<fleeky>
so one of our big goals was blender integration
<fleeky>
which we have , its not that great yet but its there
<fleeky>
the second video is realtime controlled inside blender
<Alpha64>
you know, using a vive or an oculus rift would be cool for this
<Alpha64>
the hand controllers are fantastic
<fleeky>
yeah i always wanted to do a vr interface for light/sound ops and maybe theater goers but i am currently in the really basic and horrible part of optimizing hardware and getting it to do what i need
<fleeky>
lots of prototypes , pcb design , failure
<Alpha64>
making hardware is hard
inetic has joined #ipfs
<fleeky>
yep
<fleeky>
ive actually been toying with the idea of making the devices ipfs of things devices
<fleeky>
but i doubt an arduino could handle that
<Alpha64>
those esp8266 I think can handle making http requests fine
<Alpha64>
so you could use the gateway ipfs.io or another
<Alpha64>
they pack 4MB of storage too, pretty amazing
<fleeky>
any idea what kind of pwm capabiliities the es8266 have?
<fleeky>
you could use the tlc59711 pwm driver and talk over spi with that i think
<fleeky>
actually we shouldnt drudge up this channel more
<fleeky>
if you want to keep talking we can talk in #7rnx
<fleeky>
about arduino robot stuff
<fleeky>
we are pretty off topic at #ipfs :)
<Alpha64>
well, using ipfs with a 2$ computer is quite on topic i think
<fleeky>
ok fair enough haha
<fleeky>
do you know if there are ethernet modules for the esp8266
<Alpha64>
it is what it is
cxl000 has joined #ipfs
<fleeky>
hmm ok there are ways
<fleeky>
esp32 is the newer cousin of this right ?
MrSparkle has joined #ipfs
<fleeky>
would be awesome to make network robot lights with ipfs , wouldnt have to worry about implementing encryption schemes
<fleeky>
currently we just communicate through telnet with *todo : figure out some encryption at some point* stub
gts has joined #ipfs
<Alpha64>
something i'm interested in is in fetching data from the ethereum blockchain with ipfs
<Alpha64>
so that a tiny device can react to it
<Alpha64>
through a gateway
kirby__ has joined #ipfs
maxlath has joined #ipfs
rozie has joined #ipfs
Alpha64 has quit [Read error: Connection reset by peer]
maxlath1 has joined #ipfs
jonnycrunch1 has quit [Ping timeout: 246 seconds]
maxlath has quit [Ping timeout: 240 seconds]
maxlath1 is now known as maxlath
jonnycrunch has quit [Ping timeout: 240 seconds]
jonnycrunch has joined #ipfs
krs93 has joined #ipfs
dhruvbaldawa has joined #ipfs
gts has quit [Remote host closed the connection]
gts has joined #ipfs
Boomerang has joined #ipfs
leavehouse has quit [Ping timeout: 258 seconds]
gts has quit [Ping timeout: 240 seconds]
dhruvbaldawa has quit [Remote host closed the connection]
ianopolous has joined #ipfs
girrrrrrr2 is now known as girrrrrrrl2
whenisnever has quit [Quit: WeeChat 1.9]
krs93 has quit [Read error: Connection reset by peer]
whenisnever has joined #ipfs
igorline has joined #ipfs
pcctw has quit [Ping timeout: 240 seconds]
gts has joined #ipfs
pcctw has joined #ipfs
kaotisk-irc has joined #ipfs
kaotisk has quit [Ping timeout: 240 seconds]
johnytiago has quit [Remote host closed the connection]
johnytiago has joined #ipfs
gts has quit [Remote host closed the connection]
ulrichard has joined #ipfs
athan has quit [Ping timeout: 240 seconds]
athan has joined #ipfs
maxlath1 has joined #ipfs
maxlath has quit [Ping timeout: 248 seconds]
maxlath1 is now known as maxlath
dimitarvp has joined #ipfs
frank___ has joined #ipfs
gts has joined #ipfs
<frank___>
hello,anyone here?
rcat_ has joined #ipfs
ygrek_ has joined #ipfs
mirek1337[m] has joined #ipfs
ylp has joined #ipfs
<r0kk3rz>
now thats an interesting question
<axx_[m]>
hi everyone… I just wanted some confirmation to make sure I understand properly:
<axx_[m]>
if one publishes a website with IPFS, then a little later realises there's a huge mistake on a page and replaces the page, there's no stopping the bad version of the page from being in existence, as long as someone in the P2P IPFS network has it, right?
erictapen has joined #ipfs
<axx_[m]>
in more traditionnal scenarios you can invalidate cache for instance, which usually replaces whatever is in cache, but in the case of IPFS it's distinct objects, and so they can get to coexist on the network/on people's computers, crorrect?
<Kubuxu>
yes
rcat_ has quit [Ping timeout: 248 seconds]
rcat_ has joined #ipfs
edulix has joined #ipfs
kaotisk-irc has quit [Read error: Connection reset by peer]
kaotisk-irc has joined #ipfs
dhruvbaldawa has joined #ipfs
maxlath has quit [Ping timeout: 248 seconds]
edulix has quit [Read error: Connection reset by peer]
edulix has joined #ipfs
dhruvbaldawa has quit [Ping timeout: 258 seconds]
maxlath has joined #ipfs
rcat_ has quit [Ping timeout: 240 seconds]
rcat_ has joined #ipfs
erictapen has quit [Ping timeout: 255 seconds]
Mateon3 has joined #ipfs
Mateon2 has quit [Ping timeout: 246 seconds]
Mateon3 is now known as Mateon1
maxlath has quit [Ping timeout: 248 seconds]
maxlath has joined #ipfs
xnbya has quit [Ping timeout: 248 seconds]
edulix has quit [Ping timeout: 264 seconds]
xnbya has joined #ipfs
dragrope has joined #ipfs
erictapen has joined #ipfs
erictapen has quit [Read error: Connection reset by peer]
dragrope has quit [Ping timeout: 248 seconds]
toppler has joined #ipfs
erictapen has joined #ipfs
<axx_[m]>
so this means an IPFS web would mean an adjustment in expectations and behaviours from content publishers… "the internet never forgets", but in this case it's immediate and automatic
erictapen has quit [Read error: Connection reset by peer]
engdesart has quit [Quit: yes]
sirn has quit [Read error: Connection reset by peer]
sirn has joined #ipfs
<r0kk3rz>
axx_[m]: you'd probably not want to give the raw address anyway, and use an IPNS address you can update
maxlath has quit [Ping timeout: 260 seconds]
<r0kk3rz>
the content would still be out there and retrievable if you looked hard enough
<r0kk3rz>
but thats also the case if the wayback machine scraped your site :P
obensource has quit [Ping timeout: 246 seconds]
obensource has joined #ipfs
Milijus has quit [Quit: Leaving]
}ls{ has joined #ipfs
quite has joined #ipfs
gts has quit [Remote host closed the connection]
johnytiago has quit [Remote host closed the connection]
maxlath has joined #ipfs
erictapen has joined #ipfs
erictapen has quit [Remote host closed the connection]
erictapen has joined #ipfs
ilyaigpetrov has joined #ipfs
jokoon has joined #ipfs
gts has joined #ipfs
<whyrusleeping>
fleeky: would be really really cool to have "make ipfs work on a $2 computer" as a goal
<victorbjelkholm>
axx_[m]: you're right. It's really like that today too though, and nothing that changes with IPFS
edulix has quit [Remote host closed the connection]
johnytiago has quit [Ping timeout: 255 seconds]
xnbya has quit [Ping timeout: 264 seconds]
erictapen has quit [Ping timeout: 255 seconds]
xnbya has joined #ipfs
xnbya has quit [Ping timeout: 258 seconds]
xnbya has joined #ipfs
vkoskiv has joined #ipfs
<vkoskiv>
Hey! New to IPFS. I'm wondering if this behavior is normal: My daemon went up to 2000+ connections, used 100% CPU on all cores and completely jammed up my home network
<vkoskiv>
I have a 100/100 connection, and an i5 6600k, running macOS 10.13, running IPFS 0.4.11
piotrekzurek[m] has joined #ipfs
pat36 has quit [Read error: Connection reset by peer]
pat36 has joined #ipfs
<lgierth>
vkoskiv: yeah we're testing patches for both the cpu usage and the number of connections -- the network has grown by 600% over the past 2 weeks and that's exposed a few bottlenecks
vkoskiv has quit [Ping timeout: 260 seconds]
erictapen has joined #ipfs
vkoskiv has joined #ipfs
ashark has joined #ipfs
shizy has joined #ipfs
inetic has quit [Ping timeout: 258 seconds]
pat36 has quit [Read error: Connection reset by peer]
pat36 has joined #ipfs
pat36 has quit [Read error: Connection reset by peer]
ccii has joined #ipfs
pat36 has joined #ipfs
pat36 has quit [Read error: Connection reset by peer]
pat36 has joined #ipfs
droman has joined #ipfs
erictapen has quit [Ping timeout: 246 seconds]
ulrichard has quit [Remote host closed the connection]
jokoon has quit [Read error: Connection reset by peer]
maxlath has joined #ipfs
erictapen has joined #ipfs
droman_ has joined #ipfs
anewuser has joined #ipfs
droman has quit [Ping timeout: 240 seconds]
chris6131 has joined #ipfs
chris613 has quit [Ping timeout: 248 seconds]
shizy has quit [Quit: WeeChat 1.9.1]
pat36 has quit [Read error: Connection reset by peer]
pat36 has joined #ipfs
<vkoskiv>
Is it normal for IPFS to connect to 2000+ nodes and clog up my whole network?
Jesin has joined #ipfs
erictapen has quit [Ping timeout: 240 seconds]
<gde33>
I wrote my own aggregator, eventually I decided that performance should be monitored constantly. If tasks (cpu) take longer than normal there are probably other processes that need the cycles (tune it down until resolved, if not resolved, wait for 5 sec and test again)
<r0kk3rz>
yes
<r0kk3rz>
vkoskiv: yeah its not ideal, but it will eventually connect to everyone
Anchakor has quit [Ping timeout: 240 seconds]
<gde33>
for bandwidth issues I ping local host and compare the ping with normal local host ping, If it grows by 50% since the last test the max requests is tuned down, if that doens't help pause for 5 seconds and ping local host again
rodolf0 has joined #ipfs
<vkoskiv>
r0kk3rz: There's no peer connection limit in go-ipfs at this time, right?
<r0kk3rz>
i dont think so
<vkoskiv>
r0kk3rz: Any quick solution? I can't really have it running on my main computer at the moment
<gde33>
if cpu or bandwidth doens't return to normal for a few min it stops for a few min and suggests a restart.
<vkoskiv>
My whole network just slows to a crawl
erictapen has joined #ipfs
erictapen has quit [Remote host closed the connection]
erictapen has joined #ipfs
<vkoskiv>
I've restarted it a few times, and every time it goes up to 2000+ connections, and my network goes down
<r0kk3rz>
not sure tbh i know whyrusleeping was working on a solution
<gde33>
I'm talking about my aggregator. How IPFS deals with it I have no idea. I think you just have to switch it off until patches arive. :/
<vkoskiv>
whyrusleeping: ping
<vkoskiv>
gde33: Yeah, still alpha software so I fully understand
<r0kk3rz>
its only recently that there was 1000+, now its more like 4000+
<vkoskiv>
It's an insanely cool technology, though. I don't undestand how I just became aware of it a day ago
<vkoskiv>
I've been playing around and showing to my friends
<r0kk3rz>
vkoskiv: if you arent actively serving something, id turn it off
<vkoskiv>
r0kk3rz: Yeah.
<gde33>
lol, even if he is serving something he is not serving something
<vkoskiv>
"serving" something
<vkoskiv>
More like distributing
<r0kk3rz>
whatever word you prefer :)
erictapen has quit [Ping timeout: 248 seconds]
<gde33>
vkoskiv: maybe you can lower the priority? I've never really tried to do that with anything.
<vkoskiv>
But shouldn't it be fairly trivial to add a config param + limit for peer connections? I haven't checked the source yet, and I haven't coded Go yet.
<vkoskiv>
I'd imagine there's some good reason for not having it if it is indeed trivial
<grawity>
I'd guess not all connections are equally useful
<gde33>
My reason for montioring performance and adjusting behavior is that you don't know what else the user is doing.
<grawity>
so if the limit is 500, and you simply stop after making the first 500 connections (whatever they may be), your performance might suck
<vkoskiv>
Yeah
<gde33>
if the system is idle you can use a lot more than when some task is sucking it dry.
erictapen has joined #ipfs
<vkoskiv>
So then drop connections and reconnect to new nodes till performance is optimal?
<vkoskiv>
Something like that
<grawity>
and that's no longer "fairly trivial", I assume
<gde33>
should you have to disable ipfs if you want to compile something? lol
<vkoskiv>
I literally just started delving into the specs + papers, so I've got no clue what I'm talking about yet
<cjd[m]>
500 seems high, I think bittorrent uses no more than 50 or so
<vkoskiv>
My laptop stays at a reasonable ~200, but my desktop with a wired connection goes crazy and connects to over 2000 peers
<gde33>
it should depend on the box, my laptop runs up to 150 requests if the wifi is any good. If it is crap it drops to 5 or 10.
<cjd[m]>
I remember way back in the old days, bittorrent benefitted (somewhat) from cranking up the max connections but those days are well behind us
<gde33>
(for my aggregator)
<cjd[m]>
I used to run 1600 and I was running a bad bittorrent node for doing it
<gde33>
the silly thing about lots of requests is that you don't know when they are answered.
<gde33>
If you do 100 per second you might have 300 ariving in the same second
<gde33>
the higher the number the weirder it gets
<cjd[m]>
depending on how you're handling responses, that can be a problem indeed
<gde33>
I sort of fixed that by not doing anything with the responses when they arive. A fixed number per second is parsed.
<cjd[m]>
yeah, that makes good sense
kaotisk has joined #ipfs
kaotisk-irc has quit [Ping timeout: 258 seconds]
<gde33>
and of course limit the requests if parsing is waaay behind
<gde33>
the fixed number is not really fixed. I mean adjusted by system performance.
<cjd[m]>
what are the requests ? Are they mostly DHT traffic ?
<gde33>
it is an rss aggregator
<cjd[m]>
oh ok lol
<gde33>
results need to be parsed and merged into the result set
<gde33>
so I also buffer them after parsing so that 10 or 20 can be merged in. To avoid having to move everything :P
<Kubuxu>
vkoskiv: as lgierth said: yeah we're testing patches for both the cpu usage and the number of connections -- the network has grown by 600% over the past 2 weeks and that's exposed a few bottlenecks
<vkoskiv>
Kubuxu: Good to hear!
<vkoskiv>
Kubuxu: Any explanation for the sudden increase in network size?
<vkoskiv>
I saw the youtube video by Siraj and got really hyped
<gde33>
people with deep pockets foaming from the corner of the mouth over anything blockchain
<gde33>
I really hate how all applications just suck the cpu and bandwidth dry whenever they feel like it
<r0kk3rz>
ipfs isnt anything blockchain though
<gde33>
the deep pocket guy doesn't know what a block chain is
pat36 has quit [Read error: Connection reset by peer]
<gde33>
"Decemtralized" probably sounds just as sexy
<r0kk3rz>
indeed it does mention IPFS
pat36 has joined #ipfs
<cjd[m]>
I think the tragedy of blockchain technologies is that you can never be sure which ones are designed really to make a difference in the world and which ones are get-rich-quick schemes.
<r0kk3rz>
yes certainly
<gde33>
The tragedy of the blockchain
<vkoskiv>
Kubuxu: Is that 600% increase also why the ipfs.io proxy is so slow?
<Kubuxu>
partially yes
NullConstant has joined #ipfs
<Kubuxu>
we had those fixes in works some time before but this 600% node increase surprised us a bit
<gde33>
cjd[m]: Things have to gather their audience first. If they are builders it will make a difference if they are ponzi schemers it is a get rich quick scheme
<gde33>
Early on I suggested creating a shopping software for bitcoin. People didn't even reply.
chris613 has joined #ipfs
<gde33>
If they wanted to create a new kind of money it should have been anchored onto products asap.
pat36 has quit [Read error: Connection reset by peer]
chris6131 has quit [Ping timeout: 240 seconds]
pat36 has joined #ipfs
<cjd[m]>
I feel like bitcoin had sort of a pass because when it was written, nobody had any idea of what a cryptocurency could mean, so whoever satoshi was they were very special and they really apparently cared a lot about something
<gde33>
I'm not sure, people were doing pyramid schemes, hyip's and other kinds of get rich quick schemes using endless different formulas all along
<gde33>
the only difference was selling it to a bigger audience
<cjd[m]>
yup, no doubt
<gde33>
it was a bit like stumbleupon escaped from the endless autosurf websites
chris6131 has joined #ipfs
Boomerang has quit [Ping timeout: 260 seconds]
Alpha64 has joined #ipfs
chris613 has quit [Ping timeout: 240 seconds]
<gde33>
haha the big ponzi websites got shut down.
ashark_ has joined #ipfs
<Alpha64>
?
<cjd[m]>
somehow I've decided to accept that whenever I buy into something, it's defacto a pyrimid and hopefully I just won't end up at the bottom... idk, even like euros and dollars are not really that ethically clean
<gde33>
Money Maker Group and Talk Gold
<cjd[m]>
> Money Maker Group
<cjd[m]>
excellent name
<cjd[m]>
sounds uber legit
<Alpha64>
fiat is quite literally dirty
<Alpha64>
in all aspects
<chrono[m]>
true
<chrono[m]>
but so are most the ICOs amd scam/shitcoins
<Alpha64>
nobody ever got sick from touching an ico
<cjd[m]>
Well.. in some way fiat represents the state which is supposedly elected and all of that stuff
<chrono[m]>
ha lol
<chrono[m]>
your also belive in santa clause?
<chrono[m]>
or that your money is backed by gold?
<Alpha64>
my money is backed by a loan
<gde33>
I invented a hilarious system called TAX MONEY. Basically government mints a coin every 6 months, you use it to pay tax, it expires after 12 months.
<Alpha64>
that the government took on my name
ashark has quit [Ping timeout: 248 seconds]
<Alpha64>
where do i get into that ico
<gde33>
you can pay taxes with fiat also, then you pay 200% tax
libman has joined #ipfs
<Kythyria[m]>
chrono: Cute, you believe gold or hashrate are any less imaginary than fieat?
<cjd[m]>
> your also belive in santa clause?
<cjd[m]>
I don't like this line of thinking, yes I agree things are bad, there is both corruption and ineptitude, but there is this voting thing so we don't need to go to war in order to change it, and if you don't vote because it's corrupt then you're not really helping anyone.
<Kythyria[m]>
*fiat
<Alpha64>
i believe in the holy fork
chris613 has joined #ipfs
<Alpha64>
may the hashrate bless us
<gde33>
He isn't much of an intelectual but Russel Brand showed that if you get enough public support for something reasonable the elected officals dont dare oppose the idea.
<cjd[m]>
Popper defined democracy as a system (even if it is brutal and totalitarian) in which the people have the right to judge and replace their rulers. That idea is so prescious because even if we have a horrific state, we can at least do something about it peacefully.
<Alpha64>
it's funny because i'm designing a system with a democratic system in it and it is indeed a mess
<vkoskiv>
Also: The IPFS web UI doesn't load at all on Safari
chris6131 has quit [Ping timeout: 248 seconds]
<gde33>
Alpha64: mine should work but it is only simple from the user perspective. The back end is actually horrific.
<vkoskiv>
For the local daemon
<chrono[m]>
kythyria: even more cute believing fiat is less imaginary than PoW or "value" of physical resources
<Kythyria[m]>
How much practical application does gold have?
<Kythyria[m]>
How much practical application does brute forcing SHA1 have?
gts has quit []
<Kythyria[m]>
Oh right, very little, and even less.
<chrono[m]>
wenn, you can prove your practically by eating 1k in euro notes then
<Kythyria[m]>
The only reason we use shiny yellow rocks as money is because everyone agrees that shiny yellow rocks are expensive.
<lgierth>
can we get back to topic
<lgierth>
this isn't reddit
<Kythyria[m]>
What
* cjd[m]
whistles innocently
<Kythyria[m]>
And I said all money is imaginary.
<chrono[m]>
I totally agree
<lgierth>
hi cjd o/
maxlath1 has joined #ipfs
maxlath has quit [Ping timeout: 240 seconds]
maxlath1 is now known as maxlath
ylp has quit [Quit: Leaving.]
Anchakor has joined #ipfs
rozie has quit [Quit: leaving]
<gde33>
lgierth: maybe there needs to be an off topic channel as well. Perhaps ##ipfs ?
<gde33>
humm on second thought that channel should really be about using the stuff
<zenground0>
Hello, rumor on the street is that go-ipfs may remove GET methods from its HTTP api (https://github.com/ipfs/ipfs-cluster/issues/161) in favor of POSTS. However the api docs I have found mostly use GET requests in the examples (https://ipfs.io/docs/api/). Do GET endpoints also support POSTs right now?
<miflow[m]>
I'd like to hope there will be a coin with built in ai-optimization, so it would respect fairness and democracy
<gde33>
##ipfs-crowd-control
m0ns00n has quit [Quit: quit]
aikidouke has quit [Killed (kornbluth.freenode.net (Nickname regained by services))]
<zenground0>
Also lots of gratitude to anyone who can point me towards http api routing information in the go-ipfs repo. I've been having trouble finding something like this.
Guest80404 has joined #ipfs
leavehouse has joined #ipfs
krs93 has joined #ipfs
gts has joined #ipfs
gts has quit [Client Quit]
gmoro has quit [Ping timeout: 258 seconds]
gmoro has joined #ipfs
}ls{ has quit [Ping timeout: 248 seconds]
kaotisk has quit [Read error: Connection reset by peer]
anewuser has quit [Quit: anewuser]
}ls{ has joined #ipfs
leavehouse has quit [Ping timeout: 264 seconds]
leavehouse has joined #ipfs
jdev99 has joined #ipfs
<jdev99>
test :D
slayerjain has joined #ipfs
<jdev99>
quick question :D if i add a file to my node, can i request it from the ipfs gateway?
<Magik6k>
jdev99: you can
kirby__ has quit [Ping timeout: 255 seconds]
<jdev99>
and if you add a file to your node and you would give me the hash, i could download that file?
<slayerjain>
Hi all, when I add a file to ipfs, my node hosts it, and I can access it via ipfs gateway. however, even after shutting the node down, it's still accessible - which means someone is hosting/caching it. My question is, who? and why?
<Magik6k>
jdev99: yes
<slayerjain>
jdev99 : yes!
<jdev99>
the gateway is hosting the file i think
<Magik6k>
yep, the gateway is caching it
<slayerjain>
who pays for the gateway?
<jdev99>
pay?
<slayerjain>
I mean, are there limits about time or space?
<Magik6k>
Protocol Labs, the file usually won't be kept there for long
<slayerjain>
yes, if it's being cached, it's consuming resources.
<slayerjain>
AFAIK, Protocol labs have monetised IPFS via filecoin, I was expecting IPFS to just be a lower level network.
<slayerjain>
and since there's no currency in IPFS, i think every node is free to access whatever they want.
<slayerjain>
s/access/host
<slayerjain>
My question would rather be, "How dependable is ipfs gateway? Is it also distributed? Are there performance guarantees backed by technological promise or SLAs"?
<jdev99>
i dont get this one... filecoin is a cryptocurrenca but also a storage system, so how is ipfs monetised by filecoin
<slayerjain>
isn't filecoin built on ipfs?
<Alpha64>
there is no monetization yet
<slayerjain>
Alpha64: in ipfs or filecoin?
pat36 has quit [Read error: Connection reset by peer]
<Alpha64>
ipfs
<jdev99>
how would monetization even work? like would you pay to access a certain node?
pat36 has joined #ipfs
<Magik6k>
Gateways are just hosted IPFS instances, There is no SLA (AFAIK), they are there to simplify access to IPFS and are sponsored by PL
<slayerjain>
you create a system, where you can pay owners of nodes to host some content, on the basis of storage, bandwidth,..
<jdev99>
oh like storj
<Magik6k>
basically filecoin
<slayerjain>
or maybe filecoin
krs93 has quit [Ping timeout: 255 seconds]
<slayerjain>
yeah
<jdev99>
i see :D
<slayerjain>
Magik6k : filecoin is built on IPFS right? can we build our own gateways?
<Kubuxu>
slayerjain: any ipfs node can act as gateway
<Kubuxu>
as things go, gateway should be not be used at all
<slayerjain>
Kubuxu : any node can be used as a gateway without any modifications? (as in configuration change)
johnytiago has joined #ipfs
<jdev99>
so if some guy uploads a file on mars and i request it i would host the same file
<slayerjain>
jdev99: i'm sure this would be bounded by limits
<Kubuxu>
slayerjain: your local node is almost the same thing as gateway
maxlath has quit [Ping timeout: 248 seconds]
<slayerjain>
that's good to know :)
<Kubuxu>
ours are behind ssl terminating nginx and have few config teaks
<jdev99>
what kind of limits? i mean its called interplanetary for a reason and as far as i understood, anyone who then request the same file near me would get it from my node
<slayerjain>
Kubuxu: cool! I'll also build one (nodes designed for gateway purpose ) :D
<slayerjain>
s/nodes/IPFS system
<slayerjain>
is there a background daemon clearing up files as they populate local filesystem? I mean, Assume it's a movie, I downloaded it, I'm seeding it, but for how long after I delete it locally? (is there a hidden folder as in git, where files are cached?)
<Kubuxu>
yes, it is in `~/.ipfs` by default
<Kubuxu>
you can either run `ipfs repo gc` or setup config to run GC automatically
<jdev99>
is there currently a way to add a file to a remoteserver running a node?
<slayerjain>
Kubuxu: thanks! What would be the best way to start contributing to ipfs?
dhruvbaldawa has joined #ipfs
<slayerjain>
any fun tasks for a newbie? :)
vkoskiv has quit [Ping timeout: 260 seconds]
maxlath has joined #ipfs
Jesin has quit [Ping timeout: 240 seconds]
doowa_diddee has joined #ipfs
dhruvbaldawa has quit [Ping timeout: 258 seconds]
<jdev99>
how would ipfs work with dynamic websites like one made with nodejs or asp.net or what ever technology...? and what about databases with sensitive data? how could a technology like this replace http?
<Kubuxu>
look for "help wanted" on github
<Kubuxu>
label "help wanted"
<slayerjain>
jdev99: comsider ipfs as an alternative to NFS on steriods
<slayerjain>
well not really NFS, but something like s3
<slayerjain>
(object storage, it doesn’t do block storagw AFAIK)
<jdev99>
so it will never replace http
<jdev99>
https
<slayerjain>
it depends on what you think about http
<jdev99>
dynamic websites beeing served over http :D
<slayerjain>
as a way to access resources, it could replace http
<jdev99>
oh like when you have images in your page, those could be served over ipfs
<slayerjain>
yes.
<jdev99>
or a videplatform or stylesheets
<slayerjain>
that’s my perception
<slayerjain>
i’m not how if there is a way to write a web server replying to ipfs queries
<slayerjain>
like: http requests.
<slayerjain>
s/not/not sure
<Icefoz_>
jdev99: I'm actually thinking quite hard about something like that.
<Icefoz_>
How to make database-driven apps that use IPFS for content.
redfish has joined #ipfs
<gde33>
I had a silly idea: Make an IPFSSP! Basically an ISP only for IPFS :P
<jdev99>
how would that work?
<gde33>
that is left as an exercise for the reader
krs93 has joined #ipfs
<jdev99>
and yes it would be suilly since thats the oposite of what ipfs is meant to be
<jdev99>
*silly
<gde33>
I imagine it would work like that infamous isp (I forget where) that was catching all the popular torrents for uber fast downloads not consuming backbone bandwidth
Nehagup has joined #ipfs
<jdev99>
decentralized...
<jdev99>
would be amazing but not decentralized
<gde33>
it would just be a node?
<gde33>
could take subscriptions in file coin :P
<jdev99>
true :D
<jdev99>
but imagine someone attacks your node
<gde33>
same as attacking the isp?
<jdev99>
yes
<gde33>
have the users pay part of their subscription by hosting files
<gde33>
I vaguely remember someone mentioning mesh networking and that batman would be an easy to implement things but I forget all the details
<jdev99>
batman
<jdev99>
?
<gde33>
Better Approach To Mobile Adhoc Networking
<gde33>
a routing protocol
<Icefoz_>
It's not super silly.
<Icefoz_>
You'd basically be Amazon S3 or such except host objects on IPFS.
<Icefoz_>
So instead of someone setting up a server, buying storage, configuring IPFS, running it, and so on
<gde33>
the silly part was to not offer anything besides ipfs
<Icefoz_>
You just say "please make sure this object is available", and you get billed for it.
<Icefoz_>
...
<Icefoz_>
That's a great idea.
<Icefoz_>
I'm going to do that.
<jdev99>
oh i see, well if you are connected to alot of peers, which one do you pick if two of them have the same file?
<gde33>
the middle one
<jdev99>
haha
<gde33>
with torrents I just start both and see who wins
<jdev99>
lol but than you use double the bandwidth
<slayerjain>
btw how does ipfs assign a hash so quickly to a file? Does it check a registry? or it just hopes there's no collision of keys?
<jdev99>
hashing is a function which creates unique values
<Icefoz_>
Yeah what *does* happen if there's a hash collision?
<jdev99>
not possible
<Icefoz_>
It's always possible, just hard to find.
<gde33>
jdev99: one time I even start an extra client to do a "receck" because both were writing to the same folder :P
<slayerjain>
number of possible outcomes are limited, inputs are not.
<slayerjain>
checking might be too costly
<Icefoz_>
If your hash is 128 bits and your state is 129 bits, then there must be two states that share the same hash.
<Icefoz_>
It's just hopefully exceedingly unlikely.
<Icefoz_>
but if you assume it never happens you get things like git's MD5 collission, where even trying to check a colliding file into a repo ruins the repo.
<slayerjain>
and since IPFS assigns a unique hash to every new file version.
<jdev99>
i must be very unlikely then
<jdev99>
but md5 is bad :D use sha512
<jdev99>
much more unlikely to get a collision
<slayerjain>
I won't be surprised if there's a alternative planned alternative to distinguish files.
<Icefoz_>
md5 was great in 2003 when people were still using... I dunno, DSA or whatever.
<slayerjain>
it's unlikely you'll get win a lottery, but it's possible.
<slayerjain>
s/get win/win
<jdev99>
sure
<Icefoz_>
sha128 was great until a couple years ago when people started making ASIC's for running specifically that
jdev99 has quit [Quit: Page closed]
<Icefoz_>
So you can say "as far as we can tell, if you hashed every piece of data on Earth for the next 100 years, the chances of getting a collision would be 0.0000000001%, we will assume it never happens"
<Icefoz_>
But be aware that's an assumption you are making that has historically not held true.
<gde33>
indeed every bit of data on earth might not be much soonly
<Icefoz_>
"how do we upgrade from a broken hash function" is a *great* question to ask. :D
<Icefoz_>
Thank you, slayerjain
<slayerjain>
I guess the solution is just to switch to SHA512 (from current 256) in case of hash detection.
<slayerjain>
and wait for another collision :)
<Icefoz_>
That's perfectly reasonable if you expect it and make it easy!
<Icefoz_>
I mean, MD5 was fine for 10 years or so
<Icefoz_>
SHA1 was fine for 10 years or so I guess?
<Icefoz_>
If SHA256 is fine for 10 years then great.
<Icefoz_>
The problem is that people are still using MD5 ;-)
<slayerjain>
The problem isn't finding another object with the same hash
<slayerjain>
(which is in case on if hash is used for encryption)
<slayerjain>
s/case on/case
<slayerjain>
but since hash is an identity of a file here, nothing is stopping from find 2 files with the same hash, and hence the same identity on ipfs.
<slayerjain>
(you didn't try to find such a file, you just encountered it)
shizy has joined #ipfs
<slayerjain>
so this could be a threat to such a system's resilience at scale.
ygrek_ has quit [Ping timeout: 240 seconds]
lacour has joined #ipfs
pat36 has quit [Read error: Connection reset by peer]
pat36 has joined #ipfs
Jesin has joined #ipfs
mib-kd743naq has joined #ipfs
<mib-kd743naq>
don't get carried away though
<mib-kd743naq>
yes, collision attacks on md5 are trivial, and are moderately easy on sha1
<mib-kd743naq>
*however* preimage attacks are unheard of, on any alg, including md5
<mib-kd743naq>
and in the case of IPFS ( and many uses of git for that matter ), where you need to serve something *sensible* and then replace it with something *malicious* - it's not a practical threat
<mib-kd743naq>
( let alone the sha2-256 that is used by default, this one doesn't even have the beginnings of viable collision attacks )
maxlath has quit [Ping timeout: 240 seconds]
tailorwise has joined #ipfs
igorline has quit [Ping timeout: 240 seconds]
<jfmherokiller[m]>
is it possible to completely rehash an entire repo with a new hash function?
jdev has joined #ipfs
<jdev>
to make hashcollision almost impossible, ipfs could add a salt to the hash
<Magik6k>
jfmherokiller[m]: you could get blocks using `ipfs pin ls` and do `bolck get | block put --mhtype=[hash type]` on each
pat36 has quit [Read error: Connection reset by peer]
<jfmherokiller[m]>
ok thanks
pat36 has joined #ipfs
<Magik6k>
jdev: no, it woundn't make much of a difference, at least for some attacks (like recent shattered)
<jdev>
not talking about attacks but talking about hashcollision in general so that if i request a file, i get the one i wanted
r0llerd3rby has joined #ipfs
<Magik6k>
If you mean a 'collision by accident' that's probably not possible
pat36 has quit [Read error: Connection reset by peer]
<jdev>
it is but very unlikely though, more files = higher chance of hashcollision by accident
pat36 has joined #ipfs
kirby__ has joined #ipfs
dhruvbaldawa has joined #ipfs
<mib-kd743naq>
jdev: if you truly believe that random hits like this are possible, then it would be a good use of time to just generate random numbers, treat them as bitcoin private keys, deriving their public addresses, and checking if the result has any money in it, and grabbing it if the case
Jesin has quit [Ping timeout: 264 seconds]
pat36 has quit [Read error: Connection reset by peer]
<mib-kd743naq>
( the probabilities involved are the same btw, but your "potential" payout is way better than looking for an IPFS collision )
<slayerjain>
mib-kd743naq: You again got me wrong. It’s not about attacking the system. No one is being attacked. I’m just talking about a possibility where 2 files on the network happen to have the same hash bu chance.
pat36 has joined #ipfs
<slayerjain>
s/bu/by
rendar has quit [Quit: std::lower_bound + std::less_equal *works* with a vector without duplicates!]
krs93 has quit [Ping timeout: 258 seconds]
<mib-kd743naq>
slayerjain: what I am trying to point out is that in this context a coordinated attack is indistinguishable from "possibility of 2 files on the network happening to hash to the same value by chance"
<mib-kd743naq>
and I provide a decent test for it ( with a potential return ;)
kirby__ has quit [Ping timeout: 248 seconds]
kirby__ has joined #ipfs
pat36 has quit [Read error: Connection reset by peer]
pat36 has joined #ipfs
kirby__ has quit [Ping timeout: 255 seconds]
Jesin has joined #ipfs
maxlath has joined #ipfs
<slayerjain>
mib-kd743naq you mean:
<slayerjain>
Person 2: I’m just a good ol’ dev/ops enjoying my IPFS powered systems. *Alert: there’s been a file hash collision* (no body tried to to do it, but it happened), and its time to pray that ipfs folks have a quick fix, before I break-SLAs/lose customers.
<slayerjain>
Person 1 (attacker): I need to find a file with SHA256 hash ‘x’, but damn, it may take 9*10^51 years in the worst case, and so I’m giving up.
<slayerjain>
Are the situations?
<slayerjain>
s/the/the same
<mib-kd743naq>
that's the tricky bit about *cryptographic* hash functions - the two scenarious are effectively the same, yes.
<slayerjain>
the likeliness - yes. but it may happen the next day.
<slayerjain>
any system at scale, thus must provide a walk-around.
<mib-kd743naq>
due to human inability to reason about the "space" within which this collision has to occur, it seems utterly counterintuitive that someone can categorically and with certainty dismiss a what is supposed to be a random event
<mib-kd743naq>
this is why instead of convincing you to just "trust the math", I am pointing out a practical counter-example: IFF the possibility of such a collision was practical
<mib-kd743naq>
there would be howtos plastered all over the web how to try ones luck at finding the exact key that unlocks Satoshis bitcoins
<slayerjain>
How likely is the failure of AWS's AZ?
<mib-kd743naq>
the return of such exercise would be immense, yet nobody is attempting it
<mib-kd743naq>
why is that?
<slayerjain>
forget the recent S3 outage (which was a surprising event).
<mib-kd743naq>
I don't follow your question...
<slayerjain>
It's never been down in years, AWS (and every other cloud vendor) recommends a multi-AZ deployments to make sure your system is highly available. However, when S3 failed in a single AZ (data-centre), it took down a surprisingly significant part of S3. They just anticipated S3 won't fail, and thus didn't architect multi-AZ (data centre) deployments.
<slayerjain>
s/part of s3/ part of the Web
<slayerjain>
The real answer to my question depends on the impact of such a disaster. In my use case (the system I'm gonna use IPFS for), this might be disastrous, and hard to debug.
<mib-kd743naq>
well... let's cut this short and answer your direct question
<mib-kd743naq>
to the best of my knowledge there are zero ( none whatsoever ) provisions / codepaths / etc within all IPFS implementations that protect against a potential hash function collision
<mib-kd743naq>
if it happens - you get what you get
<slayerjain>
my question is, "Does IPFS currently anticipate SHA256 collisions? If so, how?"
<slayerjain>
s/how/how does it handle it
pat36 has quit [Read error: Connection reset by peer]
pat36 has joined #ipfs
<mib-kd743naq>
adding to that: I personally would be extremely puzzled if anyone attempts to add code that deals with this eventuality in any manner
<leavehouse>
worrying about accidental sha-256 collision seems to be even less sensible than, for example, worrying about a potential future near-Earth supernova
<mib-kd743naq>
slayerjain: does the above answer your original question?
<Icefoz_>
You can do things about sha256 collisions if you try. Near-earth supernovae, less so!
<Icefoz_>
I consider the git MD5 thing an excellent example of why you need to engineer to account for hash collisions.
<mib-kd743naq>
Icefoz_: md5 has no known history of accidental hash collisions ( and that space is 128 bits )
<mib-kd743naq>
the reason md5 was phased out is because of its succeptibility to attacks, which is not what slayerjain is asking
<Icefoz_>
An MD5 vulnerability was discovered, a test engineer who knew exactly what they were doing made a unit test to generate an MD5 collision so they knew when they handled collisions gracefully, checked in, and broke git's repository.
<mib-kd743naq>
Icefoz_: IPFS does have a robust mechanism to handle collision attacks - there is a clear, backwards and forward compatible path to switch to a different hashing alogithm
<Icefoz_>
There's other valid ways of getting a collision than "by accident" or "from an attack".