--- Log opened Mon Nov 18 00:00:00 2013
--- Day changed Mon Nov 18 2013
00:54 < Luke-Jr> warren: it's not clear that doing just the first item gets some reward
00:55 < Luke-Jr> nor that 2 and/or 3 might be done without 1, in case 1 is impossible
00:55 < Luke-Jr> 3 should probably be split up between writing a fix, and getting it merged
00:55 < Luke-Jr> ie, someone who writes a fix but doesn't have the patience for getting stuff merged should still get something
00:55 < warren> Luke-Jr: devs have power to decide apportionment, so whatever.
00:56 < Luke-Jr> warren: yes, but people might see the list and give up because they don't know how to code
00:56 < Luke-Jr> it should be clear that non-developers can contribute toward 1 for part of the bounty
18:11 < petertodd> so... headers first
18:12 < sipa> i was discussion this with petertodd
18:12 < sipa> and this question came up
18:12 < sipa> what if you know about multiple header chains whose tips are better than what you currently have
18:13 < sipa> perhaps there's this situation: A-B-c, A-B-d-e
18:13 < sipa> eh wait
18:13 < petertodd> no, that's correct
18:13 < sipa> A-B-c and A-d-e-f
18:13 < sipa> and you have A and B, but not c d e f
18:14 < sipa> do you only try to fetch blocks for d e f, or do you also try to fetch c?
18:14 < petertodd> and the same problem *is* present on A-B-c, A-B-d-e
18:15 < sipa> agree, but the case with a reorganization is more revealing probably
18:15 < petertodd> so my scenario was, suppose we have an attacker who is mining blocks, but decided to withhold the actual contents. with headers first you'll find out abotu the headers, and hence the chain, but I argue you have to try to download all tree tips simultaneously, so that you can advance your fully verified tree so the majority of hashing power can move forward
18:16 < sipa> if the case becomes A-B-c vs A-d-e-f-g-h-i-j-k-l-m
18:16 < sipa> then it's probably easier to see that you should fetch c too, just to keep up with a potentially best chain, while you're fetching the potentially even better one
18:17 < petertodd> right, because d could be invalid, as an example
18:17 < sipa> indeed
18:17 < sipa> though you did already verify PoW, so that is very unlikely
18:17 < petertodd> well... :)
18:18 < petertodd> could be all sorts of crazy economic incentives, for instance if you figures out how to get the other hashing power trying to extend different tips
18:18 < sipa> we shouldn't assume it's valid of course
18:18 < petertodd> main thing is we want an algorithm that's going to get everyone to come to consensus about what fully validated chain tip to continue mining on, regardless of what crazyness is going on with the headers
18:19 < sipa> yup
18:19 < petertodd> like, suppose we had a bug where a block somehow made the networking code crash, leaving a connection in a state of limbo
18:19 < petertodd> plausible with threads for instance
18:20 < petertodd> oh, shit, this makes the blockwithholding strategy even worse you know:
18:21 < petertodd> suppose we have A-B-C-d-e-f-g-h-i, and we have fully verified up to i and are trying to make j
18:21 < petertodd> now, if there's ever any slowdown in block distribution, we could wind up with hashing power split on A-B-C, A-B-C-D, A-B-C-D-E etc.
18:23 < sipa> well the best rational strategy is probably to mine empty blocks on top of the best header chain you know
18:24 < petertodd> yes unfortunately, modulo fees
18:24 < sipa> and never build on blocks when you know there's better header chains
18:24 < sipa> modulo fees indeed
18:24 < petertodd> but that means if someone ever loses a block entirely, we're screwed
18:25 < sipa> ewww
18:25 < petertodd> lovely 'eh?
18:26 < petertodd> also, suppose we have a fork: A-B-c and A-B-d, now bandwidth is split 50:50 downloading c and d, which makes it more likely someone will create block e, which divides the bandwidth again...
18:27 < sipa> well, if block propagation is even comparable in speed to mining speed, there is certainly a problem
18:27 < sipa> headers-first doesn't change that
18:28 < sipa> but the fact that someone could create a header, announce it, and never announce the block... worries me
18:28 < petertodd> not in general no, but in this specific case yes because of how the code would now download blocks simultaneously - that wouldn't happen before
18:28 < petertodd> although, actually, "relay all blocks including orphans" may have this affect
18:28 < petertodd> s/affect/effect/
18:29 < sipa> right, but since you *know* the header strucutre already, you can make smarter decisions in what to download
18:29 < petertodd> well, but are they actually smarter decisions?
18:29 < sipa> than what?
18:30 < petertodd> then simultaneous - again, thinking about the possibility of attack or network affecting bugs
18:32 < petertodd> for instance, suppose you always tried to download the next block in the longest chain first, and then switch to another block on a timeout, but kept mining in case the next block was invalid - if you found a block, other miners doing the same thing wouldn't build upon it because it wasn't the longest chain
18:34 < petertodd> you could have 90% of the hashing power wasting it's time, while 10% is extending a slightly longer chain just by making all your nodes artifically slow down the download of the blocks in your extension
18:47 < warren> http://www.reddit.com/r/Bitcoin/new/  please vote up "Can you fix the LevelDB database corruption bug affecting Bitcoin-Qt on some platforms? 5+ BTC bounty."
18:49 < Emcy> if you guys can find it why do you think anyone else can
18:49 < Emcy> and if they can, why are they not already here
18:49 < theymos> You don't need to be a Bitcoin wizard to find a bug in a database library.
18:50 < theymos> Even an amateur programmer has some chance of finding it if they try hard enough.
18:50 < Emcy> so just an equation of eyeballs
18:50 < petertodd> It's a lot less likely for me to find the bug than a total outsider; for one thing I don't have a Mac...
18:51 < Emcy> virtualise one?
18:51 < petertodd> Emcy: the bug is likely to do with hardware IMO
18:51 < Emcy> anyone had the bug on mavericks yet
18:52 < warren> petertodd: i doubt it is hardware, it's probably something stupid in their OS
18:52 < warren> Emcy: yes, 10.8 and 10.9
18:52 < warren> no reports with 10.7, but that might just be no users
18:52 < warren> Emcy: and some users never have the bug ever
18:52 < petertodd> warren: see, if it's fsync() I figure it's likely OS stupidity + hardware
18:53 < petertodd> warren: especially given it's not easily repeate
18:53 < petertodd> *repeated
18:53 < sipa> the weirdest report is the "uncorruption"
18:53 < sipa> where a restart fixed a corrupted database
18:54 < warren> I have a hunch for the why it effects some users but not others
18:54 < sipa> which really sounds like an OS cavhe level issue
18:54 < warren> I need a mavericks machine to try it
18:54 < warren> my only mac is Litecoin's "build server"
18:54 < warren> (an old macbook with a shattered screen in a data center)
18:55 < warren> runs 10.6.8 to match Gavin's build environment for releases
18:56 < Emcy> so you just ship then stuff like that and theyll put it in a rack for you?
18:56 < warren> Emcy: people owe me big favors
19:00 < warren> theymos: sigh, it's hard to get dev things to be voted up
19:01 < theymos> It's doing pretty well.
19:01 < warren> it is?
19:01 < warren> will 8 votes get it on the front page?
19:02 < theymos> It has several votes more than its neighbors in /new. It might eventually make the front page.
19:14 < petertodd> warren: best time to get a story upvoted is to post it in the early morning in the US
19:19 < petertodd> sipa: see, what you need is an algorithm where you take the longest block header chain, and make downloading the next block to extend your block chain towards that best header tip have the highest QoS. Of course, normal networking practice is completely unable to do that. :(
19:20 < petertodd> sipa: which makes me think just fetching roughly simultaneously is by far the simpliest, and provided block interval >> block fetch + validate time it'll be alright even if not ideal
19:20 < petertodd> sipa: doesn't really look to me like we can optimize this one without risking ugly edge cases.
19:21 < warren> theymos: inappropriate to temporarily sticky something like that?
19:21 < adam3us> petertodd: what about making it so you dont know the block hash until you get the block
19:21 < petertodd> adam3us: we're talking about a scenario where block headers are distributed separately from blocks
19:21 < theymos> warren: Yeah, I think so. Stickies are really only for super-emergencies IMO.
19:21 < adam3us> petertodd: yes so then dont do that :)
19:22 < warren> theymos: screwing over all mac users for months seems like a slow motion train wreck...
19:22 < sipa> petertodd: agreed
19:22 < petertodd> adam3us: it's for sipa's headers first code - maybe what it does suggest is that the headers first be only allowed to be used for the initial block download...
19:23 < petertodd> IE, if you receive a header on a header, you just ignore it.
19:23 < adam3us> sipa: the aim is to reduce propagation delay?
19:23 < warren> theymos: likely holding back bitcoin, as well as weakening the network.   we've had a drop in listening nodes.
19:23 < petertodd> adam3us: make initial block download (and catching up) faster
19:23 < sipa> adam3us: that wasn't the original intent
19:24 < sipa> the nicest advantage imho is not needing chwckpoints anymore
19:24 < petertodd> adam3us: also, for pruning where not everyone has all blocks
19:24 < petertodd> yeah, !checkpoints is good too
19:24 < sipa> but also harder to break
19:24 < sipa> and simplifying parallel block download
19:25 < sipa> petertodd: my former attempt used the rule that block download was only delayed in case we know we're not yet caught up
19:25 < adam3us> sipa, petertodd: if there is an attack with current blocks, maybe dont do it for the most recent 6
19:25 < petertodd> sipa: btw, you realize that in this circumstance, you actually still want *a* checkpoint, but in the form of "we know that there exists a block header at height foo with total work bar, therefore don't accept anything worse than that"
19:26 < sipa> yup
19:27 < petertodd> sipa: makes sense. The code right now basically downloads all blocks that extend tip simultaneously IIRC.
19:27 < gmaxwell> petertodd: yes, but it could be a pure difficulty "checkpoint"
19:27 < petertodd> gmaxwell: oh right, it is total work and nothing else
19:27 < sipa> there's also a "fill memory with silly low-difficupty headers" attack
19:27 < sipa> which checkpoints are still useful for
19:28 < petertodd> sipa: or commit-and-choose
19:28 < petertodd> *interactive commit and choose
19:28 < gmaxwell> sipa: I still think at some point we should do the worlds safest hardfork and increase the minimum difficulty to a million or something.
19:29 < warren> what for?
19:31 < petertodd> heh, well the fastest the difficulty can drop to 1 again is 104 weeks...
19:32 < petertodd> 30 weeks to 1 million
19:32 < gmaxwell> warren: because it makes avoiding a bunch of stupid dos attacks easier.
19:33 < sipa> petertodd: i get 51 days
19:33 < gmaxwell> and its indistinguishable unless difficulty somehow falls to that, ... which it couldn't do without leaving the network insecure in any case.
19:33 < sipa> ugh
19:33 < sipa> never mind
19:33 < petertodd> sipa: diff_now/4^(2*4*weeks) == diff_future
19:34 < petertodd> sipa: don't forget the *4
19:34 < sipa> never mind, i assumed it could drop a factor 4 every 3.5 days rather than 8 weeks
19:34 < petertodd> sipa: er, I mean (weeks/(2*4))
19:34 < petertodd> sipa: lol
19:35 < sipa> the fasteat it could get from 1 to today's diff, is 51 days
19:35 < warren> gmaxwell: would testnet remain as is?
19:35 < gmaxwell> warren: sure.
19:36 < gmaxwell> It's not a serious proposal right now, but I think we should do something like that someday.
19:36 < gmaxwell> perhaps after the hashrate stablizes again.
19:36 < warren> will that ever happen?
19:36 < gmaxwell> it should
19:39 < petertodd> gmaxwell: definitely be a reasonable thing for a SPV implementation to do
19:45 < warren> Luke-Jr: ping
19:56 < warren> Luke-Jr: you earlier mentioned 25 BTC were donated toward the deterministic linux -> mac cross compile goal.  Is that still available?  cfields now has time to focus on that.
20:13 < Luke-Jr> warren: if that's what I said before, then it should be, yes
21:41 < warren> I'm not sure why people downvoted the bounty thread.
22:01 < Luke-Jr> warren: trolls will downvote anything
22:01 < Luke-Jr> reddit seems to be nearly as bad as BCT
22:59 < Emcy> Luke-Jr what weird character is your bot using
22:59 < Emcy> 5.828… BTC @  757.6774225… BCEUSD ==     22.44 TBC observe
22:59 < Luke-Jr> Emcy: which one do you consider weird?
22:59 < Emcy> some of the numbers in the tbc section are fucked up
22:59 < Emcy> is it just me
22:59 < warren> Emcy: it's just you.
22:59 < Luke-Jr> Emcy: do you have a tonal font?
23:00 < Luke-Jr> TBC only makes sense with tonal fonts
23:00 < Emcy> troll level elevated
23:00 < Luke-Jr> …
23:01 < Emcy> http://imgur.com/eyGmhsx
23:01 < Emcy> supposed to be just numbers right
23:03 < Luke-Jr> http://luke.dashjr.org/tmp/screenshots/snapshot113.png
23:04 < Emcy> are you serious
23:05 < Emcy> what are those characters supposed to be
23:05 < Emcy> gah tonal strikes again
23:06 < Luke-Jr> http://books.google.com/books?id=aNYGAAAAYAAJ&pg=PA15#v=onepage&q&f=false
23:08 < Emcy> thats mental. someone actually made a font for it
23:08 < Emcy> surely most people cant even see them though
23:09 < Luke-Jr> there are multiple fonts with tonal support
23:09 < Luke-Jr> http://eligius.st/~gateway/products/block-erupter-sapphire
23:10 < Luke-Jr> webpages can use webfonts to solve lack of widespread font support at least :D
23:11 < Emcy> what the hell is an effect
23:12 < Luke-Jr> http://books.google.com/books?id=aNYGAAAAYAAJ&pg=PA38#v=onepage&q&f=false
23:15 < wizkid057> ?!
23:16 < Emcy> hopefully they can define a kilogramme from some sort of fundamental constant someday soon
23:17 < Emcy> then we can put all this stuff to bed once and for all
23:17 < Luke-Jr> Emcy: pfft, SI is lame
23:18 < Emcy> it may or may not be but it works
23:18 < Emcy> christ its only been 45 odd years since they decimalised money
23:19 < Emcy> ive got a a couple of shillings and tuppences in a draw somewhere......
23:19 < warren> Luke-Jr: we're still willing to make tonal default in Litecoin if you join us.
23:19 < Emcy> next up, decimal time (somehow)
23:20 < Emcy> well that probably wont happen since its based of radians for a good reason
23:20 < Luke-Jr> warren: REALLY?
23:20 < phantomcircuit> Luke-Jr, i think he's just messing with you
23:20 < Luke-Jr> Emcy: SI tried decimal time originally
23:20 < warren> phantomcircuit: maybe
23:20 < Luke-Jr> Emcy: even with all their force and threats, they couldn't make people adopt that
23:20 < Emcy> whose force and threats
23:20 < Luke-Jr> Emcy: that's how SI got adopted at all
23:21 < Emcy> did the empire spread it?
23:21 < Luke-Jr> go read up about the "metric martyrs"
23:21 < Luke-Jr> people who were put in jail for refusing to adopt it
23:21 < Luke-Jr> I think there are still some today even
23:22 < Emcy> hm
23:22 < Luke-Jr> SI has never been adopted by free choice
23:23 < Emcy> https://en.wikipedia.org/wiki/Swatch_Internet_Time
23:25  * Luke-Jr sticks with good old tonal time
23:25 < Emcy> the EU mumbled something about banning the pint from sale in the UK once. There was a good upforious shitstorm about that.
23:27 < Luke-Jr> IIRC the first time the French tried to force SI on their people, they had a revolt and had to reverse it
23:28 < Emcy> you can deny that universal SI has benefitted just about everyone though. Even if its not the best system.
23:28 < Emcy> apart from the us of course
23:29 < Luke-Jr> I think I can deny that.
23:29 < Luke-Jr> while there are benefits to having a universal measure system, there are also drawbacks of having only one system
23:29 < Luke-Jr> it's like multilingual vs single-language education
23:29 < Luke-Jr> even if you only ever use one language in your life, you benefit from having learned multiple
23:30 < Emcy> if pretty much everyone speak the one "language" though, you wouldnt need to
23:30 < Luke-Jr> personally, I think the ideal (history aside) would be universal education of tonal and dozenal
23:30 < Luke-Jr> it's not a matter of need to
23:30 < Luke-Jr> it's a matter of flexibility in your brain
23:31 < Emcy> i dont follow
23:31 < Luke-Jr> if you only know one language/number system, it's somewhat "hard coded" in your brain
23:31 < Luke-Jr> if you learn multiple, you at least have the flexibility there
23:31 < Luke-Jr> even if you don't need/use the others, it's a good trait
23:32 < Emcy> if youre talking about the cognitive benefits of biligualism, im not sure that applies to measurements
23:32 < Emcy> any more than you could get from having a decent levels of maths like we exepct from most people anyway
23:33 < Luke-Jr> there's certainly benefit from multiple number systems, even if you want to debate whether that extends to measurements
23:33 < Emcy> actually knowing hex and stuff is pretty damn useful
23:34 < Luke-Jr> there are a lot of practical application that benefits from dozenal and/or tonal units, which is why mankind has always evolved toward using tonal/dozenal units historically
23:34 < Emcy> i learned how to go from hex to denary to duonary (?) and back once
23:34 < Luke-Jr> (decimal units have only come about by unnatural means)
23:36 < Emcy> well shit if its gonna change now
23:36 < Emcy> you think bitcoin consensus is hard.......
23:37 < Luke-Jr> heh
23:37 < Luke-Jr> the great thing about tonal is that it doesn't need a consensus
23:38 < Emcy> well not when its best advocates go around breaking peoples fonts.....
23:39 < Luke-Jr> it doesn't break fonts ;p
23:39 < Luke-Jr> your fonts are just missing symbols
23:39 < Luke-Jr> easily solved by installing a better font
23:39 < Emcy> i like tahoma
23:40 < Emcy> its antialised and doesnt have any stupid serifs
23:41 < Emcy> cat just blatantly came up here and clawed me in the nipple wtf
23:42 < Luke-Jr> lol
23:42 < Emcy> why does the internet like cats again, they are murderous apex predator beasts
23:42 < Luke-Jr> Emcy: they taste good?
23:42 < Emcy> you live in the south right?
23:44 < wizkid057> florida isnt really a southern state anymore
23:45 < Emcy> i thought luke lived in georgia
23:45 < Luke-Jr> Florida
23:46 < Luke-Jr> at the moment
23:46 < whatnick> Big cats share the same tapeworms with humans
23:46 < Emcy> nice
23:46 < whatnick> http://scientiarules.wordpress.com/tag/origin-of-human-tapeworm/
23:47 < Emcy> house cats parasitise thier human hosts to better do thier bidding
23:47 < Luke-Jr> not mine :>
23:47 < Emcy> you wouldnt know
23:47 < Luke-Jr> she goes on the table, I toss her
23:47 < Luke-Jr> lol
23:48 < Emcy> well i was talking about the brain parasite they carry
23:48 < Emcy> but yeah my cat has recently decided she is finished with jumping up places and just wails until someone physically lifts her instead
23:49 < Emcy> im probably infected so i have to comply
23:49 < Emcy> she is lso beating up the dog more often
23:50 < Emcy> which i have to let happen due to dog psychology
23:53 < maaku> Emcy: can cats train dogs?
23:54 < Emcy> sure seems like they can
23:54 < Emcy> operant conditioning moderated by the claw
23:57 < Luke-Jr> lol
23:57 < Luke-Jr> debating whether to just let her live her life out spayed and alone; or let her have kitties and cook her once they weak
23:58 < Luke-Jr> wean*
23:59 < Emcy> spay unless you enjoy your house seiged by beefy tomcats
23:59 < Luke-Jr> meh, can get rid of excess kitties too I'm sure
23:59 < Luke-Jr> well, not so sure
--- Log closed Tue Nov 19 00:00:02 2013