00:01:58 -!- L8D has joined.
00:02:11 -!- zzo38 has joined.
00:06:54 -!- elliott has quit (Ping timeout: 246 seconds).
00:07:38 -!- elliott has joined.
00:08:02 -!- elliott has changed nick to Guest69134.
00:11:39 -!- ^v has joined.
00:26:06 -!- boily has joined.
00:29:05 -!- ProofTechnique has joined.
00:33:33 -!- L8D has quit (Ping timeout: 246 seconds).
00:35:14 -!- hjulle has joined.
00:48:02 <boily> ProofTechellonique!
00:51:57 -!- Andodaan has joined.
00:52:11 -!- hjulle has quit (Ping timeout: 264 seconds).
01:04:57 -!- iamevn has joined.
01:06:24 -!- Kasran has joined.
01:10:01 -!- callforjudgement has joined.
01:10:28 -!- ais523 has quit (Remote host closed the connection).
01:11:07 -!- callforjudgement has changed nick to ais523.
01:13:58 <HackEgo> [wiki] [[User:Kasran]] N http://esolangs.org/w/index.php?oldid=42180 * Kasran * (+71) hello!
01:27:18 <orin> god dammit. these ads on youtube music videos that are, themselves, music videos
01:28:56 <ais523> bonus points if two music videos have each other as ads?
01:29:00 * boily hands some AdBlock potions to orin
01:29:14 <boily> hellorin. aren't you oren usually?
01:29:34 <ais523> I assume some people archive ads on YouTube
01:29:49 <ais523> perhaps some day someone will have the ad for a video be the same as the video itself
01:30:21 <orin> boily: yah, my passport says "oren" but sometimes I spell it other ways
01:30:32 <Andodaan> The realization is that the encapsulating music video is really an ad.
01:31:29 <orin> I would'nt mind so much if the two musics were the same genre
01:32:19 <Kasran> chiptune demo video opening with bruno mars' new single
01:32:21 <orin> which shawn mendes and BABYMETAL are not, at all
01:32:23 <ais523> my passport doesn't say "ais523" :-(
01:32:53 <boily> my has a boily in there ^^
01:33:24 <boily> orin: the only way to get babymetal out of my mind is some heavy touhou remixes.
01:35:36 <orin> adadadadadada dokkyu! zokkyu dokkyu zokkyu dokkyu yada yada yada neva neva neva!
01:36:11 -!- ais523 has quit (Read error: Connection reset by peer).
01:36:26 * boily mapoles orin with a swift swing
01:36:26 -!- ais523 has joined.
01:37:19 <boily> Kasran: IOSYS usually does the job for me, but I do enjoy some Alstroemeria. what are your favourite circles?
01:39:08 <orin> shinra bansho, hatunetu-miko's, IRON ATTACK!
01:39:26 <Kasran> I'm not super into Touhou music, so I'm not sure I'd be able to pick out any favorites. I do like what I've heard from Crow's Claw, though.
01:40:02 <boily> orin: heh! listening to hatsunetsu at this moment right now :D ♪
01:40:37 <boily> Kasran: you should try Demetori, particularly their later albums. some epic rocking in them.
01:41:24 <Kasran> bleh!!!! And alright, I'll take a look, thanks.
01:42:16 <orin> I also like anything from Odyssey Music/T. Stebbins/Eurobeat Brony/ whatever other pseudonyms he has
01:43:45 <orin> An amazing male vocalist who does songs in a whole bunch of fandoms
01:44:10 <orin> https://www.youtube.com/watch?v=4LpGkyR3hfI
01:44:15 -!- Guest69134 has changed nick to elliott.
01:44:18 -!- elliott has quit (Changing host).
01:44:18 -!- elliott has joined.
01:48:17 -!- Lymia has quit (Ping timeout: 265 seconds).
01:51:12 <orin> Apparently he is also named Ken Blast and also mortimer Mortimer
01:52:01 -!- oren has joined.
01:52:25 -!- orin has quit (Quit: ssh to uoft is slowing up).
01:53:56 <oren> no that was a result of a slow connection
01:54:48 <oren> by convention Eurobeat song and artist names are limited to numbers and uppercase letters. No idea why
01:57:33 <oren> A typical song listing is of the form "PRETTY GIRL ROCK / SPEEDMASTER feat CINDY"
01:58:54 -!- boily has quit (Quit: ARTICULATED CHICKEN).
02:02:03 <HackEgo> oren is a Canadian esolanger who would like to obliterate time zones so that he can talk to his father who lives in the same house.
02:02:28 <oren> hes asleep already
02:21:47 -!- Phantom_Hoover has quit (Remote host closed the connection).
02:28:20 -!- adu_ has joined.
02:40:20 <zzo38> I was reading something David Sirlin wrote recently, and now I can see that he even designed a Magic: the Gathering card.
02:45:59 -!- ais523 has quit.
02:46:03 -!- callforjudgement has joined.
02:46:04 -!- callforjudgement has changed nick to ais523.
02:46:16 -!- ais523 has quit (Changing host).
02:46:16 -!- ais523 has joined.
03:00:36 -!- Lymia has joined.
03:26:02 -!- ^v has quit (Ping timeout: 256 seconds).
03:29:03 -!- ^v has joined.
03:30:28 -!- Andodaan has quit (Read error: Connection reset by peer).
03:47:16 <oren> Screw this crap. I'm making my own image editor
03:49:08 -!- augur_ has quit (Ping timeout: 265 seconds).
03:49:58 <coppro> oren: what is this crap?
03:50:06 -!- vodkode_ has quit (Ping timeout: 265 seconds).
03:51:06 -!- L8D has joined.
03:51:13 <oren> coppro: mtpaint
03:53:30 -!- ^v has quit (Ping timeout: 265 seconds).
03:55:34 -!- augur has joined.
03:55:44 <oren> I'ma make a better one. with scripting. and whores
03:57:46 -!- ais523 has quit.
03:57:49 -!- ^v has joined.
03:58:00 -!- ais523 has joined.
03:58:52 <oren> specifically, I want to be able to select a quadrilateral out of the image and deform it into another quadrilateral
04:00:36 <ais523> oren: are they both parallelograms?
04:01:35 <oren> they are roughly trapezoids
04:04:43 <ais523> affine translation might still work, on that data, but it's far from guaranteed
04:04:47 <ais523> err, affine transformation
04:08:28 <coppro> the way I'd do it is by taking the diagonals to anchor your coordinate system, and define the transformation in terms of that. Not sure if it's linear
04:10:14 -!- GeekDude has quit (Quit: {{{}}{{{}}{{}}}{{}}} (www.adiirc.com)).
04:10:39 -!- _AndoDaan has joined.
04:26:23 -!- L8D has quit (Ping timeout: 264 seconds).
04:31:53 -!- vodkode_ has joined.
04:33:37 -!- shikhin has quit (Ping timeout: 256 seconds).
04:36:21 -!- zzo38 has quit (Ping timeout: 250 seconds).
04:40:40 -!- iamevn has quit (Quit: iamevn has quit (Quit: iamevn has quit (Quit: iamevn has quit (Quit: iamevn has quit (Quit: iamevn has quit (Quit: iamevn has quit (Quit: iamevn has quit (Quit: iamevn has quit (Quit: iamevn has quit))))))))).
04:51:39 -!- callforjudgement has joined.
04:51:50 -!- ais523 has quit (Disconnected by services).
04:51:51 -!- callforjudgement has changed nick to ais523.
05:14:10 -!- bb010g has joined.
05:15:53 <fizzie> Image processing tools generally call that a "perspective" transformation.
05:16:57 <ais523> assuming the trapezia are isoceles and the two parallel sides are horizontal, I've also seen it called "keystone"
05:18:30 <fizzie> http://www.imagemagick.org/Usage/distorts/#perspective etc.
05:19:42 <fizzie> (I've only seen the "keystone" term in a projector context.)
05:20:06 <ais523> haha, POSIX defines the syntax of yacc using a yacc grammar
05:20:16 <ais523> thus leading to a standard in need of bootstrapping
05:20:57 <ais523> YAELI: a language for which the syntax of yacc is an interpreter for that language, ideally in as non-contrived a way as possible
05:22:24 <pikhq> POSIX defines all of its syntax using a yacc grammar.
05:23:19 <ais523> that doesn't make it any less of a circular definition
05:25:04 -!- Novtopro has joined.
05:27:01 -!- ais523 has quit (Read error: Connection reset by peer).
05:27:14 -!- ais523 has joined.
05:30:53 -!- L8D has joined.
05:35:06 -!- ais523 has quit.
05:35:16 -!- ais523 has joined.
05:46:53 -!- Novtopro has quit (Quit: Textual IRC Client: www.textualapp.com).
06:01:45 -!- _AndoDaan has quit (Ping timeout: 245 seconds).
06:18:49 -!- Kasran has quit (Quit: Page closed).
06:24:20 -!- fractal has quit (Ping timeout: 272 seconds).
06:29:02 -!- fractal has joined.
06:35:10 -!- oerjan has joined.
06:36:05 -!- adu_ has quit (Quit: adu_).
06:47:07 <oerjan> fizzie: fungot and zemhill got k-lined...
06:47:17 <lambdabot> boily said 19h 28m 43s ago: I know.
06:47:31 <ais523> I'm going to assume that was in-context
06:47:40 <ais523> but it's much less than a year, so no new record
06:48:20 <ais523> <oerjan> fizzie: fungot and zemhill got k-lined... <boily> I know
06:48:39 <HackEgo> 497) <CakeProphet> monqy: help how do I use lambdabot to send messages to people. [...around half an hour later...] <CakeProphet> @messages <lambdabot> quicksilver said 1y 2m 18d 19h 54m 29s ago: you use @tell
06:48:46 <ais523> and that's the previous record
06:49:02 <ais523> and still the current record
06:50:14 <oerjan> >25 years without a cavity, and BAM! 5 in one year ;_;
06:50:36 <ais523> have you changed diet?
06:50:39 <oerjan> and i have no one to blame but my shrinks
06:51:00 <ais523> I'd better not give too much advice, though, or you might turn into Sgeo
06:51:27 <oerjan> ais523: more chocolate, less brushing teeth, pretty obvious cause really
06:51:52 <ais523> I actually gave up chocolate accidentally in march last year (possibly even february)
06:51:54 <ais523> and decided to just run with it
06:52:50 <oerjan> i started eating chocolate in the middle of the night because it seemed to help with my foot cramps...
06:53:17 <oerjan> but i had already done that a year ago, when the checkup was clean
06:53:38 <oerjan> well maybe not as much
06:54:42 <oerjan> also, i'm unlikely to become like Sgeo, my way is to ignore all advice completely hth
06:58:58 -!- ais523 has quit (Read error: Connection reset by peer).
06:59:07 -!- ais523 has joined.
07:07:55 <HackEgo> [U+FFFD REPLACEMENT CHARACTER]
07:08:26 <oerjan> someone didn't have unicode set up properly
07:11:38 <HackEgo> [U+1D350 TETRAGRAM FOR FAILURE]
07:12:31 <oerjan> pretty much the same thing, there
07:12:59 -!- vodkode__ has joined.
07:13:17 -!- vodkode__ has quit (Client Quit).
07:13:41 <oerjan> oh wait it's not supposed to look the same
07:13:53 <oerjan> pretty much a failure, then.
07:14:49 -!- _AndoDaan has joined.
07:20:43 <HackEgo> [U+8CA0 CJK UNIFIED IDEOGRAPH-8CA0]
07:21:29 <oren> see this is stupid. why does the tetragram have its iching meaning in it but the kanji has a stupid hex number
07:22:00 <oerjan> i've been assuming it's because the kanji's meaning might depend on language
07:22:30 <oren> not to such an extent...
07:22:56 <oerjan> i guess it's because they hate humans, then
07:23:52 <oren> damned racists
07:24:15 <oerjan> technically that's speciesist hth
07:40:58 -!- vodkode_ has quit (Ping timeout: 272 seconds).
07:43:25 <oerjan> <orin> why does irssi use your local username as default nick <-- because you haven't saved your nick setting?
07:44:05 <oerjan> actually i think that's part of the server setting
07:44:22 -!- Patashu has joined.
07:45:37 <oerjan> hm nick isn't listed as an argument on permanent servers
07:48:40 <oerjan> well it's in .irssi/config, anyway
08:09:32 -!- zadock has quit (Quit: Leaving).
08:09:40 -!- L8D has quit (Ping timeout: 245 seconds).
08:12:22 -!- _AndoDaan has quit (Remote host closed the connection).
08:15:05 -!- SopaXorzTaker has joined.
08:35:20 -!- shikhin has joined.
08:55:07 <HackEgo> [wiki] [[Brainloller]] http://esolangs.org/w/index.php?diff=42181&oldid=32144 * 212.13.18.233 * (+108) Added php implementation on github
08:59:45 -!- SopaXorzTaker has quit (Remote host closed the connection).
09:21:52 <izabera> aren't you tired of those stupid console programs that don't have a line editing library?
09:22:19 <izabera> and you press the arrows and get ^[[A ^[[B
09:23:40 <izabera> well, fear no more! https://github.com/izabera/bin/blob/master/rrlwrap
09:24:41 <izabera> now you can just run rrlwrap otherprogram and the power of readline will be yours
09:28:23 <ais523> izabera: how does that differ featurewise from ledit, another program that does the same thing?
09:28:39 <izabera> first of all, the author is much sexier
09:29:13 <izabera> second, the author didn't know ledit
09:29:37 <izabera> the author did know rlwrap but it was a fun experiment anyway
09:29:37 <ais523> author sexiness has never been a reason to choose between programs
09:29:44 <ais523> otoh, the author of ledit didn't know of rrlwrap either
09:29:49 <ais523> so that condition's kind-of a wash
09:31:03 <izabera> this has the added bonus of being less than 100 lines long \o/
09:31:30 <ais523> it's not entirely a bot
09:31:36 <ais523> it's a human and bot using the same account
09:31:40 <oerjan> ooh myndzi is working properly again
09:31:42 <ais523> which can get confusing when they're both speaking
09:32:14 -!- AnotherTest has joined.
09:34:15 <izabera> third reason: this is the only important part of my code and it works on its own
09:35:04 <ais523> it's written in shellscript?
09:37:38 <oerjan> i'd ^celebrate but someone killed fungot
09:38:30 <ais523> why would fungot be klined?
09:39:06 <ais523> well I assume they're on the same connection
09:39:08 <ais523> so that makes more sense
09:46:21 -!- dianne has quit (Quit: byeannes).
09:52:41 -!- ais523 has quit.
10:11:35 -!- oerjan has quit (Quit: leaving).
10:13:05 -!- boily has joined.
10:21:52 -!- SopaXorzTaker has joined.
10:43:31 <fizzie> And yes, they're connecting from the same place.
10:44:15 <fizzie> "You are banned from this server- Spam is not allowed on freenode. Please contact kline@freenode.net if you think this kline has been set in error."
10:45:38 -!- fungot has joined.
10:45:44 <fizzie> Well, it seems to have been temporary.
10:55:51 <Jafet> Maybe fungot needed to make some urgent international funds transfers.
10:55:51 <fungot> Jafet: so what of it doesn't really work
10:56:26 -!- Phantom_Hoover has joined.
10:58:40 <boily> fungot: stop bein sentient.
10:58:40 <fungot> boily: yeah, i didn't know much
10:58:45 <fungot> boily: ' s if you can implement call/ cc k receiver) ( receiver ( lambda ( x)
10:59:00 -!- boily has quit (Quit: SCRATCH CHICKEN).
11:30:11 -!- Tritonio has joined.
11:31:27 -!- Patashu has quit (Ping timeout: 244 seconds).
11:39:15 -!- Tritonio has quit (Ping timeout: 252 seconds).
11:46:13 -!- shikhin has quit (Ping timeout: 252 seconds).
11:48:17 -!- shikhin has joined.
11:53:39 -!- mroman has joined.
11:53:50 -!- AnotherTest has quit (Ping timeout: 245 seconds).
12:33:01 -!- ProofTechnique has quit (Ping timeout: 250 seconds).
13:02:12 -!- ProofTechnique has joined.
13:12:50 -!- hjulle has joined.
13:33:02 -!- `^_^v has joined.
13:39:00 <HackEgo> fun fact 0 = 1 | fact n = n * fact (n - 1)
13:40:57 <b_jonas> that looks like Standard ML
13:42:07 <Taneb> That was my assumption
13:42:13 <HackEgo> /j汣!ŕElDP3A5BJ.Uj%{c. ]4IӍjmVA0"SFDI1b}aVMd[⁉9.VfAp(/#ͷJcXlzՈɊe:;qn \ ȪJo](*BE}Io,;Hy_X@;sݿ!urnk!00`ZD(T⩟ЧCO_J0ۻ
13:43:30 <Taneb> One day I am going to be all boring and switch to the three-letter nick I have registered that is just my initials
13:46:13 <Taneb> I am kind of glad all 4 of those are vaguely pronouncable
13:46:47 <FireFly> Leading ŋ is kinda tricky to pronounce
13:47:28 <Taneb> FireFly, I manage it somehow
13:48:50 <b_jonas> fungot, how can a video encoder use two gigabytes of memory?
13:48:50 <fungot> b_jonas: ( inherently)? because i don't know
13:49:09 <fungot> Available: agora alice c64 ct darwin discworld enron europarl ff7 fisher fungot homestuck ic irc* iwcs jargon lovecraft nethack oots pa qwantz sms speeches ss wp youtube
13:49:16 <fungot> b_jonas: well it tastes bad. they should know better then the movie list two friends of mine and i are in completely different places in the template to be nothing in the rules
13:49:58 <Taneb> I know someone who runs a fungot-like channel-regular imitation bot in another channel
13:49:58 <fungot> Taneb: it's not my fault, i believe.) hmm... long digression, guys!".
13:52:31 <Taneb> Does USD form a module over the ring of 100ths of integers?
13:53:13 <Taneb> And thus exchange rates form a homomorphism!
14:12:34 -!- SopaXorzTaker has quit (Ping timeout: 246 seconds).
14:19:29 -!- zadock has joined.
15:00:07 -!- AndoDaan has joined.
15:28:53 -!- AndoDaan has quit (Quit: Going, going, gone.).
15:57:59 -!- bb010g has quit (Quit: Connection closed for inactivity).
16:19:21 -!- zadock has quit (Quit: Leaving).
16:20:46 -!- SopaXorzTaker has joined.
16:33:46 -!- oren has quit (Quit: Lost terminal).
16:40:25 -!- GeekDude has joined.
16:52:45 -!- J_Arcane has quit (Quit: ChatZilla 0.9.91-rdmsoft [XULRunner 32.0.3/20140923175406]).
16:55:42 <mroman> anybody working on an interesting esolang?
16:55:55 <mroman> that... uhm.. has functions and some way of #include?
16:56:07 <mroman> or anybode worked on one?
16:56:13 <mroman> preferably with an implementation available
16:58:39 -!- J_Arcane has joined.
17:02:47 -!- lleu has quit (Changing host).
17:02:47 -!- lleu has joined.
17:06:38 -!- Lymia has quit (Ping timeout: 265 seconds).
17:20:02 -!- dianne has joined.
17:35:25 -!- cpressey has joined.
17:37:33 <cpressey> so, one of things I've never liked about monads is that people are apt to talk about "THE list monad" as if there is only monad that could be sensibly applied to lists
17:37:46 <cpressey> but i think i just figured out why this happens
17:39:09 -!- g2watson has joined.
17:39:21 -!- g2watson has changed nick to orin.
17:42:55 <cpressey> i'd ask on #haskell but i half-expect to be talked down to, there
17:43:29 <cpressey> HI Phantom_Hoover HOW'S YOUR BLOG HOW'S BITCOIN
17:43:34 <elliott> it's not true for every type
17:43:40 <elliott> it is true for lists though
17:44:14 <cpressey> elliott: so it's just that "the X monad" is "the monad for X that seems to make the most sense to the most people"?
17:45:10 <elliott> cpressey: there is only one list monad
17:45:26 -!- shikhin_ has joined.
17:45:28 <mroman> what are the other list monads :D?
17:45:45 <elliott> cpressey: but generally "the X monad" is going to mean the monad that X actually has an instance for, so yeah.
17:46:17 -!- shikhin has quit (Disconnected by services).
17:46:18 -!- shikhin_ has changed nick to shikhin.
17:46:36 <elliott> cpressey: okay, let's put it this way: "the State monad" doesn't mean "the monad that State is", "State" there implicitly carries the semantics we associate with it
17:46:39 <elliott> rather than just the type itself
17:47:05 <elliott> cpressey: just like "the ring <some set>" in mathematics
17:47:39 <cpressey> i apologize for not knowing that the list object only has one possible monad (if that's true)
17:48:46 <orin> is a monad a mathematical thing
17:49:41 <Phantom_Hoover> it's a concept taken from the academic study of the austro-hungarian empire's administrative structure
17:50:32 <mroman> cpressey: the haskell-guys seem to have that opinion that you can't define another monad, yes.
17:50:39 <orin> I'm just asking which link to click on wikipedia's biguation page?
17:51:12 <cpressey> longerwordthatbasicallymeanstrivial
17:52:02 <orin> sre we talking about "Monad (category theory)" or "Monad (functional programming)"
17:52:25 -!- oerjan has joined.
17:52:37 <cpressey> "pathological" except with the connotation of also trivial
17:52:58 <cpressey> anyway n/m question about general usage of word answered thank you very much
17:53:49 <cpressey> Phantom_Hoover: BING thank you
17:53:57 <oerjan> cpressey: hi thanks to a dmm comic i realized what chrysoberyl means :P
17:54:01 -!- AndoDaan has joined.
17:54:23 <cpressey> Phantom_Hoover: oh, there is also a word you used years ago that i've been meaning to ask you about (but now i have the channel logs so i can probably find it by grepping)
17:54:51 -!- SopaXorzTaker has quit (Remote host closed the connection).
17:55:46 <cpressey> so, there are no degenerate list monads? the usual one is unique?
17:56:08 <oerjan> what would a degenerate list monad be?
17:56:16 <oerjan> Maybe could count as one...
17:56:23 <cpressey> oerjan: i feel like such a nerd since i learned it through, um, gemology
17:57:33 <oerjan> well i _actually_ learned it from wikipedia's article, but the comic pointed me there
17:59:10 <quintopia> hi chris. 'sbeen a while. anything new for us?
18:00:10 -!- zadock has joined.
18:00:32 <oerjan> cpressey: i am not sure whether the usual Monad is the only possible one for [] but i wouldn't be extremely surprised. maybe something that permutes elements could also work, but that wouldn't be "degenerate" as such
18:01:02 <cpressey> doesn't have to be strictly "degenerate" or any other adjective; just has to exist
18:01:33 <cpressey> to distinguish "the usual list monad" vs "the unique list monad" which is (apparently) what i care about
18:03:54 <oerjan> the fact that the ZipList Applicative cannot be extended to a Monad is a bit subtle, so to prove it even more generally...
18:06:31 <Taneb> It can if all lists are infinite, but then you're using different lists
18:06:50 <oerjan> yes, it also can if all lists have the same length
18:07:05 <oerjan> then you use a different return
18:10:16 <oerjan> "is the list monad unique" doesn't seem to give any sensible google hits
18:11:23 <fungot> \o| c.c \o/ ಠ_ಠ \m/ \m/ \o_ c.c _o/ \m/ \m/ ಠ_ಠ \o/ c.c |o/
18:11:24 <myndzi> | c.c.c | ¯|¯⌠ `\o/´ | c.c.c | `\o/´ ¯|¯⌠ | c.c.c |
18:11:24 <myndzi> /< c.c >\ /< | | >\ c.c /| | /< |/^\ c.c /|
18:12:05 <oerjan> (belated demonstration)
18:15:25 -!- ^v has quit (Read error: Connection reset by peer).
18:16:24 <oerjan> <Taneb> Does USD form a module over the ring of 100ths of integers? <-- 100ths of integers isn't closed under multiplication
18:19:05 <oerjan> <elliott> it is true for lists though <-- is there a proof around?
18:19:25 <oerjan> it would of course give the ZipList thing as a corollary
18:21:33 <oerjan> btw the State s monad mentioned is one example where it's definitely not unique: the type is isomorphic to ReaderT s (Writer s) which has a completely different instance
18:21:36 -!- FreeFull has quit (Ping timeout: 264 seconds).
18:22:38 <oerjan> (also WriterT s (Reader s), although i think those two _are_ isomorphic.
18:22:48 <oerjan> oh and s must be a Monoid.
18:23:05 <oerjan> @unmtl WriterT s (Reader s)
18:23:05 <lambdabot> Plugin `unmtl' failed with: `WriterT s (Reader s)' is not applied to enough arguments, giving `/\A. (Reader s) (A, s)'
18:23:13 <oerjan> @unmtl WriterT s (Reader s) a
18:23:51 <oerjan> @unmtl ReaderT s (Writer s) a
18:24:14 <oerjan> am i doing something wrong or doesn't it know about those
18:25:00 <oerjan> @@ @unmtl @unmtl WriterT s (Reader s) a
18:25:01 -!- FreeFull has joined.
18:25:31 <int-e> looks like it doesn't know that (a b) c is a b c
18:26:07 <int-e> so what's holding up ghc-7.10.1 now?
18:26:31 <oerjan> is it post the deadline?
18:27:59 <int-e> Yes, last Friday, I thought.
18:28:23 <oerjan> and all tickets are gone from https://ghc.haskell.org/trac/ghc/wiki/Status/GHC-7.10.1
18:28:23 <int-e> Anyway they're tweaking release notes, so there's some hope...
18:31:16 <oerjan> *although i think those two _are_ isomorphic as monads
18:36:00 <elliott> oerjan: and yeah I know State wasn't unique in that sense
18:36:18 <elliott> oerjan: though to be fair that requires Monoid s
18:36:20 <oerjan> the typeable kind bug has been fixed yes
18:36:32 <elliott> there is probably no other monad *parametric* in s?
18:36:55 <Vorpal> Random question to anyone who knows: What is the best (quality/file size ratio) lossy audio format? mp3 or ogg? Or doesn't it matter as long as you use a good encoder and reasonable parameters?
18:37:00 <elliott> maybe you can just pick the "right" one all the time
18:37:08 <elliott> no you'd still need s to have an element
18:37:38 <Vorpal> elliott, never heard of that, doubt the player on my phone supports that though
18:37:44 <elliott> probably rockbox or something does
18:38:14 <elliott> vorbis handles artifacts better than mp3, at least, so at low bitrate...
18:38:25 <elliott> but probably LAME is so good that at the kind of bitrates you'd expect to use these days it doesn't matter
18:38:33 <Vorpal> elliott, hm I'm rather attached to PowerAmp, the library search and management aspect of it is unparalleled on phones (of what I tried so far)
18:38:39 <elliott> which is strictly better than both mp3 and vorbis probably
18:39:00 <orin> mp3 is fine unless you have really good speakers
18:39:17 <elliott> just use anything it doesn't matter you won't hear any artifacts with reasonable settings
18:39:39 <elliott> orin: actually mp3 is transparent no matter what speakers you have with a good encoder/settings and a not-totally-pathological sample
18:39:48 <Vorpal> orin, well I have the music in flac currently, which is just a waste of space, so I need to convert it to something. Might as well go for the best one then.
18:39:57 <elliott> (please feel free to publish ABX results contradicting this)
18:40:13 <elliott> (please don't feel free to tell me lossless "just sounds better" without them)
18:40:53 <Vorpal> elliott, the only reason to use lossless from what I understand if you need to process it (as you would in a studio after recording the music)
18:40:59 <orin> I don't have any music that isn't mp3 so I have no data. My father *claims* that vinyl is best
18:41:14 <oerjan> cpressey: hm i think i may have a very degenerate monad that works for _finite length_ lists.
18:41:50 <elliott> your father either doesn't understand audio at all, or just likes the specific distortion and artifacts caused by vinyl
18:41:54 <Vorpal> orin, I personally hate the typical vinyl sound, it doesn't sound as clear as a CD in general. Though since it is analogue it technically has more information.
18:42:00 <elliott> (and in particular the evolving nature of them as you listen to a record over time)
18:42:21 <elliott> but you can capture any given play of a vinyl just fine in digital audio and play that back and it'll sound just as good
18:42:29 <orin> I think he just likes the music that came out in the days of vinyl
18:42:43 <elliott> oh, and vinyl sometimes has better mastering than CD releases
18:42:52 <elliott> in which case it of course sounds a lot better
18:43:36 <orin> elliott: i think i know what you mean, vinyl records are quieter
18:43:45 <Vorpal> elliott, I think the sheet music sounds superior to all of these
18:43:50 <oerjan> return x = [x]; [x] >>= f = f x; l >>= f | all ((==1).length) l' = concat l | otherwise = [] where l' = map f l
18:43:57 <elliott> if you want a quick primer on digital audio in terms of how sampling and analogue vs. digital etc. works, https://www.xiph.org/~xiphmont/demo/neil-young.html is nice
18:44:11 <Vorpal> elliott, no pesky acoustics getting in between you and the pure music
18:44:26 <elliott> orin: yeah (not always. sometimes they just use the exact same master they were going to use for the CD and press it to vinyl. except it gets actually more distorted by the processing needed to be able to put it on vinyl)
18:45:04 <elliott> really, most people are never even going to notice artifacts in an even 128 kbps MP3 encoded with a good version of LAME
18:45:17 <elliott> I do like lossless audio but just because I'm obsessive, not because it is actually better for me in any way whatsoever.
18:45:20 <Vorpal> elliott, why would they not add the most dynamic range they can? Seems stupid to not make use of the entire spectrum. And surely CD audio has enough dynamic range for anything...
18:45:41 <elliott> Vorpal: well I can't summarise the loudness war in a single IRC line
18:45:45 <Vorpal> Well anything non-pathological
18:45:47 <orin> because WE LIKE IT LOUD DUDURDUDUR
18:45:51 <Vorpal> elliott, I thought that was just radio...
18:45:58 <elliott> no, it affects CD mastering
18:45:59 -!- heroux has quit (Ping timeout: 244 seconds).
18:46:08 <Vorpal> Why? It makes little sense there
18:46:19 <elliott> just read the wikipedia article or something >_>
18:47:08 <elliott> (it's also not 100% objectively terrible, really, it produces a certain sound. lots of compression is obviously wildly inappropriate for certain music, though.)
18:47:10 <Vorpal> elliott, so, I'm probably going for lame, since I remember there was/is an issue with cover art in ogg using PowerAmp.
18:47:43 <elliott> just use one of the --preset options and leave it at that
18:47:46 <elliott> (the default should be just fine)
18:47:53 -!- heroux has joined.
18:48:23 <Vorpal> elliott, I would like a feature to do compression up to a certain minimum level on the fly. That would be useful when feeding music into the car stereo while driving and you hit a quiet piece you can't hear about the engine. And when you turn it up and it gets to a loud section you suddenly can't hear the car engine instead
18:48:30 <elliott> --preset medium if you want to save some space with most likely no perceptible loss of quality, --preset standard if you want some kind of stronger transparency guarantee
18:48:54 <Vorpal> You said "good version" of lame
18:48:55 <Vorpal> LAME 64bits version 3.99.5 (http://lame.sf.net)
18:49:04 <Vorpal> Or should I try to update it
18:49:12 <Vorpal> This is debian stable after all
18:49:56 <Vorpal> Well Ubuntu 14.04 LTS has the same version. So screw it
18:50:00 <elliott> "anything from the last decade"
18:50:24 <Vorpal> Hah this is the last version
18:50:28 <elliott> I just mean that in the Old Days mp3s actually did suck, because all the encoders did.
18:50:45 <elliott> and you actually did need super-high bitrates to get reasonable results.
18:50:53 <Vorpal> Wow, a package where debian stable has the current version
18:51:08 <elliott> tbf that version was released over three years ago
18:51:12 <Vorpal> elliott, is that why loudr still provides 320 kbps MP3 download...
18:51:18 <elliott> LAME is one of those programs that is pretty much complete
18:51:31 <elliott> Vorpal: no that's just because people want them
18:51:53 <Vorpal> elliott, I would prefer VBR, bandcamp has that iirc, it is a bit smaller...
18:52:05 <elliott> yeah but people see 320 and think "that's the maximum, it must be better".
18:52:16 <Vorpal> loudr just provides 320, FLAC and apple lossless
18:52:41 <elliott> right. that makes them look higher quality than saying "~190 kbps" and it's also simpler.
18:52:58 <Vorpal> And I'm too lazy to convert to something sane (in which case you should probably start with lossless anyway?)
18:53:01 <orin> I usually get my music from youtube so my perceptions might be ... pedestrian
18:53:11 <elliott> I mean, listening to music is a subjective psychological experience. so if people like their placebos, then cool.
18:53:24 <Vorpal> I actually prefer bandcamp as a platform when the music is available there
18:53:32 <elliott> Vorpal: I wouldn't transcode but you could probably do it without losing too much if you needed to save space
18:53:44 <orin> I usually can't tell the difference between a CD and the ipepd mp3
18:53:49 <elliott> "too much" = "anything noticeable", I guess. I don't know how bad transcodes really are with decent source material and settings.
18:54:12 <Vorpal> elliott, well, I have a 64 GB memory card in my phone, 17 GB of that is music, about 5 is various other stuff, and the rest is free
18:54:18 <Vorpal> so I'm not quite hurting for space yet
18:54:54 <Vorpal> elliott, also covert art embedding is really stupid. Because that will be the same across an album. And now you have 10+ copies of the same image
18:55:08 <elliott> Vorpal: you can just put cover.jpg or album.jpg or whatever the names people use are in the same folder
18:55:12 <Vorpal> elliott, On the other hand, cover.jpg files show up in the bloody album app on the phone
18:55:21 <elliott> anyway the cover is going to be smaller than the audio data :P
18:55:46 <Vorpal> elliott, sure, but this album has 1500x1500 png cover art embedded!
18:55:49 <orin> just reduce the cover art to 64x64 and put it in
18:55:51 <Vorpal> Which is pretty stupid
18:56:19 <Vorpal> orin, A bit more than that given the screen. I usually go for 500x500, that works well when blown up on the lock screen as well
18:57:00 <Vorpal> But yes I have seen <10 MB music files with cover art of 2 MB
18:57:34 -!- Lymia has joined.
18:58:41 <Vorpal> lame -V2 input.wav output.mp3
18:58:49 <Vorpal> Not recommending --preset in --help?
18:59:33 <Vorpal> Whaat... lame can't read flac? How annoying, oggenc can iirc
19:00:01 <orin> convert to wav first?
19:00:09 <Vorpal> Well yes, but that is annoying.
19:00:21 <Vorpal> find . -iname '*.flac' -print0 | xargs -0 -n1 -P4 lame -V2 has a certain beauty
19:00:31 <elliott> Vorpal: -V2 is --preset standard
19:00:59 <elliott> http://wiki.hydrogenaud.io/index.php?title=LAME#Recommended_encoder_settings
19:01:07 <Vorpal> I wonder how to convert to wav without touching ffmpeg (spending 10 hours reading man page)
19:01:32 <elliott> tbh though I would just use one of the three presets in http://wiki.hydrogenaud.io/index.php?title=LAME#Technical_information that aren't the 320 one
19:02:02 <orin> I use avconv (ffmpeg's new name) for a lot of stuff
19:02:04 <elliott> twiddling any more nobs than a -V switch is pointless and harmful, anyway
19:02:17 <elliott> orin: that is a very flamewarry way of describing the ffmpeg/libav split :P
19:02:17 <Vorpal> orin, avconv is not ffmpeg's new name. That is a fork and what not
19:02:22 <elliott> (I think Debian is switching back to ffmpeg?)
19:02:32 <Vorpal> I think I heard that too elliott
19:02:44 <Vorpal> Also I always had issues with avconv compared to real ffmpeg
19:02:58 <orin> anyway I use it for taking out pieces of a video
19:03:36 <Vorpal> Yes it definitely can be used for that. But it is a swiss army knife, and I don't use it enough to remember the complicated syntax
19:03:54 <elliott> ffmpeg/avconv should just use the LAME library, anyway
19:04:02 <elliott> as in, it almost certainly does
19:04:08 <Taneb> <oerjan> <Taneb> Does USD form a module over the ring of 100ths of integers? <-- 100ths of integers isn't closed under multiplication <-- bah, foiled again
19:04:11 <elliott> Vorpal: flac can decode to stdout btw
19:04:16 <elliott> so you can probably just pipe that into lame
19:04:54 <Vorpal> elliott, then I will have to deal with messy file names with space. Also parallelizing it would be annoying
19:05:35 <elliott> what is this about filenames
19:05:43 <elliott> I just mean flac ... filename | lame ... rather than lame ... filename
19:06:16 <Vorpal> elliott, yes, sure, but that means lame doesn't auto-set a reasonable output file name based on input, which it does when just provided a file name
19:06:26 <cpressey> oerjan: i'm just fascinated that it seems to be an open(ish) question
19:06:42 <Vorpal> for i in *.flac; do sox "${i}" "${i/.flac/.wav}"; done # not parallel, but quite fast
19:06:48 <Vorpal> find . -iname '*.wav' -print0 | xargs -0 -n1 -P4 lame -V2 # speeeed
19:07:07 <cpressey> I mean I could see the Maybe monad being unique
19:07:21 <elliott> it's not so much an open question as nobody has bothered writing it up, as far as [] goes, I think
19:07:26 <orin> Vorpal how many cores do you have???
19:07:32 <elliott> it's one of those things that's "just obvious"
19:07:38 <Phantom_Hoover> has someone found http://blog.jle.im/entry/a-non-unique-monad-instance yet?
19:07:39 <elliott> if I am proved wrong I will eat several hats
19:07:42 <Vorpal> orin, which is why I used -P4
19:07:53 <elliott> Phantom_Hoover: yeah not all types have a unique monad instance
19:08:00 <elliott> though some of them have them unique up to a choice of monoid or such, I suspect
19:08:07 <orin> then with 2 programs per run, use P2?
19:08:10 <Vorpal> elliott, Though .wav lost the metadata, of course
19:08:18 <Vorpal> Well that can be copied
19:08:26 <elliott> cpressey: for something like State, the thing is that we associate semantics with State s a beyond just Thing s a = s -> (a, s)
19:08:52 <elliott> so when we say "the State monad", we're bringing the intended semantics in, whereas "the Thing monad" would be legitimately ambiguous
19:13:03 <Taneb> (,) Integer has multiple valid Monad instances
19:13:19 <Taneb> Eg, the Sum writer, the Product writer
19:13:36 <fizzie> Vorpal: I tried out the Android photosphere and panorama things. But the subject matter was such that you can't really judge stitching quality or anything.
19:13:42 <elliott> that falls under "unique up to a choice of monoid"
19:15:05 <Taneb> elliott, ah, I missed the goalposts moving
19:15:27 <Vorpal> elliott, this is strange, I'm using X11 forwarding, and a bloody wine program runs better over it than a qt program
19:15:28 <elliott> well I never claimed that every type has a unique monad instance
19:15:35 <elliott> I knew that counterexample from the start
19:15:54 <orin> Qt is probably poorly written then
19:15:55 <fizzie> Vorpal: This is directly out of the device: https://dl.dropboxusercontent.com/u/113389132/Misc/20150324-pano_20150321_221800.jpg -- I think you could stitch that up pretty much as badly as you like and it still wouldn't be too noticeable.
19:16:01 <elliott> Vorpal: that doesn't surprise me
19:16:03 <Taneb> Well, howabout "Sum Integer -> (Sum Integer, a)" has a more interesting non-uniqueness
19:16:08 <elliott> wine's UI stuff is a lot simpler
19:16:19 <Taneb> ie, State monad and composition of Reader and Writer
19:16:29 <elliott> Vorpal: and also I suspect Qt don't care about X11 forwarding
19:16:37 <elliott> since it's unpleasant and Wayland
19:16:58 <Vorpal> elliott, I think this qt is too old. Thought it is probably PyQt as well on top of everything
19:17:16 <fizzie> (Okay, there's a noticeable jagged edge in the pillar in front of the stage.)
19:17:21 <FireFly> Speaking of audio codecs I found http://xiph.org/~xiphmont/demo/neil-young.html to be a good read
19:17:34 <Vorpal> I'm forwarding from debian stable, I doubt it is too waylandish yet
19:17:59 <orin> has anyone here ever heard of "A Logical Approach to Discrete Math0"
19:18:30 <orin> by David Gries and F.B Schnieder
19:19:21 <elliott> Vorpal: I just mean nobody cares about X11 forwarding because that's not how networked GUIs will work under Wayland, which is The Future that everyone is preparing for.
19:19:48 <elliott> (rapidly becoming the present, really... I think Fedora plans to partially switch next version and then fully switch by default the version after?)
19:20:03 <elliott> FireFly: hehe, I linked that earlier
19:20:08 <Vorpal> elliott, right, I'm just trying to quickly access a GUI program on a different computer, and no x11vnc won't work well (because the target screen is like 3 times as big as the laptop I'm at)
19:20:27 <Vorpal> elliott, hm what happens to catalyst and the nvidia drivers?
19:20:30 <elliott> it doesn't get into lossy compression or anything but it's like half of what I know about digital audio
19:20:47 <elliott> Vorpal: you could use xpra
19:20:56 <elliott> Vorpal: and those drivers do kernel modesetting anyway
19:21:03 <Vorpal> elliott, for forwarding? Never heard of it, will look into it
19:21:07 <cpressey> elliott: well yes; that's similar to "the ring of reals" thing; when people say that, I know they mean "the usual ring over the reals" even though there are probably others
19:21:10 <elliott> yes you have I told you about xpra before :P
19:21:29 <elliott> cpressey: well, are there others? it depends on what you mean by "the reals"
19:21:40 <cpressey> elliott: yeah, i mean, i don't know
19:21:44 <Vorpal> what about modesetting? Does that mean the drivers will work out of box for opengl under wayland?
19:21:46 <elliott> it can't mean a set, because there are multiple constructions of the reals and only set theorists care about the difference
19:22:11 <Vorpal> I assumed the user space component (libGL, libGLU and so on) would need updating to work with wayland
19:22:14 <cpressey> elliott: pretend i said "rationals" if you like :0
19:22:20 <elliott> it's basically like what "the X" means in category theory -- "we know that all Xs are, say, uniquely isomorphic, so 'the X' just means whichever X you want, because there is only one in the ways we care about"
19:22:21 <Vorpal> And I thought the drivers provided them
19:22:40 <elliott> so "the reals" probably means "the ordered Archimedean field"
19:22:43 <orin> holy jesus "XWayland is an X Server running as a Wayland client" oh god why???
19:22:56 <elliott> orin: so that you can run X11 applications on a Wayland desktop?
19:23:16 <Vorpal> orin, backward compatibility?
19:23:22 <orin> and we need to redirect everyhtinh thru another program!?!??
19:23:22 <elliott> why do nerds get so worked up about simple things... it's literally just software, chill :p
19:24:00 <Vorpal> elliott, I should try that response at work :D
19:24:38 <elliott> orin: ok, since you must be an expert to know how obviously unreasonable this is, what would you do?
19:24:43 <Vorpal> "We are not meeting the real time scheduling deadline!" "It's literally just software, chill :p"
19:25:35 <Vorpal> (Yes I work on real time control software, partly safety critical as in human lives, partly just "we lose a lot of money if it doesn't work")
19:26:21 <elliott> hopefully the level of discourse at your workplace is slightly higher than holy jesus oh god why
19:26:30 <orin> elliott: I would have the new display thingy support the old API directly.
19:26:41 <elliott> orin: that's not how it works. X11 is a network protocol
19:27:03 <elliott> also, the whole point is to drop the cruft of the X11 protocol from the core display server.
19:27:18 <elliott> (yes, multiple libraries are used to access X11 servers. e.g. Xlib and XCB.)
19:27:46 <elliott> also, there's not really such a thing as "the Wayland server".
19:28:02 <elliott> there's just Wayland compositors, which are relatively simple to write, because they don't have to do much.
19:28:17 <elliott> ...but you knew all this before dismissing it, right?
19:28:23 <Vorpal> elliott, yes, obviously. Though missing two scheduling deadlines in a row (or more than x in y, which I don't remember on the top of my head) will trigger an emergency stop. So that is probably not the worst way for stuff to fail in. And it failed in the lab (with simulated actuations) like that of course at some point or other
19:28:43 <orin> well my key issue here is "will this change cause my stuff to run slower or possibly not work?" and the answer seems to be "yes"
19:28:46 <elliott> Vorpal: please don't get me killed :P
19:28:59 <Vorpal> elliott, do you work in a mine?
19:29:32 <elliott> orin: I don't see any reason for a non-negligible speed penalty. also, GTK and Qt and SDL and ... support Wayland natively, so most things will not go through the X11 forwarding soon enough.
19:29:47 <orin> just like pulseaudio caused problems for liek 5 years before is worked
19:29:50 <Vorpal> elliott, if you don't work in one, then you are probably safe from me at least.
19:29:56 <cpressey> elliott: yes, also, that's the point (when you just start making general statements about structures) where I usually get lost :)
19:30:01 <elliott> I think the current default GNOME Wayland setup actually uses XWayland for all the applications rather than native Wayland and it works just fine.
19:30:10 -!- cpressey has quit (Quit: leaving).
19:30:12 <elliott> the overhead should be roughly nothing
19:30:19 <oerjan> <elliott> it's not so much an open question as nobody has bothered writing it up, as far as [] goes, I think <-- did you see my finite length only counterexample above
19:30:27 <elliott> I think you're just looking for an excuse to complain though so whatever
19:30:29 <Vorpal> elliott, what about gaming performance through XWayland though?
19:30:42 <elliott> Vorpal: well it's still going to be using GL, right
19:30:58 <elliott> maaaaybe there'll be one additional copy?
19:31:03 <elliott> but I'm not sure, I think there might not be
19:31:04 <Vorpal> elliott, input lag though?
19:31:17 <elliott> it's not like xwayland is some huge bulky thing
19:31:29 <orin> elliott: essentially I'm aware that ANY change to software tends to cause users problems (NEW probles which don't have workarounds yet)
19:31:35 <Vorpal> Not that there weren't extra layers before...
19:31:36 <elliott> it's just like running an X11 server except instead of talking to graphicsy KMS stuff directly it talks to a wayland compositor
19:31:52 <elliott> orin: ok. then I suggest running debian stable
19:32:20 <elliott> oerjan: that is a strange instance
19:32:20 <orin> I run Xubuntu stable and I rarely update it
19:32:32 <elliott> orin: great. then you don't have to complain about changes because you won't get them
19:32:33 <Vorpal> elliott, also btw, real time systems are so nice to work with, you never see a GUI freeze up. The system just flows nicely all the time, irrespective of load. I'm seriously considering switching to a rt-patch kernel...
19:32:42 <Vorpal> On my personal computer I mean
19:32:58 <Vorpal> Still, that won't solve all of it of course
19:32:59 <elliott> Vorpal: I could never work on anything directly connected to people's lives :/
19:34:04 <Vorpal> elliott, good testing and enough safety layers that if any one of the layers fail everything will stop in a safe way is the key to good sleep while programming real time systems
19:34:46 <Vorpal> And hardware interlocks for the really critical parts of course.
19:35:59 <elliott> please don't posix_madvise any cars I am in
19:36:19 <Vorpal> elliott, If someone hits the emergency stop button, that breaks a constant current and electrically cuts off the output. Then the control software is informed of this, but it is not involved in actually halting the actuations. This is what people refer to as fail-safe
19:36:30 <Vorpal> Well, one specific example of fail safe
19:36:41 <orin> do cars even use UNIX?
19:37:05 <elliott> an expensive car probably has a couple of linux systems in it these days I think
19:37:10 <elliott> (but not for anything safety-critical)
19:37:16 <elliott> okay, maybe not "probably"
19:37:33 <Vorpal> elliott, there may be real time linux systems, but the brakes won't be running on Linux
19:37:39 <fizzie> The elevator in a hotel I was in recently had this thing where, for maybe 10-15% of trips, it started moving without closing the doors, went up for maybe a cm or two, then fell back and stopped; then closed the doors and did the normal thing.
19:37:43 <fizzie> It felt exactly like a hardware interlock stopping the software from doing a stupid thing, and didn't really leave me at ease.
19:37:49 <elliott> Vorpal: I was thinking for, like, in-car touchscreen type things.
19:37:51 <Vorpal> elliott, nobody has safety certified real time linux, that is the reason for that
19:37:52 <oerjan> <elliott> so "the reals" probably means "the ordered Archimedean field" <-- *+complete
19:38:02 <elliott> oerjan: I knew I was forgetting something
19:38:43 <Vorpal> elliott, we use real time linux for some modules, but not for the modules which have been deemed safety critical. They run on some expensive certified RTOS.
19:39:08 <Vorpal> Which begs the question... Which code is actually most well tested? That expensive RTOS, or Linux
19:39:14 <orin> I once saw a restaurant sign booting up... it was, for unfathomable reasons, RedHat.
19:39:18 <Jafet> http://www.independent.co.uk/life-style/gadgets-and-tech/researchers-hack-cars-to-remotely-control-steering-and-brakes-8733723.html
19:39:20 <Vorpal> One is certified yes, but what does that actually mean?
19:39:53 <elliott> Vorpal: about as scared as always under late capitalism
19:39:53 <orin> Who certifies it?
19:40:19 <Vorpal> elliott, I can almost guarantee it is not formally verified in most cases. Not an entire OS. Specific algorithms maybe.
19:41:25 <Vorpal> Jafet, I wonder if the button to lock/unlock the car doors on the key is secure, it probably isn't using high end RSA keys, lets put it like that...
19:41:42 <Vorpal> Depending on the car, a replay attack might work
19:41:51 <orin> elliott: I like that "era name": `In 2015, as the Late Capitalism Era continued to smolder...'
19:41:53 <elliott> have you used one of those fancy RNG card things for banks
19:42:10 <elliott> where you press a button and you put in the number it gives you into your bank's website
19:42:15 <elliott> I've always wondered exactly how they work
19:42:23 <elliott> they'll spit out like 5 numbers in a minute before telling you to wait for the next one
19:42:29 <orin> I have never heard of such thing
19:42:31 <elliott> I can think of many schemes you could use
19:42:32 <Vorpal> elliott, I log in using my debit card, which I put into a card reader with a small display and a numpad
19:42:46 <Vorpal> elliott, and that is challenge reply
19:42:58 <Vorpal> Secure enough? Maybe, who knows...
19:43:00 <elliott> the thing is, the card I am talking about is not networked at all
19:43:09 <elliott> it is just a button and a display
19:43:31 <orin> elliott: giant one-time pad on flash memory?
19:43:33 <elliott> *all* communication must necessarily be one-time, before the card is sent out to the customer
19:43:42 <Vorpal> elliott, well nor is the one I'm using. I get a 9 digit code on my screen, type it into the card reader.
19:43:54 <elliott> orin: maybe... the bank would have to store all the random data though
19:43:56 <Vorpal> elliott, quite similar to this one https://gfs.nb.se/privat/bilder/ovrigt/card_reader_logincard.jpg
19:44:13 <elliott> orin: it could just store a seed, but seeking to the "next" number might be slow
19:44:16 <Vorpal> elliott, then the card PIN, and I get a 9 digit replay code
19:44:17 <elliott> like I said I can imagine tons of systems
19:44:30 <elliott> but I wonder which one is actually used
19:45:10 <Vorpal> elliott, what about these: http://www.tokenguard.com/images/tokens/SID700.gif I have that for VPN to work, it gives a new code every 30 seconds... But how do they keep the clock synced?
19:45:25 <elliott> Vorpal: right, it is that kind of thing
19:45:46 <elliott> isn't that the thing that uses dual ec drbg because of the nsa
19:46:01 <Vorpal> elliott, hm maybe? It looks quite similar at least
19:46:30 <elliott> "The RSA SecurID token is a rather old, already broken and a rather expensive solution." hm
19:46:34 <elliott> well anyway one way to do it could be like...
19:46:43 <Vorpal> elliott, you need more than just that one though, password as well
19:46:47 <elliott> you have a key shared by the bank and your card
19:47:00 <fizzie> I had a SecurID at Nokia.
19:47:01 <Vorpal> elliott, for https://gfs.nb.se/privat/bilder/ovrigt/card_reader_logincard.jpg you mean?
19:47:07 <elliott> the card spits out some jumbling of the key with the current minute you're in, and an index
19:47:12 <elliott> (like "the Nth code of the minute")
19:47:23 <elliott> and then your bank just probes that space
19:47:27 <elliott> I think it's actually rate-limited though
19:47:43 <elliott> like pressing it repeatedly gets you fewer codes in a minute than pressing it slowly over the course of the minute? I'm not sure
19:47:51 <elliott> Vorpal: no for the thing I'm talking about
19:47:53 <oerjan> <elliott> oerjan: that is a strange instance <-- basically i define everything to be [] except what the first and second monad laws and fmap law force not to be
19:47:58 <elliott> don't worry I don't care about your bank's system at all :p
19:47:59 <izabera> i just noticed that there are both edirc and erdic o_O
19:48:37 <Vorpal> elliott, the one I have is rate limited and the challenge is time limited in it's validity, but not time limited in the generation algorithm (I tried the same challenge twice with a few minutes in between on the device, gave the same response)
19:49:08 <elliott> izabera: there used to be both newsham and news-ham but that was my fault.
19:49:12 <Vorpal> of course, a challenge is only valid for one entry on the website though
19:49:36 <fizzie> And the bank account has a card reader with three different functions. One is plain identification, one a challenge-response and the last one a sign-a-payment thing. They differ in how many things you need to type in.
19:50:12 <elliott> izabera: it would have been good if you made edirc as a pun on erdic though.
19:50:35 <izabera> ha ha erdic it was totally a pun
19:51:03 <Vorpal> fizzie, there are 4 buttons on the one I have (login, sign, buy and something I don't remember), only ever use two though (login and sign)
19:51:37 <Vorpal> though all are challenge-response based
19:52:04 <fizzie> The "identify" on mine only needs the PIN.
19:52:21 <Vorpal> for me it is challenge + PIN for every function
19:52:48 <fizzie> ISTR challenge wants one code, and sign a code and a monetary amount.
19:53:09 <Vorpal> though challenge length can vary on a given function depending on what you transfer. Like when I made a bit transfer I had to type in the sum I transferred as well as the usual random numbers
19:54:28 -!- AndoDaan has quit (Read error: Connection reset by peer).
19:54:52 -!- AndoDaan has joined.
19:56:02 <fizzie> I got a Google account phishing email at my personal (non-gmail) address the other day.
19:56:31 <fizzie> It wasn't very professional, although they had at least copied the logos.
19:57:15 <Vorpal> fizzie, I assume you use two factor auth for google?
19:57:48 <fizzie> I'm afraid I can't comment, but common sense says that would be a reasonable guess.
19:58:02 <elliott> is that covered under your NDA too
19:58:21 <fizzie> Not explicitly. It's just very generic.
19:58:34 <Vorpal> well a lot of sites have been pushing it lately
19:58:57 -!- Patashu has joined.
19:59:25 <Vorpal> google, github, gandi, linode, launchpad too iirc(?)
19:59:35 <Vorpal> and I'm sure I have seen more offering it
19:59:54 <Vorpal> Did bitbucket offer it? I can't remember
20:00:13 <fizzie> We do support it for regular Google accounts. I think you can also use a Yubikey: https://www.yubico.com/products/yubikey-hardware/yubikey-neo/
20:00:33 <fizzie> If you want to pay for the hardware, that is.
20:00:54 <Vorpal> fizzie, issue with yubikey iirc, is that it is limited to the number of accounts it can track?
20:01:33 <fizzie> Maybe? I don't know details.
20:01:38 * int-e growls at Microsoft.
20:02:00 <fizzie> I wouldn't have expected, but...
20:02:09 <elliott> yubikey supports like a billion things
20:02:19 <elliott> https://www.yubico.com/products/yubikey-hardware/
20:02:28 <elliott> even PGP if you buy the right one
20:03:08 <Vorpal> elliott, hm maybe it was the original one that was fairly limited then
20:03:26 <Vorpal> And then there is the new U2F thingy
20:03:46 -!- L8D has joined.
20:03:53 <elliott> yes, yubikey supports that
20:03:57 <elliott> they even have one that just does that
20:04:00 <int-e> (We're switching our calendars to Exchange... so you can mark appointments as private. Common sense would suggest that this means that the details of the appointment are only visible to yourself. And it seems to be true, as long as you use MS' own clients. But the server happily gives the information out, and the EWS provider for Thunderbird does *not* filter the information of other people's...
20:04:07 <int-e> ...private appointments...)
20:04:27 <pikhq> Yubikey is pretty delightful.
20:04:33 <Vorpal> elliott, Does the normal one support it? Then a neo might be worth buying, rather than storing the two factor auth keys on your phone...
20:04:51 <int-e> http://tinyurl.com/pz5pcje
20:04:59 -!- Lymia has quit (Ping timeout: 265 seconds).
20:05:00 <int-e> And that's what MS has to say on this.
20:05:02 -!- ^v has joined.
20:05:17 <elliott> Vorpal: I literally linked you a table with all the things the models support. :p
20:05:42 <Vorpal> elliott, yes I'm looking at it currently
20:05:55 <Vorpal> Still, I would need NFC support, and $50 is quite a bit of money
20:07:57 <elliott> what's your salary and how often would you need a new yubikey though
20:09:53 <Vorpal> Well depends on if they break backward compatibility and how long the thing lasts
20:10:12 <pikhq> They don't and apparently ages.
20:10:37 <int-e> does it support firmware updates? . o O ( If so, every time you cross the US border and they find it in their security checks :P )
20:11:13 <pikhq> And that's the reason. :)
20:11:51 <Vorpal> pikhq, well okay, but look at the original standard yubikey, it is fairly limited compared to the neo, and since they don't support firmware upgrades, I believe early neo doesn't support U2F?
20:11:57 -!- nycs has joined.
20:12:19 -!- `^_^v has quit (Disconnected by services).
20:12:20 -!- nycs has changed nick to `^_^v.
20:12:37 <elliott> int-e: they could probably do enough with just temporary access to it, really
20:13:05 <pikhq> You're not going to see new functionality for the thing, but it's not as though support for it's going to suddenly go away.
20:13:07 <elliott> it's kind of irrelevant though since they could also just demand that you do whatever.
20:13:09 <Vorpal> pikhq, so that can be considered breaking backward compatibility the day when TOTP is deprecated in favor U2F
20:13:31 <Vorpal> pikhq, and apparently U2F currently doesn't support NFC
20:13:37 <Vorpal> so no login on the phone
20:14:10 <Vorpal> "Please note, current U2F standards do not support NFC for mobile devices. For more information, you can access the full U2F Specifications on the FIDO Alliance website."
20:14:26 <Vorpal> Probably won't be possible to update a NEO to add that support?
20:16:26 -!- Lymia has joined.
20:21:46 -!- Patashu has quit (Ping timeout: 245 seconds).
20:25:23 -!- Koen_ has joined.
20:25:51 <Vorpal> elliott, hm a ubikey neo and putting your ssh keys on it as well would be cool. And client ssl certificate.
20:26:22 <elliott> Vorpal: uh, just make sure you don't lose it
20:26:30 <elliott> (I don't think you could use password-protected keys there)
20:27:01 <Vorpal> elliott, hm... well you could have a password protected offline copy stored in a bank deposit
20:27:13 <Vorpal> Or something along those lines
20:27:36 <elliott> the point is that someone would get access to all your ssh accounts just from stealing your key
20:28:33 <orin> elliott: keep it around your neck at all times then
20:29:17 <elliott> ideally it would be embedded in your skin and only removable surgically
20:29:27 <orin> (these things are waterproof right)?]
20:29:31 <elliott> (let's hope it doesn't have a tracking device in it)
20:30:02 <orin> elliott: Oh shit
20:30:18 <Vorpal> elliott, hm surely it would be pin protected as well? I think the GPG key variant when the device is acting as a smartcart requires a pin
20:30:27 <Vorpal> from some googling at least
20:30:37 <orin> well that's life in our corporatist paradise
20:31:23 <elliott> Vorpal: maybe... how do you give the yubikey the pin without giving the computer you're using license to print as many signaturse as it wants
20:32:16 <Vorpal> elliott, the issue I have is storing keys on my computer, even encrypted, what if the key is being handled in memory at the moment an attack happens
20:32:53 <elliott> it would need a bulky PIN entry thing
20:33:05 <elliott> I guess it would still require you to press it
20:33:13 <elliott> kinda yuck to give your computer the PIN
20:33:30 <fizzie> *ahem* voice input for digits *ahem*
20:33:34 <fizzie> Very good in public. :p
20:33:45 <elliott> is that what they have you working on :P
20:33:53 <Vorpal> elliott, Speaking of which, what if an attack breaks out of the browser sandbox. Then it has control of your X session... And if you have a root terminal open at that point. Ouch
20:33:55 <fizzie> I'm afraid I can't comment, but no.
20:34:34 <fizzie> I do remember reading one whisper-related research paper, but I can't recall whether it was about whispering recognition or whispering synthesis.
20:34:47 <Vorpal> fizzie, do you call old versions of google now "google then"?
20:34:48 <orin> put a touchscreen on it and use handwriting input
20:35:04 <elliott> Vorpal: I think Wayland will be a bit more secure there
20:35:18 <orin> that would be less bulky but more expensive obviously
20:35:18 <elliott> in that it will have more security than X's "none"
20:35:30 <fizzie> Yes; the way people have explained it to me, Wayland will improve on X's security model by having one.
20:35:44 <Vorpal> elliott, so now you break screen magnifiers and screenshot apps?
20:36:02 <elliott> oh no how terrible, we'll have to give like three programs special permissions :p
20:36:11 <Vorpal> Also password managers where you copy the key means that your key can be spyed on when in the clipboard
20:37:07 <fizzie> All this authentication talk makes me ashamed of not having enabled 2FA for my personal Google account. :/
20:37:23 <Vorpal> elliott, Apparently you can use gpg on yubikey, and you can use a gpg-agent for ssh login, so that way you can use a yubikey for ssh...
20:38:00 <Vorpal> Though there is no direct ssh key or ssl client cert support
20:40:25 <orin> fizzie: I just avoid having anything important on there
20:41:45 <orin> and I deal with my bank mostly physically anyway
20:43:42 <orin> It is cheap to clone a credit card, expensive to make a mask that looks like me and copy my passport
20:44:14 <orin> Although with modern 3d printing...
20:44:28 <fizzie> Vorpal: Here's the other place I did a photosphere thing: http://goo.gl/0SvqVQ
20:44:47 <Vorpal> fizzie, I didn't see the first one?
20:44:50 <fizzie> "Not shown on Google Maps. Why? Pending approval..." they don't like me :/
20:44:57 <fizzie> Oh, right. Well, it's the same place as the panorama.
20:45:38 <Vorpal> fizzie, that thing is hell on my browser on this old laptop? Can't you just link it like a normal image?
20:45:58 <Vorpal> fizzie, I don't mind it being a bit distorted
20:46:25 <Vorpal> fizzie, I think it is trying to do GL stuff, this is a core 2 duo with intel graphics, from before when intel graphics were any good
20:46:33 <fizzie> Except that I don't have it dumped from the phone yet.
20:46:48 <fizzie> I don't know if you can link to the underlying image in Views.
20:46:57 <Vorpal> fizzie, I didn't see any google maps integration in google camera?
20:47:29 <fizzie> Vorpal: Yeah, it's really strange. Apparently you're supposed to be able to "share" it from Photos directly to Maps, but Maps doesn't even show up in the sharing menu.
20:47:44 <Vorpal> fizzie, so how did you upload it to maps?
20:47:54 <orin> What is the roundhouse?
20:48:13 <orin> is that a nightclub or a theatre?
20:48:14 <fizzie> Vorpal: The "official" (as in, mentioned in the support group) workaround is to set "full-resolution uploads" on in G+, upload to a G+ post, then go to Views with a regular browser and import it in from "my posts".
20:48:21 <Vorpal> fizzie, Also I wouldn't want to share it to maps, in any case, I don't get any royalty for it from google...
20:48:26 <fizzie> orin: It's a venue, I think it's mostly about music?
20:48:48 <Vorpal> fizzie, heh I actually laughed out loud at that complicated workaround
20:49:26 <fizzie> Vorpal: I mean, it's just an image, I think you can also just upload it to Views from a computer, and use any means you like to get the image off the phone. Although I don't know if it loses some metadata.
20:49:49 <fizzie> FWIW, at least my phone makes a pretty low-res photosphere.
20:49:53 <Vorpal> fizzie, what is "Views" with a capital V anyway?
20:50:17 <Vorpal> And didn't you have a nexus 6?
20:50:27 <Vorpal> I guess low res is a way to hide the seams
20:50:47 <orin> some google thing... they've switched from prepending a G to just capitalizing everything
20:50:52 <fizzie> Vorpal: As far as I can tell, Views is the part of Google Maps that deals with photospheres. The linking of multiple ones into a navigable thing, and the web-viewer, and so on.
20:51:15 <fizzie> Vorpal: https://dl.dropboxusercontent.com/u/113389132/Misc/20150324-pano_20150321_191802.jpg is the one I had dumped out of the phone, as a regular image.
20:51:16 <Vorpal> fizzie, is this different than street view?
20:51:39 <Vorpal> fizzie, neat, what is that of
20:51:51 <fizzie> http://www.roundhouse.org.uk/
20:52:00 <fizzie> It's some old railway turntable thing, I think.
20:52:04 <FireFly> That photosphere thing is cool, how does that work?
20:52:23 -!- vodkode_ has joined.
20:52:27 <Vorpal> FireFly, turned into a venue?
20:53:09 <Vorpal> FireFly, the phone uses the gyro to keep track of where you are looking and then try to stitch together an image similar to any other panorama
20:53:32 <Vorpal> well probably gyro and compass
20:53:51 <Vorpal> FireFly, it also shows some guide points to help you get complete coverage
20:54:31 <fizzie> Hey, I can download the other photo back out of G+.
20:54:50 <fizzie> "2015 - 1.jpg". What a good file name.
20:55:02 <Vorpal> fizzie, it would be fun to try photo mapping btw, as in using photos to re-create a 3D scene with reasonable textures, it is all the rage nowdays
20:55:59 <fizzie> Yes. I even tried that out briefly back in Finland on some buildings of the university campus, but couldn't get a good enough set of photos to make the software work.
20:56:12 <fizzie> https://dl.dropboxusercontent.com/u/113389132/Misc/20150324-ps.jpg there's the other thing.
20:56:55 <fizzie> Heh, a rather nasty seam in the bridge. They look okayish at a glance, but...
20:57:43 <fizzie> I think I'll stick to panotoolsing it up. But admittedly this is more convenient.
20:58:15 <Vorpal> fizzie, oh? what software?
20:58:23 <Vorpal> I don't know what software to use
20:58:31 <fizzie> I had a good set of links somewhere.
20:58:48 <fizzie> Let's see if they're not dead yet.
20:59:07 <Vorpal> fizzie, I forgot if I uploaded mine
21:00:59 <fizzie> http://www.cs.cornell.edu/~snavely/bundler/ does the basic unsorted-images-to-sparse-3D-point-cloud thing, http://www.di.ens.fr/pmvs/ can take that further into a mesh-and-texture thing, http://www.di.ens.fr/cmvs/ is a computational-cost-reducing trick for it, and http://ccwu.me/vsfm/ is kind of a full toolchain/frontend for it all.
21:01:19 <fizzie> It's all typical research code, so at least back when I was trying, getting it to compile was a full evening's job.
21:01:59 <fizzie> VisualSFM changelog doesn't include dates, cleverly, so it's hard to tell if it's actually seen any work lately.
21:02:09 <Vorpal> fizzie, hm didn't google research thingy have a phone with stereo cameras that could do this on the fly?
21:03:11 <fizzie> I remember something else vaguely similar too.
21:03:20 <fizzie> From a Tested.com video or something.
21:03:54 <fizzie> And of course Street View does 3D reconstruction, but that's based on the LIDAR stuff.
21:04:38 <fizzie> There was a LIDAR unit at the Computer History Museum's self-driving car exhibit.
21:04:47 <Vorpal> fizzie, actually, does it do 3D reconstruction?
21:05:21 <fizzie> I'm not sure how widely they actually do it.
21:05:34 <Vorpal> You mean, some areas only?
21:07:10 -!- ProofTechnique has quit (Ping timeout: 264 seconds).
21:08:05 <fizzie> I have this vague idea (based entirely on pre-Google knowledge) that the LIDAR data from Street View cars is involved in the 3D buildings of Google Earth (which I guess is now kinda-sorta part of the new web Google Maps).
21:08:15 <fizzie> But I could be entirely wrong there.
21:08:32 <Vorpal> 3D buildings on google maps? Didn't know that
21:09:04 <fizzie> Yes. If you use the new thing, satellite imagery is replaced with something relatively Earth-like.
21:09:28 <fizzie> Except you only have three levels of tilt (straight from above, and two oblique ones) and the 4 cardinal directions for rotation.
21:09:39 <fizzie> I assume it's all WebGL.
21:09:44 <Vorpal> fizzie, not like this computer could handle it, Also wouldn't it be limited to areas with street view coverage only?
21:10:11 <fizzie> I think it's limited to much less than that. Like, major city centres.
21:13:16 <fizzie> Helsinki doesn't have 3D buildings (in Maps) at all.
21:13:25 <fizzie> Although it does in the actual Google Earth.
21:13:59 <Vorpal> fizzie, how does it handle complex building shapes? Like "not a box"?
21:14:13 <fizzie> I think it's been impressively good, actually.
21:14:28 <Vorpal> fizzie, I wonder if it based on building plans though
21:14:32 <fizzie> There's a funny thing that it's clearly somehow optimized for the view direction.
21:14:42 <fizzie> (Possibly just the render aspect.)
21:14:49 -!- `^_^v has quit (Quit: This computer has gone to sleep).
21:14:54 <fizzie> If you rotate it 90 degrees, it's very... well, it's hard to describe. But wrong-looking.
21:15:05 <Vorpal> fizzie, I'm fairly certain that the building outlines on some Swedish maps are, because they are basically identical between different providers.
21:15:07 <fizzie> Then it'll fix things up after it loads more stuff.
21:15:46 <fizzie> If this magazine article is right, the 3D buildings in Maps are based on reconstructions from aerial imagery.
21:16:27 <Vorpal> Well that I don't believe quite
21:16:45 <Vorpal> fizzie, if so I wonder what something like the Eifel tower would end up looking like
21:17:23 <fizzie> They've had the "landmark" buildings as manually constructed 3D models for quite a while.
21:17:32 <fizzie> I doubt they'll let the autogenerated stuff override that.
21:17:51 -!- `^_^v has joined.
21:18:07 <Vorpal> fizzie, but I wonder how the algorithm would handle a building like that
21:20:40 <fizzie> http://www.google.com.au/earth/explore/showcase/3dimagery.html is about the only official thing I can find.
21:21:32 -!- cpressey has joined.
21:22:29 <cpressey> oerjan: elliott: I guess an example is "the group over the rationals"; you obviously wouldn't say that because there are obviously 2 "good" candidates. and I'm pretty sure there are an infinite number of others
21:23:15 <cpressey> <Q,|,*> where * is multiplication and a|b = 2(a+b) I think is a ring, and there would be many more where that came from
21:23:59 <cpressey> and maybe a ring theorist would say "that's not REALLY different from the standard one" and well, no, it's not very interesting, they're both rings and ring theorists care more about rings than the sets of things that form the rings
21:24:11 <cpressey> programmers DO care about the concrete data, a lot.
21:24:32 <cpressey> i just needed to say that. i'll be afk for a bit
21:25:55 <fizzie> Vorpal: Oh, here it finally is. http://google-latlong.blogspot.co.uk/2012/06/never-ending-quest-for-perfect-map.html "Since 2006, we’ve had textured 3D buildings in Google Earth, and today we are excited to announce that we will begin adding 3D models to entire metropolitan areas to Google Earth on mobile devices. This is possible thanks to a combination of our new imagery rendering ...
21:26:01 <fizzie> ... techniques and computer vision that let us automatically create 3D cityscapes, complete with buildings, terrain and even landscaping, from 45-degree aerial imagery."
21:26:46 <fizzie> As for quality, it's not hard to find places where it's made approximations (for "thin" structures etc.) but it's been better than I expected.
21:29:08 <int-e> a|b = 2(a+b) doesn't look very associative
21:29:48 <int-e> or perhaps we should start with the existence of a unit
21:33:31 <int-e> is $pos-- while (substr($data, $pos, 1) =~ /[\x80-\xBF]/); okay perl or is there a nicer way of doing that? (The task is to rewind $pos to the start of a unicode character in UTF-8.)
21:35:40 -!- ProofTechnique has joined.
21:36:23 -!- L8D has quit (Ping timeout: 252 seconds).
21:37:31 <tromp> try matching the string up to position $pos against /[\x80-\xBF]*$/ and use length of match
21:40:33 -!- HMC_A has quit (Ping timeout: 250 seconds).
21:41:18 -!- AnotherTest has joined.
21:41:46 <int-e> Hmm, so ... substr($data, 0, $pos) =~ /[\x80-\xBF]*$/; $pos = $-[0];
21:44:45 <int-e> actually no, that has an off-by-one error.
21:44:57 <tromp> yep, you want $-[0]-1
21:45:15 <int-e> or /.[\x80-\xBF]*$/
21:45:57 <tromp> that is different if the string is all \x80-\xBF]*
21:45:58 <int-e> (oh but what's the value of $-[0] if there was no match...)
21:46:11 -!- fungot has quit (*.net *.split).
21:46:11 -!- zemhill_ has quit (*.net *.split).
21:46:12 -!- jameseb has quit (*.net *.split).
21:46:18 <tromp> leaving out the . ensures a match
21:46:43 -!- ^v has quit (Read error: Connection reset by peer).
21:47:01 <tromp> anyway , the nicest way involves avoiding the need to compute positions in the first place
21:47:06 -!- ^v has joined.
21:47:26 -!- HMC_A has joined.
21:47:43 -!- fungot has joined.
21:47:43 -!- zemhill_ has joined.
21:47:43 -!- jameseb has joined.
21:48:28 <int-e> thanks for the ideas
21:49:14 -!- oerjan has quit (Quit: Nite).
21:54:38 <Vorpal> fizzie, how does the 3D reconstruction thing handle stuff like trees I wonder
21:54:55 <fizzie> I think they're faked.
21:55:08 <Vorpal> fizzie, no I mean, does it filter them?
21:55:14 <Vorpal> The research stuff you talked about
21:55:24 <Vorpal> fizzie, also does it need/make use of geotagged images?
21:55:24 -!- Koen_ has quit (Quit: The struct held his beloved integer in his strong, protecting arms, his eyes like sapphire orbs staring into her own. "W-will you... Will you union me?").
21:55:48 <Vorpal> I remember some camera app for my phone that could record compass direction as well
21:55:52 <fizzie> I don't think at least the VisualSFM workflow needs or makes any use of existing geotags.
21:57:10 <Vorpal> fizzie, I will endeavor to try that stuff out at some point. You said compiling it was tricky?
21:57:19 -!- `^_^v has quit (Quit: This computer has gone to sleep).
21:58:20 <fizzie> Back then, yes. But I didn't keep any notes.
21:58:24 <Vorpal> So this was a while ago?
21:58:37 <fizzie> Yes, at least a year or two.
21:58:47 <fizzie> I think the "Ubuntu" installation guide on the VisualSFM page -- http://www.10flow.com/2012/08/15/building-visualsfm-on-ubuntu-12-04-precise-pangolin-desktop-64-bit/ -- more or less worked out on Debian.
21:59:08 -!- Phantom_Hoover has quit (Ping timeout: 256 seconds).
21:59:37 <fizzie> It must've been more like three years, because this is the same guide I remember reading, and it's talking about 12.04.
22:00:52 <fizzie> Anyway, there was nothing especially difficult, just minor nuisances. The kind of things the Ubuntu guide mentions, though I'm sure its workarounds are not the only possible ones.
22:01:56 <fizzie> Also, if you want a feeling of accomplishment, I'd suggest starting with some small physical object that you can walk around and photograph from all angles, and running the thing on that first.
22:02:21 <fizzie> Possibly even including somewhat from above.
22:02:27 <Vorpal> fizzie, I don't think I uploaded this one before: https://www.dropbox.com/s/pjrwlsdckl3wr81/PANO_20150308_132129.jpg?dl=0
22:02:36 <Vorpal> fizzie, I tried out that fish eye pano mod
22:03:23 <fizzie> Because I think the main problem with my "huge building not entirely visible from any place" thing was that it couldn't figure out how the pieces fit together, so it made me something like 20 unconnected components that were parts of the building.
22:04:00 <fizzie> So it's regular photos and just a fisheye projection for the result, or what?
22:04:11 <Vorpal> fizzie, well I don't mind if the roof is missing, it would be interesting it could accurately reconstruct the view along a street at ground level
22:04:26 <Vorpal> Even if there are some trees blocking the view every now and then
22:04:36 <Vorpal> fizzie, it is from the google camera app!
22:04:45 <fizzie> Oh! I didn't even know it has that.
22:04:56 <Vorpal> fizzie, you go to non-sphere panorama and then you switch from landscape mode
22:04:58 <fizzie> VisualSFM did pieces of different walls for me, it just didn't combine them into a single whole.
22:05:38 <Vorpal> fizzie, it is the [box] ... button you click on, and then you select the circle icon
22:05:42 <fizzie> Ohhh. I didn't even see that menu.
22:05:50 <Vorpal> No it wasn't really explanatory
22:05:57 <Vorpal> I had no idea it was doing fisheye
22:06:29 <fizzie> Do you know what the pillow-shaped thing is? The guide points looked like a 3x3 grid, so just a wideangle thing?
22:06:51 <Vorpal> fizzie, I think so, though I'm not 100% sure
22:07:05 <Vorpal> fizzie, by the way, how does the google camera app figure out the lens parameters?
22:07:14 <fizzie> Wow, the fisheye has a lot of points.
22:07:17 <Vorpal> I would love to figure those out myself for use with hugin
22:07:52 <fizzie> You can optimize for them, up to some degree.
22:08:01 <fizzie> I don't know how the stock camera app does it, though.
22:08:47 <Vorpal> fizzie, that is not the stock camera on my phone though
22:08:52 <Vorpal> so that is why I wonder
22:09:03 <Vorpal> since the data in question is not in the normal EXIF data
22:09:10 <Vorpal> unlike what my real camera gives me
22:09:27 <fizzie> My old phone (the N900) put some lens stuff in the EXIF data.
22:09:45 <fizzie> Although I think it was missing some stuff, because Hugin didn't know the focal length multiplier.
22:09:46 <Vorpal> fizzie, here is another phone camera image https://www.dropbox.com/s/p8nfprekuyjupby/PANO_20150314_150829.jpg?dl=0
22:10:16 -!- Phantom_Hoover has joined.
22:14:24 <cpressey> alright, so | as I defined it isn't associative.
22:15:01 <cpressey> so is there only one ring over the rationals?
22:15:13 <cpressey> that seems so unlikely, somehow
22:16:28 -!- AndoDaan has quit (Quit: Going, going, gone.).
22:17:06 <Vorpal> fizzie, so here is a bunch more panos https://www.dropbox.com/sh/eejt7nnbu5ephrq/AAAR-Jrfg0V8nFPwYKG8_VNGa?dl=0
22:17:57 <Vorpal> fizzie, also a normal photo of amazing ice
22:19:18 <orin> cpressey: perhaps there's only one ring over rationals where multiplication is comutive?
22:20:54 <cpressey> what about <Q,+,|> where a|b = 2*a*b
22:21:14 <cpressey> but this is kind of straying from my main point
22:22:02 <cpressey> which i don't remember well enough to put into words at the moment
22:22:17 <fizzie> Vorpal: Fancy. Although the fisheye images kind of look "too perfect" to be realistic fisheye images.
22:22:44 <fizzie> Vorpal: No chromatic aberration near the outer edge and so on.
22:23:01 <cpressey> quintopia: I stand corrected. I do have something new. But it's not terribly esoteric: http://catseye.tc/installation/Matchbox
22:23:54 <Vorpal> fizzie, what do you think about the iced branches above the stream?
22:24:08 <Vorpal> fizzie, it looked more amazing when you were there
22:27:01 <fizzie> V. fancy, too. Running water and ice tend to be a good combination. I think there was some photography collection at the Yle (Finnish public broadcasting thing) website once.
22:27:07 <fizzie> No ice or snow hereabouts.
22:27:27 -!- Phantom__Hoover has joined.
22:27:58 <Vorpal> fizzie, two more normal images of fancy ice added (from a while ago, might have showed it earlier?)
22:32:33 <Vorpal> fizzie, ooh found another nice image I haven't uploaded, might have some seams though
22:33:18 <Vorpal> still uploading though
22:33:41 <orin> Ok nevermind Wayland is ok... http://en.wikipedia.org/wiki/Mir_%28software%29
22:34:23 -!- AndoDaan has joined.
22:34:29 <Vorpal> fizzie, https://www.dropbox.com/sh/eejt7nnbu5ephrq/AAAR-Jrfg0V8nFPwYKG8_VNGa?dl=0#lh:null-PANO_20150222_143800.jpg
22:35:46 <orin> What projection is that?
22:36:22 <Vorpal> orin, not sure, it is the google camera app photo sphere thingy
22:36:40 <Vorpal> fizzie, that is on the top of a ridge that road goes over, though that isn't clear from the image I think
22:37:03 <Vorpal> I think the app was confused, because that looks decidedly flat in the image
22:37:15 <fizzie> I think the photoshpere is just regular equirectangular image.
22:37:59 <Vorpal> fizzie, anyway isn't this ice amazing: https://www.dropbox.com/sh/eejt7nnbu5ephrq/AAAR-Jrfg0V8nFPwYKG8_VNGa?dl=0#lh:null-DSC_0069.JPG
22:38:16 <fizzie> The aspect ratio is exactly 2, which would make sense for a 360x180 degree equirectangular image, and the top/bottom edge distortions look like that too.
22:38:30 <fizzie> (Since both edges are really just points.)
22:38:59 <Vorpal> fizzie, yeah is is annoying that you get varying pixel sizes in panoramas
22:39:39 <fizzie> Nice ice. No pun intended.
22:39:57 <fizzie> I was taking some macro photos of our frozen balcony glasses back in Finland, but haven't had time to put them anywhere.
22:40:14 <Vorpal> fizzie, that is phone camera, not really macro
22:40:27 <Vorpal> fizzie, from the Xperia Z2
22:41:07 -!- AnotherTest has quit (Remote host closed the connection).
22:42:55 <Vorpal> fizzie, hm I have an awkward vertical 4K video from my phone here, about 5 seconds. Yeah I don't think I'm very good at videos...
22:43:33 <fizzie> https://dl.dropboxusercontent.com/u/113389132/Misc/20150324-20100215_001.jpg this is kind of in the same category, although a bit less frozen. (From 2010, taken with the N900.)
22:44:07 -!- cpressey has quit (Quit: leaving).
22:44:25 <Vorpal> Holy crap, this 4K video is 149 MB and just 24 seconds long
22:44:29 <Vorpal> I'm not uploading that
22:45:33 <Vorpal> fizzie, it may be obvious from this image, but the path was really a stream at this point https://www.dropbox.com/sh/eejt7nnbu5ephrq/AAAR-Jrfg0V8nFPwYKG8_VNGa?dl=0#lh:null-DSC_0364.JPG
22:45:44 <Vorpal> it is also quite a steep slope
22:47:15 <fizzie> I took photos of all the Google Android sculptures they have on the lawn.
22:47:28 <Vorpal> fizzie, you should try photo mapping
22:47:46 <Vorpal> as in 3D reconstruction
22:48:21 <fizzie> I don't think I have enough frames for that, didn't really have that in mind. Which is silly, because back in Finland I was wondering where I'd find a nice recognizable shape for a statue.
22:48:25 <Vorpal> also btw, I wonder how accurate that 3D mapping is? Could you reconstruct with a street with meter accuracy?
22:49:06 <fizzie> Well, maybe next visit.
22:49:08 <Vorpal> fizzie, wouldn't recording video be perfect for this? Just extract the video frames...
22:49:17 <Vorpal> Or is the quality too bad?
22:49:33 <fizzie> There's the motion blur problem.
22:49:57 <Vorpal> You could try to filter for non-blurry images
22:50:25 <fizzie> Yes. I'm not sure if any of those 4 links I mention explicitly do video, but I'm sure I've seen some stuff related to it.
22:51:22 <fizzie> The first thing (Bundler) is associated with this fancy thing: http://www.cs.cornell.edu/projects/bigsfm/
22:51:37 <fizzie> It's about doing large-scale models from large photo datasets that weren't really intended for that purpose.
22:52:08 <fizzie> While the VisualSFM workflow is kind of more oriented around you taking a bunch of pictures of a particular thing you want to reconstruct.
22:56:55 <fizzie> People having been doing SfM from the Google Street View images, I know that.
22:57:23 <fizzie> http://ieeexplore.ieee.org/xpl/login.jsp?tp=&arnumber=5457551 is something a quick search turned up.
22:57:41 <fizzie> I don't know if we've published any papers in the topic.
22:58:46 <fizzie> Wonder how many individual photos the Street View collection is. It's probably a big number.
23:02:10 <fizzie> Heh, "Incremental plane-based reconstruction and exploration of buildings in old postcards".
23:03:50 <fizzie> Title was more interesting than the contents, to be honest.
23:04:07 <fizzie> (Not Google-related at all, just browsing.)
23:05:32 <fizzie> http://vision.princeton.edu/projects/2014/readjustment/paper.pdf is probably relevant wrt how accurate you can get with photos.
23:13:35 <fizzie> I like this stuff; it's a shame people aren't doing the same with audio.
23:14:56 <fizzie> Record some arbitrary signal with N microphones, reconstruct the 3D environment around it. Might be "slightly" more ill-conditioned problem.
23:16:44 <Vorpal> fizzie, though I was more wondering on the global measurement accuracy. Could you for example use this to improve something like openstreetmap?
23:18:58 <fizzie> How physical-reality-accurate is OpenStreetMap trying to be, anyway?
23:19:35 <Vorpal> fizzie, well, as good as possible, though it depends on the individual contributor I assume
23:20:28 <Vorpal> fizzie, I have contributed some stuff to it for example. And it depended on my input data in each case how accurate I could be.
23:20:51 <fizzie> Physical accuracy might not be absolutely critical (I mean, below the scale of, say, 1m) for many of their use cases, though.
23:21:24 <Vorpal> fizzie, but air-based imagery in hilly terrain has some issues when not taken from straight above
23:21:35 <fizzie> That's spooky, I went to maps.google.com and it knows where I am.
23:21:45 <Vorpal> In some cases, such as mapping trails in nature reserves, all I have is my GPS traces, with maybe a 10-20 meter accuracy if conditions are bad
23:21:57 <fizzie> I mean, on a desktop browser. And I don't mean the city, I mean the exact location.
23:22:18 <Vorpal> fizzie, I always click "don't share location" when I get such questions
23:22:57 <fizzie> I mean, it didn't show the "website wants to know your location" thing. Although it's quite possible I've given it permission.
23:23:16 <fizzie> Incidentally, IP geolocation puts our home DSL into Topcliffe, UK.
23:23:33 <fizzie> That's somewhere north of Leeds, and very much not where we are.
23:24:22 <fizzie> (It's something like 300 km off.)
23:24:40 <fizzie> Well, maybe 200-300. Anyway.
23:24:56 <Vorpal> fizzie, maybe google checks the ping from all their servers and uses that?
23:25:54 <fizzie> Heh. Or maybe it's tied to my logged-in Android device with location reporting. (Hey, they paid for it.)
23:26:27 <Vorpal> fizzie, anyway, you should be able to deal with hills and such when mapping using SFM on the ground I think.
23:26:35 <Vorpal> If it is accurate enough
23:26:48 <Vorpal> And you have some way to calibrate or measure a base distance
23:27:44 <fizzie> Mmmaybe. If you don't get a single frame out of the whole thing you're mapping, though, you might end up with accumulating error, like when using accelerometer-based things.
23:31:13 <fizzie> "When you activate the My Location feature, Google Maps asks your web browser for your location. Typically, your browser uses information about the Wi-Fi access points around you to estimate your location."
23:31:30 <fizzie> Maybe they know of an access point in this building, or somethin... wait a minute, this desktop doesn't have wifi.
23:31:56 <Vorpal> fizzie, I was thinking you would carefully take photos of a city block or two that is on a hill, and some features on flat terrain nearby, then you could create a 3D model, and calculate a orthographic projection straight above from it, and use that to create the map
23:32:41 <fizzie> "If no Wi-Fi access points are in range, or your computer doesn't have Wi-Fi, it may resort to using your computer's IP address to get an approximate location." Well, it's not doing that. Unless Google's getting my ISP's subscriber information or something, which sounds unlikely.
23:32:55 <fizzie> Maybe I should check in Javascript console what the JS geolocation API returns.
23:33:00 -!- ProofTechnique has quit (Ping timeout: 250 seconds).
23:33:11 -!- Lymee has joined.
23:33:19 -!- Lymia has quit (Ping timeout: 265 seconds).
23:34:21 <fizzie> Hmm. Well, doing that in the JS console did pop up the "... wants to use your computer's location" box, so they're not using that.
23:34:26 <fizzie> I'm going to guess it's device-based, then.
23:45:10 -!- shikhin_ has joined.
23:47:54 <orin> are you on a mac?
23:48:01 -!- iamevn has joined.
23:48:18 -!- shikhin has quit (Ping timeout: 265 seconds).
23:59:36 -!- AndoDaan_ has joined.