log☇︎
109300+ entries in 0.026s
asciilifeform: worth a try
asciilifeform: soo which knob should i max, Framedragger ?
asciilifeform: i even spoke with career dbists, answer was 'your application is monstrous abuse and you need a cluster' ☟︎
asciilifeform: did.
asciilifeform: aha
asciilifeform: disk is not the bottleneck on this box
asciilifeform: what'll it give
asciilifeform: say, 64GB
asciilifeform: http://btcbase.org/log/2016-12-30#1593233 << all knobs set to defaults ☝︎
asciilifeform: ullage!1111
asciilifeform: even if it ~does~ go whrrrrr.
asciilifeform: if drum printer dun go 'bang' when it gets the octet - it ain't 'printable.'
asciilifeform: then take standard reiser. but in that case you grunt under entirely arbitrary procrusting .
asciilifeform: picture reiserfs , but without the idiot fortranistic hard limits on nodes, lengths, etc
asciilifeform: yeah but one that doesn't motherfucking grind to a halt when read 1000/sec omfg ☟︎
asciilifeform: how much of 'what trinque wants here' is unobtainable by simply abusing the fs ?
asciilifeform: *nearby
asciilifeform: mircea_popescu: no, it is specifically 'doesn't work because ~nearly there is a cock-shaped shadow~
asciilifeform: trinque: my understanding is that these ways typically involve clusters of machines, duplicate db, and very elaborate/failure-prone synchronizers
asciilifeform: aite
asciilifeform: http://btcbase.org/log/2016-11-19#1570951 << see also ☝︎
asciilifeform: imagine your cock didn't work if 10,000 others in town were in use.
asciilifeform: it is monumentally retarded, as a mere possibility in a system
asciilifeform: tolerable
asciilifeform: but so must orthogonal ~operations~. under nk circumstances is blocking on UNRELATED op, tolerablr
asciilifeform: all of the orthogonal tools must be mercilessly cut apart, yes.
asciilifeform: elaborate?
asciilifeform: my fs can eat GB/s r/w without breaking a sweat
asciilifeform: is it 1971? am i on a drum disk?
asciilifeform: and why the FUCK is 10,000 writes/sec 'too much'
asciilifeform: why the FUCK should a READ block because ~wholly unrelated datum is being written~
asciilifeform: i've developed a loathing, inexpressible in words, for postgres and all things like it
asciilifeform: ben_vulpes: 'this shape' being 'html as-seen-by-reader as the only storage format'
asciilifeform: trinque: mircea_popescu will probably barf if the 2 genesisen get displayed on same screen. it isn't an eggog in vtronics, but is in his head.
asciilifeform: in ~all~ cases.
asciilifeform: ~displaying a www page~ oughta be an O(1) op, really
asciilifeform: i've seriously considered reimplementing phuctor in this shape. as it is, it loses more from the slow writes idiotically queuing up, and the wedged reads that result, than it wins from fast structured queries.
asciilifeform: hence i ask 'why have 2'
asciilifeform: the mention of pg_notify
asciilifeform: anywhere
asciilifeform: at all
asciilifeform: yes but why have the sql db then
asciilifeform: so update the html?
asciilifeform: why have 2 representations of same thing on disk ?
asciilifeform: trinque: why would you generate the site from a db? more than once, i mean
asciilifeform: http://btcbase.org/log/2016-12-30#1592962 << news to me ☝︎
asciilifeform 'can't even', takes break, off to play with 10kg joystick, and then with pet
asciilifeform: phf: USES nonprintables in the magic ?! even deeper retardation.
asciilifeform: one can debate whether the persians are right to cut hands off thieves. but the hands of folx who write programs like this, i cannot see any reason why they should stay attached.
asciilifeform: the '.' operator in 'diff -e' is the magic.
asciilifeform: and THEY CAN BE ANY OCTET
asciilifeform: the alternative, the correct one, is 'next N bytes are payload'.
asciilifeform: if there's a forbidden char or string --- that is called magic.
asciilifeform: it's magic
asciilifeform: ick.
asciilifeform: phf: 'ed-style' diff outputs are the Right Thing, but done the ~proper~ way, with NO INBAND MAGIC, and not the monkey way.
asciilifeform: (i did not have any lines beginning with '.', so i have nfi what the escape would have looked like. but there WOULD be one.)
asciilifeform: http://wotpaste.cascadianhacker.com/pastes/3NXcp/?raw=true << what FUCKGOATS release would have looked like, in that mode.
asciilifeform: ffs.
asciilifeform: NO MOTHERFUCKING INBAND MAGICS
asciilifeform: phf: diff -e has own serious problems (it 'enmagics' the '.' character, for instance.)
asciilifeform: stock gnudiff is brain-meltingly dumb re minimality of diffage.
asciilifeform: phf: does it needleman-wunsch ?
asciilifeform: in the case of the orphanages, they had 0 constructive purpose. they were like the 'death glands' on that one species of octopus. snip'em and you get octopus that lives for +2 yrs and no other effect.
asciilifeform: the important thing to observe , in each case, is whether it changed semantics in a way that can be ~understood~
asciilifeform: ditto 'malleus'.
asciilifeform: but some -- did. the orphanage removals certainly did.
asciilifeform: ben_vulpes: not each (nixing the win32 #ifdefs did not, for so long as nobody is dumb enough to try to build for win32)
asciilifeform: (which, if it is left up to me, will look like teco macros. so if anyone does not like this shape, had better come up with another.)
asciilifeform: as soon as there is a viable replacement.
asciilifeform: phf: it is a terrible format and is not to remain.
asciilifeform: different semantics - different proggy. with different behaviour . different outputs for same inputs. etc
asciilifeform: ben_vulpes: semantics is , more or less, what the proggy ~is~
asciilifeform: ben_vulpes: same as in any other proggy
asciilifeform: and unquantifiably so.
asciilifeform: it WILL have variant semantics.
asciilifeform: but it is also not clear to me whether this can be done and the result still referred to as 'trb'.
asciilifeform: one theoretical solution to every type of blackhole other than the (theoretical) 'nsa sprays shit directly into the pipe on the backbone' is to make trb actually multiprocess
asciilifeform: and it isn't clear what will happen in re miners, for instance, when it does.
asciilifeform: but it isn't happening tonight.
asciilifeform: pete_dushenski: the only long-term answer is full wotnetization of the nodes.
asciilifeform: ben_vulpes: the 'a' and 'b' are historic artifacts from my torture room. but notice, gnudiff ignores the name.
asciilifeform: mircea_popescu has been running public nodez longer than i , and iirc has pretty good instrumentation, he might have something to add to this thread .
asciilifeform: i suppose for completeness one ought to include a '5' -- foolish folx who think that 4GB / non-ECC ddr4 / etc. is a trb node
asciilifeform: 4 is where 1,2,3 can be ruled out with some confidence. could be mircea_popescu's 'magic packet', or just about anything, i have ~0 useful data.
asciilifeform: type3 (this taxonomy is strictly from asciilifeform's notes, and is no canonical animal of any kind) is the 'thinking man's shitflood', where requested inventory is not a DOS by virtue of ~quantity~, but ~quality~ (flood of questionable but not immediately/cheaply rejectable material)
asciilifeform: type2 ( pete_dushenski's ) is the garden variety shitflood. which is sometimes solved by ip ban, but only in the case of 'shrapnel addressed to occupant', i.e. idiot prb nodes wildly spamming crapolade, and not in the 'bullet with your name on it' case, where somebody actually has a sybil constellation drowning your trb node in liquishit, with no SINGLE ip misbehaving in any way ☟︎
asciilifeform: and for so long as block verification is single-processor, there will remain type1
asciilifeform: (the one unattributable to overload of any type)
asciilifeform: mircea_popescu: ultimately for so long as peers are unauthenticated and speak unauthenticated plaintext , there will be type4 blackhole.
asciilifeform: and it is not in fact held up by the fs thing.
asciilifeform: (that lead to same state - wedged node)
asciilifeform: mircea_popescu: blackhole, in my current understanding, is at least 4 distinct things
asciilifeform: one thing that palpably helps is the aws banhammer.
asciilifeform: also i see some 'connect() failed after select(): Connection refused' which iirc is bleeding edge prb kicking trb out ☟︎
asciilifeform: spray of crapolade from $somebody, designed to bog down your node and have it lose all peers
asciilifeform: pete_dushenski: this'd be the genuine article
asciilifeform: ben_vulpes: the thing phf refers to , is in use still, whenever i make (yes) xp box! < 400MB! (lighter weight than, e.g., africa-linux)
asciilifeform: and can you get a core dump out of the thing pete_dushenski
asciilifeform: pete_dushenski: from your telling, it seems that there was no 'after'. so let's have the 'before' and 'during'