popescu: i dunno
that you'd want
to make
txn files really.
they're small, just load
their block in memory.
mod6: <+BingoBoingo>
To
the Republic I pose New Header for
the gray lady with
the rabbit hat. "Qntra - [Fill-this-blank] Of
The Most Serene Republic" << Chronicle ?
BingoBoingo: <mircea_popescu> what was it, Mainstream News Outlet ? << "Propaganda Mouthpiece"?, Whatever it is
this question now goes
to
the Peerage for suggestions.
mircea_popescu: they are still
turds. i don't want
to seek in a fucking 2gb file
to find block N
a111: Logged on 2016-11-01 03:24 mircea_popescu: i have nfi how satoshi reasoned
to arrive at
the current nonsense of 2gb monoliths, but it was dumb.
BingoBoingo: To
the Republic I pose New Header for
the gray lady with
the rabbit hat. "Qntra - [Fill-this-blank] Of
The Most Serene Republic"
a111: Logged on 2016-11-01 03:20 mircea_popescu: also i have nfi what "db" we can
trust.
they're all shit
mircea_popescu: pop a fresh 8x
TB ssd in a 1+0 raid array,
ten years later
they're within 95% of factory speeds. because never fragmented, because nothing ever deleted.
mircea_popescu: and it's also immensely ssd friendly : you don't
touch most blocks. say you keep i dunno, 12 or something like
that, user-configured, in memory.
this way, you only write
them
to disk once.
mircea_popescu: or w/e, what was
the limit, something stupid like 32k files ?
mircea_popescu: moreover,
the disks / kernel are exactly optimized for
this
task. fuck around with ~mb ish item on disk
BingoBoingo: <mircea_popescu> i have nfi how satoshi reasoned
to arrive at
the current nonsense of 2gb monoliths, but it was dumb. << FAT32 file size limit
mod6: sure, i
think going block -> certainly fits into
the same paradigm as eatblock/dumpblock.
mircea_popescu: i have nfi how satoshi reasoned
to arrive at
the current nonsense of 2gb monoliths, but it was dumb.
☟︎ mircea_popescu: actually, improving logging
to add some hooks for profiling in
there may be a good idea.
mircea_popescu: mod6 note i'm not even discussing a special fs.
that's a while down
the road. for now, all
that's needed is some serious profiling work,
to see how we store
the
thing. prolly as block = file ? and just use a normal ext4 or w/e
mircea_popescu: and
then of course
they'll discover
this nice bug
that's been
there for 10 years and utterly ruins you. so you gotta update. except
the only update available is made by florian "i eat shit for breakfast" weimer and well... it contains 3 extra holes
to be burned in 2025
mod6: yeah, me either.
they're garbage.
mircea_popescu: (large codebase, many changes, many "developers" === shit, by definition, no way out of
this)
mod6: Agree, we've def. discussed a few
times.
mircea_popescu: mod6 iirc it keeps popping up w/e we discuss
the bitcoin db monstrosity, which is about each 9 months.
mircea_popescu: just
the benefit of NOT haviong a whole pile of code in
there doing "db" is already immense ; but
there's actually no gain from slowing down
the disk seeks etc
through extra wrappings
mod6: Ah indeed, an 'interesting idea'. I'm excited
to see where we can
take
that.
mircea_popescu: the
thing with
the db is, i keep
thinking about it and i keep failing
to convince myself a db is even needed in
the first fucking place.
truth be
told blockchain is
the
textbook example of "you don't db
this shit son, use
the filesystem"
mircea_popescu: mod6 yeah raw
tx / log fixing sounds pretty great actually.
BingoBoingo: Anyways, reall Halloween
truth. When Elliot was in
that Spielberg movie ET, Elliot did not dress up
the alien as a ghost. Elliot dressed
the alien up as a klansman.
a111: Logged on 2016-10-31 23:37 mircea_popescu:
http://btcbase.org/log/2016-10-31#1560997 <<
this is exactly how it goes, on one end of
that : everyone has his own boxes / his own process for acquiring boxes. you can't buy
them boxes and mail
them over, not really, and if you could it'd be a specialized job and wouldn't work like
this. and
this isn't even
the only angle, people are also variously allergic
to divers proteins involved etc.
a111: Logged on 2016-10-31 21:45 gabriel_laddel:
http://btcbase.org/log/2016-10-31#1560440 < 2.5k usd. Per month, until I bitched about it and it was knocked down
to $600/mo - anyways, it isn't about
the money, it's about him doing absolutely nothing with it and making a mess
that someone will have
to clean up.
a111: Logged on 2016-10-31 21:33 gabriel_laddel: When my life is less shit I'll be setting up a method by which in-WOT people can get it for free. You all, of course, will be able
to sell
the same product I am.