log☇︎
120100+ entries in 0.078s
asciilifeform: trinque: i dun have a replacement, can relax for nao. but would like to know which of the knobs actually do something useful.
asciilifeform: q is, how much can be removed from http and still result in e.g. proper display on typical heathen browser, wget-able tarballs, etc
trinque: I'm just telling you tools will want things you're muntzing, so why bother trying to fix
asciilifeform: i'd like to shoot it behind the shed already.
trinque: I'm not here to defend HTTP, you are, by trying to fix it
asciilifeform: and example of just the kind of thing that makes http's designers subhuman
trinque: "do I already have the item at $path" and "how big is the item at $path" are interesting questions without the server beginning to transmit the item itself. ☟︎
asciilifeform: having to ask the 'did it change' question per-file, is terrifyingly bad design
asciilifeform: 'telegraph office is three doors down'
asciilifeform: 'sorry i only have 1 bit of storage', picture this.
asciilifeform: all protocols require some minimal space to work in.
asciilifeform: there gotta be to hold the http header already.
trinque: there is no magic amount of space on the receiving end that will always be there!
asciilifeform: then it oughta be in sections, and the sections in small manifest. merkle-style. ☟︎
trinque: lets say you use your "magic file" for mirroring, manifest, then what when the manifest is very large?
asciilifeform: what other http knobs are there that, despite that neither i nor deedbot nor anything any of us have made , ever use, but somehow essential ??
asciilifeform: something that i actually keep around
asciilifeform: i'm trying to find something that does
asciilifeform: i dun see why this requires a 'send ONLY the header' special knob to exist.
asciilifeform: well naturally you wouldn't ctrl-c with hands, but break yer connection after you got what you want, i.e. the header
trinque: bahaha, yes, you can time your control-c *just so* that you get only headers ☟︎
asciilifeform: without loading whole thing
asciilifeform: my copy of wget doesn't seem to use HEAD for this, tho. it puts out a GET and simply aborts if you ctrl-c
trinque: and furthermore the idea that providing just the headers, which already exists as a codepath, and then bailing, adds mb, is idiotic
trinque: should the fs require you read() whole file to know file metadata?
asciilifeform: aaaand this is how stateolade gets into 'stateless' protocol
trinque: I will not use an item that can't give me metadata for a given path without loading the potentially gigantic item itself
asciilifeform: we don't use it ? it dies. that simple ☟︎
asciilifeform: trinque: what do we have here that issues HEAD ?
asciilifeform: the http protocol is not the place for manifests.
asciilifeform: if you want to know that file has changed, you update a manifest, and other side fetches that.
trinque: you're going to what, download the whole hundreds of gb every time you sync? ☟︎
trinque: I have the gentoo distfiles on http
asciilifeform: mirroring is to be done the sane way, by fucking loading the file.
asciilifeform: and don't intend to
asciilifeform: i've never written proggy that issues 'HEAD' ☟︎
trinque: ^ how are you going to know whether item over there changed without HEAD
asciilifeform: prolly also other things.
asciilifeform: even in this one, i can already see obvious candidates for snippycutters
trinque waits patiently for a gossipd atop which to redo hypertext ☟︎
asciilifeform: BingoBoingo: that's more or less exactly what i was looking for, ty. not to use, of course, but to adaize.
asciilifeform: and imho selling 10+ yr old server, prolly complete with fan dust cake, for 'you must ask for quote!' takes some cheek. ☟︎
trinque: eh the no-prices thing is an orc haggling thing
BingoBoingo: Beating the garbage truck to the dumpster seems to be a substantial activity here.
asciilifeform: it is also possible that this is one of those orcistans where there is not such a thing as a scrapyard, errything gets picked up by scavengers and reused same day ☟︎
BingoBoingo: Still haven't found it. Most of the local scrapping industry tends to center around hawking junk finds at la Feria Tristan Narvaja on Sundays.
asciilifeform: i suspect that it has better boxes than the one linked above
BingoBoingo: Still, they get an email along the lines of "I am scaling up my business here in Uruguay and need a bunch of 1U server boxes. What can you get as whole computers or empty cases? ☟︎
BingoBoingo: asciilifeform: Gotta inquire because the shops dun know how to web right
asciilifeform: BingoBoingo: where is the price on that page ?
BingoBoingo: A lot of the locally availabe refurb stuff is fat guys like http://www.okcomputers-uy.com/productos/productos_masinfo.php?id=4382&secc=productos&path=0.42.108
BingoBoingo: <trinque> BingoBoingo: when do you anticipate having a final number for 1U of space, and when do you plan on beginning to rack customer machines? << I am putting out solicitations for hardware again to the local shops today. Will squeeze in a pricing post today for the Republic's consideration.
BingoBoingo: <asciilifeform> mod6: try as i might, i can't picture how a photo would help to exploit yer box ( aside from the 'tells enemy which rack to dynamite' aspect, but mp_en_viaje addressed that one earlier ) << Ah, datacenter fellow expressed some bemused curiosity about the unbranded Qntra machine. Apparently everything else is plastered in brand names because latino rather than assembled into generic steel chicom chassis
asciilifeform: ave1: and plz post the output of gnat -v . ☟︎
asciilifeform: ave1: you see this for all of the invocations ? and 0 speed diff when removing ( not -gnatN, but remove, with sed ) ~all~ of the inline directives ?
ave1: strange that in disasm I see call <fz_...>, but will try more!
ave1: Ok I will check further into the code, bbl
asciilifeform brb, teatime
asciilifeform: you can trivially verify the latter by commenting out the Inline param and observing 5-20x speed differential
asciilifeform: this is a known bug in 2016, and is a red herring, they still get inlined
asciilifeform: lemme guess, you disasmed the bin and saw them as functions ?
asciilifeform: ave1: tell me how you determined that they were not inlined
ave1: So, I started to try commandline options
ave1: I was debugging the code and saw that the functions where not inlined "FZ_Clear", "FZ_*" etc.
asciilifeform: why. did. you. use. this. flag.
asciilifeform: consider answering the question, ave1 ?
asciilifeform: why the fuck did you USE this flag tho
ave1: as for the switch, in old docu it is here: https://gcc.gnu.org/onlinedocs/gcc-4.1.2/gnat_ugn_unw/Switches-for-gcc.html#ndex-g_t_0040option_007b_002dgnatm_007d-_0028_0040command_007bgcc_007d_0029-125
asciilifeform: i'm not a telepath, but somehow nobody thinks to preface their observations with 'i was using gnat xxxxx, which i got from...'
ave1: o wait, that is only in combination with "-gnatN", so ignore the Inline_Always comment
ave1: i.e. when running with "-gnatwa", I get " warning: pragma appears too late, ignored"
ave1: also, The "Inline_Always" pragma's seem to in the wrong part of the code (but does not help moving"
ave1: asciilifeform, I'm reading trough chapter 2 and I've tried to get inlining to work, but so far have failed, nothing gets inlined outside of the defining module!
mp_en_viaje: aaand in other "keeping the toilet seat up" domestic arguments, http://78.media.tumblr.com/e538aadd26355d0ade5c9cb9088e80d0/tumblr_oj309lRovF1uzx8ido1_400.gif
a111: Logged on 2018-02-02 01:29 asciilifeform: aha. recall, there was at least 1 occasion of trilema server growing legs and walking away
mp_en_viaje: http://btcbase.org/log/2018-02-02#1779327 << not to mention what's probable TWO occasions of phuctor disks. ☝︎
trinque has services that can get by on a pretty humble machine, static HTML and IRC bots
trinque: I would rather see a sampling of potential machines than to guess at what might be available.
trinque: BingoBoingo: when do you anticipate having a final number for 1U of space, and when do you plan on beginning to rack customer machines?
deedbot: trinque updated rating of BingoBoingo from 1 to 3 << went on a long march and resurrected the Herald
asciilifeform: but this is for later.
asciilifeform: and that's the compleet story under amd64 , re stdio
asciilifeform: technically it wants ~three~ ttys, one for normal i/o, one for eggogs, one for rng
a111: Logged on 2017-07-19 19:00 asciilifeform: i have 'p' structured such that all it wants i/o-wise is a 'getbyte' and 'putbyte' that operator can replace with whatever
asciilifeform: ^ for reasons unknown to me, ada's standard character i/o routines mutilate newlines. so ffacalc imports the standard unixland ones.
a111: Logged on 2018-02-02 01:11 mp_en_viaje: http://btcbase.org/log/2018-02-01#1779163 << the theory is that very smooth, object files can be linked together directly. the testing of this theory is the most valuable part of trying it early.
asciilifeform: http://btcbase.org/log/2018-02-02#1779288 << ftr there are two methods for this; depending on which lang the 'main' procedure was written in. see also https://gcc.gnu.org/onlinedocs/gnat_ugn/Interfacing-to-C.html , has example of each. ☝︎
mod6: apologies for the disruption. carry on!
mod6: my first inclination is to cut down as many attack vectors as possible. real or imagined. the imaginary ones can be rooted out through a bit of thinking / conversation with intelligent folks.
mod6: im not sure if that was english.
mod6: this may be an asinine thread i started here, but i'd rather be thinking this way - defensive, even if asinine, and then think/talk it out.
asciilifeform: if it e.g. sends random page of ram as a coded dns packet every full moon, it will do this regardless of whether anyone photographed it, lol
mod6: anyway, this is probably nonsense.
mod6: yeah, I don't think ours will even have that anyway. BUT, the point stand that, I didn't build it; who knows what nasty pitfalls are in there.
asciilifeform: no particular reason why you'd leave this item plugged in, afaik
asciilifeform: mod6: these are only a problem when plugged in
mod6: so, i've heard like nasty things about some of these boxes with like the lights out management thing, and whatever.
asciilifeform: hey i was curious re whether the hypothetical can be made to make sense
mod6: annnnyway, this is probably nonsense. carry on!
asciilifeform: aha. recall, there was at least 1 occasion of trilema server growing legs and walking away ☟︎