<cryptix>
(the one from yesterday was resolved by increasing ulimit)
brixen has joined #ipfs
M-jgrowl has joined #ipfs
M-oddvar has joined #ipfs
ogzy has joined #ipfs
<ogzy>
hi all, i am impressed with the IPfs project, i am working on DDoS mitigation and wondering whether any project is under development based on IPFS?
<zignig>
I think so , he's really busy most of the time.
<zignig>
just ask here or on github.
<whyrusleeping>
cryptix: are you running latest master?
<whyrusleeping>
goprocess panics are my most favoritest
<whyrusleeping>
ogzy: whatsup?
<ogzy>
i am working on DDoS mitigation and the IPFS is the exactly what i was thinking, the papers are not published and i am wondering whether any work on DDoS mitigation is going on related with IPFS for DDoS mitigation
<ogzy>
whyrusleeping, ^
stoopkid has joined #ipfs
krl has joined #ipfs
<cryptix>
whyrusleeping: yup - upgraded my debian vps to latest than startet seeing them when refs'ing some data
chriscool has quit [Ping timeout: 264 seconds]
<whyrusleeping>
cryptix: cool, can you file an issue for that?
<whyrusleeping>
i'm going to work on fixing it right now, but having a tracking issue that i didnt create will help
<cryptix>
saw it twice in a row now it wont...
<whyrusleeping>
ogzy: well, what do you mean by DDoS mitigation? thats a rather vague thing
<whyrusleeping>
the ipfs paper is on ipfs.io
<cryptix>
need to run now - will file the panics later
<cryptix>
also, transfers get stuck and i need to restart nodes.. and lots of 'not conntacted. dialing' in the logs.. self dailing is still a thing, i guess?
<ogzy>
whyrusleeping, instead of using mitigation techniques at a single point, distributing the content and making the browsers act like a content provider is automatically mitigating the ddos, even the server is down, the content will be on Internet, thats the point i was thinking
<whyrusleeping>
cryptix: yeah.... self dialing is still a thing :/
<whyrusleeping>
ogzy: yeah, that just happens as a result of the distribution mechanism ipfs uses
<ogzy>
whyrusleeping, so i wonder whether any enhancements or missing issue of IPFS to improve the DDoS attack mitigation part
shyamsk has quit [Ping timeout: 264 seconds]
shyamsk has joined #ipfs
<whyrusleeping>
the distribution mechanism can be made to be smarter
shyamsk has quit [Ping timeout: 255 seconds]
elsehow has quit [Ping timeout: 264 seconds]
Guest73396 has joined #ipfs
shyamsk has joined #ipfs
<zignig>
speaking of which , whyrusleeping does an IPNS publish send out to more nodes ( bigger DHT radius ) than normal adds ?
<zignig>
if it doesn't , it should ;)
<whyrusleeping>
zignig: lol, why?
<whyrusleeping>
if it properly puts to kademlia, then pushing to more wont really help all that much
<ion>
Push to all the nodes
<whyrusleeping>
lol
<whyrusleeping>
just precache it on everyones nodes
<whyrusleeping>
if you notice ipns resolves taking longer recently its because the behaviour was changed from 'grab the first value for /ipns/X you see and assume its right'
<whyrusleeping>
to
<zignig>
whyrusleeping: it will help in resilience , not in search.
<whyrusleeping>
'grab 20 values for /ipns/X and sort through them'
<whyrusleeping>
zignig: well, its only 'resilient' if you can find it
<zignig>
indeed... ;)
<whyrusleeping>
if a thousand nodes have it, but nobody ever knows the ask them for it, its as good as gone
wowaname has joined #ipfs
<zignig>
how is it searched for just the hash of /ipns/<nodeid>
<whyrusleeping>
so the key in the dht is '/ipns/<nodeid>'
<whyrusleeping>
so the dht uses h(key) to find closest peers
wowaname has quit [Read error: Connection reset by peer]
<zignig>
and the fact that it is signed prevents others from faking it.
<whyrusleeping>
yeap
<whyrusleeping>
and nodes will validate a record upon receiving it
<whyrusleeping>
so i cant send a fake record to overwrite your correct one
elsehow has joined #ipfs
wowaname has joined #ipfs
<zignig>
and loacal checking will prevent a sybil attach.
<zignig>
s/attach/attack
<multivac>
zignig meant to say: and loacal checking will prevent a sybil attack.
<zignig>
multivac: can you punch your self in the powersupply please.
amstocker has quit [Ping timeout: 255 seconds]
elsehow has quit [Ping timeout: 240 seconds]
<whyrusleeping>
right?
ReactorScram has joined #ipfs
* zignig
needs to read the kademlia paper again.
elsehow has joined #ipfs
shyamsk has quit [Ping timeout: 255 seconds]
shyamsk has joined #ipfs
elsehow has quit [Ping timeout: 264 seconds]
elsehow has joined #ipfs
wowaname has quit [Read error: Connection reset by peer]
wowaname has joined #ipfs
elsehow has quit [Ping timeout: 260 seconds]
ReactorScram has quit [Ping timeout: 265 seconds]
<whyrusleeping>
when cryptix says the codes broken, i stay up and fix that shit
wowaname has quit [Read error: Connection reset by peer]
* zignig
passes whyrusleeping a double redbull chai mocha with a twist of lemon.
<ipfsbot>
[go-ipfs] whyrusleeping created dial-smarter (+1 new commit): http://git.io/vCE85
<ipfsbot>
go-ipfs/dial-smarter 9538115 Jeromy: refactor dialing to not panic, and to be smart about ordering...
<whyrusleeping>
zignig: ethiopian yergacheffe, brewed strong in a pourover
<whyrusleeping>
:D
<zignig>
mmmm
<zignig>
coffee
<whyrusleeping>
i dont do any of that weird energy drink stuff
<ipfsbot>
[go-ipfs] whyrusleeping opened pull request #1837: refactor dialing to not panic, and to be smart about ordering (master...dial-smarter) http://git.io/vCE4x
<whyrusleeping>
it doesnt really make you feel all that good
<whyrusleeping>
but, anyways, cryptix's bug should be fixed in that PR. sleep time.
<zignig>
i do from time to time, but not like some of my workmates.
<whyrusleeping>
ooooh, yeah. my last roommate would have three a day
<whyrusleeping>
i cant imagine
<zignig>
;/
<zignig>
gnite.
<whyrusleeping>
gnite!
* zignig
has astralboot nearly working for real , back to programming.
groxx has quit [Quit: Connection closed for inactivity]
dignifiedquire has quit [Quit: dignifiedquire]
ygrek has quit [Ping timeout: 265 seconds]
ReactorScram has joined #ipfs
pguth2 has quit [Remote host closed the connection]
queicherius has quit [Remote host closed the connection]
<_p4bl0>
hello
<_p4bl0>
I just updated to the new i386 build and it seems name resolve and name publish does not work
<_p4bl0>
is that a known bug or not?
<_p4bl0>
hm
<_p4bl0>
it worked now
<_p4bl0>
I guess it may have been a network error the first two times
Wolf480pl has quit [Ping timeout: 244 seconds]
ogzy has quit [*.net *.split]
ogzy has joined #ipfs
<stick`>
i386? is that still a thing? :-)
<ion>
m68k for life
pguth2 has joined #ipfs
NeoTeo has joined #ipfs
r1k0 has quit [Quit: Leaving]
<ipfsbot>
[go-ipfs] rht force-pushed ipfs-path-resolve from 172eeb8 to 9021f97: http://git.io/vcmFY
<ipfsbot>
go-ipfs/ipfs-path-resolve eddd641 Matt Bell: Resolve '/ipfs/root/some/path' paths in 'ipfs resolve' command...
<ipfsbot>
go-ipfs/ipfs-path-resolve 8d370a3 Matt Bell: Added help text to 'resolve' command to show DAG path resolution....
<ipfsbot>
go-ipfs/ipfs-path-resolve 3a2b83d Matt Bell: resolve cmd: Fully resolve IPFS paths after resolving IPNS names...
hjoest3 has joined #ipfs
hjoest2 has quit [Read error: Connection reset by peer]
Guest73396 has quit [Ping timeout: 256 seconds]
stoopkid has quit [Ping timeout: 268 seconds]
bedeho has quit [Ping timeout: 246 seconds]
<ipfsbot>
[go-ipfs] rht force-pushed ipfs-path-resolve from 9021f97 to 740447e: http://git.io/vcmFY
<ipfsbot>
go-ipfs/ipfs-path-resolve 04482cb Matt Bell: Consolidated 'resolve' tests...
<ipfsbot>
go-ipfs/ipfs-path-resolve 2e0c072 rht: Use core path resolver for 'resolve command' [1]...
<dignifiedquire>
can you gist me the whole output of the install?
<slothbag>
im on nodejs 4.2.1, npm 2.14.4
<dignifiedquire>
right then upgrade npm, just run npm update -g npm
<slothbag>
ok thanks
<dignifiedquire>
latest npm version is currently 3.3.6
<dignifiedquire>
which fixes the annoying behaviour that you are seeing that installs fail because of invalid peerdependencies
<dignifiedquire>
if you find any other issues please feel free to report them
<achin>
slothbag: "ipfs object stat <hash>"
<achin>
this will technically download <hash>, but it won't download anything that <hash> references
<ion>
Hopefully /ipfs/<hash> isn’t a terabyte in size. :-P
<slothbag>
@achin thanks, yeah i'm worried about what @ion says
<ion>
go-ipfs might have safeguards against unrealistically large objects, i don’t know.
<achin>
in theory it could be a very large block, but if it was created with "ipfs" then it won't be (since ipfs doesn't create really large things)
<achin>
(by "ipfs" i mean the "ipfs" command from go-ipfs)
whalebacker has quit [Remote host closed the connection]
Myron has joined #ipfs
judson_ has joined #ipfs
<judson_>
A quick question. I added this to ipfs yesterday: QmRwYme1JKKiC2WCa9SpyX1kHJJgC9sUZrjP6xZF5qLVJc/ . I turned off my local ipfs daemon. I'm sure nobody pinned it. So, how is it still available? Is it awaiting garbage collection? How long will it stay up if, for example, I never turn on my local daemon again? Just a bit confused. Thanks.
<richardlitt>
Sprint for this week updated: https://github.com/ipfs/pm/issues/39. Thanks all. jbenet mappum noffle: add your to dos to the sprint, if you have any this week. No worries if not. :)
<dignifiedquire>
cc daviddias krl jbenet whyrusleeping
slothbag has quit [Remote host closed the connection]
equiatomic has joined #ipfs
lporras has joined #ipfs
JasonWoof has quit [Ping timeout: 246 seconds]
reit has quit [Ping timeout: 240 seconds]
ogzy has joined #ipfs
<krl>
dignifiedquire: this is part of the gateway
<krl>
(go-ipfs at the moment)
<dignifiedquire>
krl: I see, but we could serve any html/js/css that we want there correct? (thinking about improving those views and things like image/markdown renderers)
<krl>
dignifiedquire: yes, and that sounds like a great idea
martinkl_ has joined #ipfs
<dignifiedquire>
krl: can you point me in a direction where that is generated atm? (go-ipfs is a big code base for someone new ;) )
<krl>
dignifiedquire: hmm, i have not touched the go code much myself either
<dignifiedquire>
krl: fair enough, will just grep for bootstrap keywords
<josephdelong>
If I want to make a webpage that allows the user to interface with their local IPFS installation; is that only possible using Go at the moment?
rehelmin has joined #ipfs
Animazing` has joined #ipfs
hjoest has joined #ipfs
<gamemanj>
josephdelong: I'm pretty sure that already exists...
<gamemanj>
josephdelong: it's the webui
<josephdelong>
That is true but I was looking to roll my own
<josephdelong>
Am I crazy?
<whyrusleeping>
ion: looks like files i accidentaly added... meh
<whyrusleeping>
should probably remove them
sasha has quit [Ping timeout: 268 seconds]
<josephdelong>
I’m willing to learn Go I just dont want to be barking up the wrong tree
<whyrusleeping>
oh, the .gxlastpubver is the version of that subpackage that was last published by gx
<whyrusleeping>
which is a packaging tool i have been trying to build
<ion>
Ah, found it.
Guest73396 has quit [Ping timeout: 268 seconds]
demize has quit [Quit: Do Androids Dream of Electric Sheep?]
zz_r04r is now known as r04r
domanic has joined #ipfs
therealplato has joined #ipfs
therealplato1 has quit [Ping timeout: 255 seconds]
rehelmin has quit [Quit: Leaving.]
demize has joined #ipfs
domanic has quit [Ping timeout: 244 seconds]
martinkl_ has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
elsehow has joined #ipfs
demize has quit [Client Quit]
fazo has quit [Quit: WeeChat 1.3]
SDr has quit [Ping timeout: 252 seconds]
rehelmin has joined #ipfs
demize has joined #ipfs
SDr has joined #ipfs
SDr has quit [Changing host]
SDr has joined #ipfs
<bengl>
daviddias: congrats on getting registry-static working w/ ipfs
<daviddias>
bengl: thank you :D
<bengl>
and yeah, you're gonna need a TB
<bengl>
npm is *big*
<daviddias>
and there is more! what we want to do is be able to mount read only 'mfs' namespaces from other nodes
<daviddias>
so that people can do npm install modulename and they use IPFS to look them in the network without even knowing the hash or previously downloaded the registry
<daviddias>
ahaha true
<daviddias>
when I so your warning of ~300Gb, I worried about the health of my computer
<bengl>
yeah it's closer to half a TB now
<bengl>
i stopped keeping a mirror on my laptop
<ReactorScram>
Every node.js package is ~ a TB?
<ReactorScram>
s/every/all/
rehelmin has quit [Quit: Leaving.]
<bengl>
ReactorScram: no, all of them together are. and if you're running registry-static, you're downloading all of them.
<ReactorScram>
Yeah I couldn't phrase it right
<ReactorScram>
Does that include the whole Git history of them all too?
<bengl>
no
<ReactorScram>
wow
<bengl>
but it does include every version published to npm
simonv3 has joined #ipfs
<bengl>
also it's more like ~0.5TB right now, i think (it's been a while since i checked)
<bengl>
looks like my mirror is sitting at about 450GB right now
<ReactorScram>
I've never tried fully mirroring something like that, but I did configure my Arch desktop to be a local mirror when installing Arch on my laptop and it was so much fun
<sonatagreen>
can you give an example of a use case for what you would do with a pub/sub system if you had it?
<sonatagreen>
and i can say how one would use bitmessage to do that thing
<ReactorScram>
As opposed to regular p2p? Because the world has been friggin ACHING for a "send file from computer A to computer B" for probably 30 years
rehelmin has joined #ipfs
<ReactorScram>
sonatagreen: I guess have something that's updated on IPNS, like a blog, and use pub / sub to push it to other nodes instead of RSS polling it?
<sonatagreen>
right, that's something BM has solved
<whyrusleeping>
yeah, thats on the todo list
<sonatagreen>
you can subscribe to a broadcast address to get notified when it posts something
<whyrusleeping>
implementing a pubsub is easy, preventing abuse is harder
<Xe>
^^^^^^
<sonatagreen>
BM is all about preventing abuse through proof of work
<whyrusleeping>
yeah... PoW is not my cup of tea...
<sonatagreen>
there are only so many possible solutions to the abuse prevention problem.
<sonatagreen>
do you actually have a preferred approach? or are you hoping that a miracle will occur?
rehelmin has quit [Client Quit]
<ReactorScram>
this part worries me: The difficulty of the
<ReactorScram>
proof‐of‐work should be proportional to the size of the message and should be set such that an average
<ReactorScram>
computer must expend an average of four minutes of work in order to send a typical message
<whyrusleeping>
my current ideas are somewhat near trusted filtering combined with proof of work, the same way that vermouth is combined in a dry martini
<whyrusleeping>
(open the bottle, then put it back on the shelf)
<ReactorScram>
I don't like buying new computers so that means in a few years it would take my computer half an hour to send each message
<whyrusleeping>
nodes should be able to come to a topical conclusion about how well they trust a given peer, based on how cooperative they are in the network
<whyrusleeping>
(useful messages sent received, requests fulfilled, etc)
<sonatagreen>
so a trust/pow hybrid?
<whyrusleeping>
yeah
<whyrusleeping>
where the work would be an expenditure of bandwidth
<whyrusleeping>
nodes can join a pub/sub group for a given key, as long as they are cooperative and aid in the publishing
<whyrusleeping>
and depending on node trust levels, you can orient the distribution graph to put more trusted nodes at more advantageous positions
<sonatagreen>
How does a newly-introduced node get started?
<sonatagreen>
i.e., how do you deal with a node that has no history behind it
<whyrusleeping>
by being placed at the bottom of the distribution graph (worse latency to receive messages)
<whyrusleeping>
and then further new nodes may be placed beneath that node
<sonatagreen>
i'm not sure i grok...
<sonatagreen>
how would a discussion forum work on this?
<whyrusleeping>
alright, so lets assume a simplified distributed pub/sub system with a single publisher, multiple subscribers
<sonatagreen>
no see, that simplification is the sticking point
<whyrusleeping>
hold on
<sonatagreen>
if you have a single publisher that abstracts away the reason why it's hard
<whyrusleeping>
the simplification is for me to explain the graph to you
<sonatagreen>
I think I got that part
<whyrusleeping>
okay
<sonatagreen>
it's basically a phone tree, right?
rehelmin has joined #ipfs
<whyrusleeping>
for the most part, yeah. with some redundancy and ways of flagging non-participatory nodes
<whyrusleeping>
for a given pub/sub group, a base level of trust could be set before allowing that node to publish
<whyrusleeping>
so they can join, receive messages, and then help out by routing messages until they gain a certain level of trust
<sonatagreen>
so you earn the right to post new messages by helping to route other people's messages?
martinkl_ has joined #ipfs
<whyrusleeping>
yeah, thats the gist of the idea i had
therealplato1 has joined #ipfs
r1k0 has joined #ipfs
therealplato has quit [Ping timeout: 240 seconds]
<whyrusleeping>
its still somewhat centralized, as the creator of the group gets to begin the trust chain
<whyrusleeping>
but every peer has the same opportunity to create a group
<sonatagreen>
how do you discover new groups?
<whyrusleeping>
by their key, on the dht
<sonatagreen>
i mean
<sonatagreen>
if someone creates a group
<sonatagreen>
how do you find out about it
<sonatagreen>
do they have to post a notification to some existing group, or out of band?
<whyrusleeping>
same way you find about anything else
<whyrusleeping>
you could do OOB or posting in an existing group
<sonatagreen>
mm, fair enough
voxelot has quit [Ping timeout: 255 seconds]
<whyrusleeping>
the difficult part is ensuring that trust cant be easily gamed
<sonatagreen>
is it possible to merge groups?
<whyrusleeping>
i would think not, because groups would be based around a given key
<whyrusleeping>
although, i can imagine ways of doing so
<whyrusleeping>
(havent actually thought about it)
<sonatagreen>
right, gaming trust is the problem
<sonatagreen>
due to sibyl attacks
<whyrusleeping>
yeah, so you can do a couple things towards avoiding that
amade has quit [Quit: leaving]
<whyrusleeping>
you can ensure that the amount of damage you could cause is less than the benefit you provide by gaining a certain level of trust
<sonatagreen>
ok, so, scenario
<whyrusleeping>
you can also limit joins to a group
<whyrusleeping>
N/hr or something
<sonatagreen>
node A joins. over time, B-K join also, and some of them get filed below A.
<sonatagreen>
A dutifully routes messages and gains trust.
<whyrusleeping>
They would likely be filed under a second node as well, the redundancy i mentioned earlier
<sonatagreen>
but, surprised, B-K are actually run by the same person as A, and only exist for the purpose of trust farming.
<sonatagreen>
*surprise
<whyrusleeping>
so, the way we would resist that is by not letting A make decisions on the graph structure until it reaches a certain trust level
<sonatagreen>
why should A get more trust for passing messages to its sockpuppets than for simply existing?
<sonatagreen>
it seems pointless and wasteful
<whyrusleeping>
yeah, so is proof of work.
<sonatagreen>
and if you're going to reward pointless waste, might as well formalize it
<sonatagreen>
proof of work is at least optimized for being hard to fake and not wasting /other/ people's resources
<whyrusleeping>
in your scenario it sounds like only A's resources are being wasted
<sonatagreen>
well, in order to know how much trust to accord A, the entire network has to keep track of the existence of all the sockpuppet nodes and their reports of 'yep, i got the message', right?
<sonatagreen>
that's way more overhead for verification than just verifying a pow
<whyrusleeping>
trust is really just a number
<josephdelong>
Does anyone know if the “ipfs add” function is working on the Web API yet? I keep getting “"This command can't be called directly. Try one of its subcommands.” with a code 1
<whyrusleeping>
and the only reports that need to be accounted for are the 'i didnt get a message from X' messages
<whyrusleeping>
josephdelong: what endpoint are you hitting?
<sonatagreen>
in fact, you could spam the system just by adding nodes, since subscribing necessarily has to (1) not require trust, and (2) notify existing users
<josephdelong>
thank you so much
<josephdelong>
@mappum that totally worked
<mappum>
:)
<whyrusleeping>
sonatagreen: right, and earlier i said that one protective mechanism was to limit joins to a group
<sonatagreen>
ok, but then your spammer is filling the queue, preventing legitimate users from joining, at modest cost to themselves (because they only have to generate a limited number of requests in order to max you out)
<whyrusleeping>
okay, so look at another trust metric to rank new join order. walk a trust graph or something
<whyrusleeping>
for every system, you can pretty easily find a 'what if this...
<whyrusleeping>
scenario where youll have to do something a little different
<sonatagreen>
'trust graph', like a web of trust?
<sonatagreen>
and how does one join the trust graph?
<ReactorScram>
sonatagreen: So bitmessage is meant to be like email?
<sonatagreen>
your plan is 'to figure out who to trust, first you have to know who to trust'
<sonatagreen>
that's not a solution
<whyrusleeping>
yeap, sounds like a good problem
<sonatagreen>
ReactorScram, you can do email but it's also discussion forums
<ReactorScram>
sonatagreen: POW makes sense for a discussion forum where you want to reduce spam, but for sending an email to a friend, 4 minutes is a lot of CPU time
<mappum>
sonatagreen: you listed "solutions to the abuse prevention problem", but there is one more: cryptocurrency payment channels (e.g. you pay micropayments to your peers for the resources you use)
<ReactorScram>
Especially since as I said the difficulty will probably go up in time
<sonatagreen>
oooh, yes thanks mappum
<mappum>
ReactorScram: actually while PoW was invented for spam prevention, it actually doesn't work well for that since botnet operators are more effective than legitimate users at producing the PoW
<sonatagreen>
ReactorScram, that's a good point, Bitmessage is mostly for communication between untrusted parties
<ReactorScram>
mappum: oh that sucks
<ReactorScram>
So I'll always be getting junk emails?
<mappum>
no, machine learning filtering works very well (which is why gmail does so well)
<ReactorScram>
I'd like to not use gmail one day
<sonatagreen>
if you have someone you know you want to get messages from, you can use something without spam prevention
<mappum>
yeah, one day maybe people can just run filtering as good as that on their own
<mappum>
sonatagreen: right, in that case checking a signature is the spam prevention
<sonatagreen>
botnet might be able to produce PoW cheaper, but spamming requires way more proof of work than legit messaging because you're sending so many more messages
<sonatagreen>
spamming is like broadcast publishing, except you want to send to people who didn't opt in
<mappum>
let's assume a normal machine can post once per 10 minutes, then a operator of a 100k botnet can post 10k posts per minute
<mappum>
even if it just scales linearly, they can still produce a lot of spam
<ReactorScram>
And the POW is used because bitmessage sends every message to every node, and spammers could flood the network easily?
haadcode has quit [Ping timeout: 250 seconds]
<ion>
4 minutes of latency is tough if you’re doing e.g. livestreaming.
hellertime has quit [Quit: Leaving.]
NeoTeo has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
<ion>
Plus another 4 minutes to get comments back from viewers (thinking of a distributed twitch.tv alternative).
<ReactorScram>
If I read the FAQ right, it's limited to 220 KB of attachments / message, so you'd be hard-pressed to even send one frame of HD video
<sonatagreen>
well, that's why you integrate with ipfs
<ion>
You’d send pointers to IPFS objects.
<whyrusleeping>
ion: yeah, the system i was describing earlier should be able to acheive pretty low latency
<ReactorScram>
I guess it's nice to have all these alternatives but I can't think what I would ever do with email that takes 4 minutes of CPU humping to send
<ReactorScram>
If you send over IPFS you would end up leaking your IP wouldn't you?
<ion>
Not if using a layer that provides privacy.
<ReactorScram>
hm
<ReactorScram>
And the reason bitmessage is needed is because IPFS doesn't have a broadcast layer?
noein has left #ipfs ["Ex-Chat"]
<demize>
mappum: In my experience spambayes is as good ore even better than gmail's spam filter, but it requires quite a bit of training since you're starting from scrotch.
<Gisle>
Can I set ipfs to just use one or two network interfaces?
<whyrusleeping>
Gisle: currently we assume that we can use all of them
<whyrusleeping>
it wouldnt be too difficult to change that though
<whyrusleeping>
mind filing an issue requesting that feature?
<sonatagreen>
ReactorScram, exactly
<ion>
whyrusleeping: The 0.0.0.0 and :: in Addresses.Swarm can’t be changed at the moment?
<whyrusleeping>
ion: oh yeah, you can change those
<whyrusleeping>
but saying something like, 'only use en01 and wlp3s0' isnt easy
<Gisle>
That works. I also see an address filter, any docs on that?
<whyrusleeping>
'ipfs swarm filters --help'
<spikebike>
interesting, a friend is building 100+ thin clients base on small linux boxes. Needs around 50GB for a shared filesystem (read only) for apps
<spikebike>
I think IPFS would be a good fit, but the local 16GB storage might fill
<ReactorScram>
So you're trying to have them read files off the network because they need more space than they have?
<spikebike>
right, and more reliable would be nice as well
<spikebike>
so 100x16GB to provide 57TB reliably (and fast)
<spikebike>
if places on a single nfs server there are performance issus
<spikebike>
issues
<ReactorScram>
something about that math isn't adding up for me
<spikebike>
well each client will rarely need all 57TB
<spikebike>
and there could be a main server with all 57TB of course.
<ReactorScram>
right
<spikebike>
but ideally most of the transfers would from thin client on same switch to thin client on same switch
<ReactorScram>
It would be cool to have IPFS' caching so that the most basic things like shared libraries don't have to be pulled over the network every time
<spikebike>
even a strict LRU would be fine
<ReactorScram>
Yeah LAN peering would be amazing. Not sure if it's set up yet
voxelot has joined #ipfs
<spikebike>
but someway to ensure it doesn't cause the local disk to fill
ipfs_intern has joined #ipfs
<spikebike>
well even just picking the lowest latency peer with the content would be fine
<spikebike>
or maybe lowest number of hops
<Gisle>
Filters doesn't seem to help. Is there more i need to provide than just /ip4/x.y.0.0/ipcidr/16 ?
captain_morgan has joined #ipfs
gggthgr has joined #ipfs
ipfs_intern has quit [Quit: Page closed]
david______ has joined #ipfs
gggthgr has quit [Ping timeout: 240 seconds]
<david______>
if i am making a web portal so how can i let the user to add a file to ipfs repo from the web portal
<ipfsbot>
[go-ipfs] whyrusleeping created feat/sysdiag (+1 new commit): http://git.io/vC2Da
<ipfsbot>
go-ipfs/feat/sysdiag 2a1eb40 Jeromy: WIP: system diagnostics command...
<ipfsbot>
[go-ipfs] whyrusleeping force-pushed feat/sysdiag from 2a1eb40 to 3e3f7aa: http://git.io/vC2Qg
<ipfsbot>
go-ipfs/feat/sysdiag 3e3f7aa Jeromy: WIP: system diagnostics command...
rehelmin has quit [Quit: Leaving.]
chriscool has joined #ipfs
rehelmin has joined #ipfs
david______ has quit [Quit: Page closed]
<ipfsbot>
[go-ipfs] whyrusleeping force-pushed feat/sysdiag from 3e3f7aa to 800c5d2: http://git.io/vC2Qg
<ipfsbot>
go-ipfs/feat/sysdiag 800c5d2 Jeromy: system diagnostics command...
<whyrusleeping>
spikebike: ReactorScram peers on the same LAN will find eachother via mdns, as long as the network supports it
rand_me has joined #ipfs
<ipfsbot>
[go-ipfs] whyrusleeping opened pull request #1838: system diagnostics command (master...feat/sysdiag) http://git.io/vC2NW
<ipfsbot>
[go-ipfs] whyrusleeping force-pushed feat/sysdiag from 800c5d2 to 0bb88a6: http://git.io/vC2Qg
<ipfsbot>
go-ipfs/feat/sysdiag 0bb88a6 Jeromy: system diagnostics command...
atrapado has quit [Quit: Leaving]
<whyrusleeping>
vendoring is hard...
<ipfsbot>
[go-ipfs] whyrusleeping force-pushed feat/sysdiag from 0bb88a6 to 85b5b23: http://git.io/vC2Qg
<ipfsbot>
go-ipfs/feat/sysdiag 85b5b23 Jeromy: system diagnostics command...
tsenart has joined #ipfs
<achin>
<3 progress and <3 robots that let us know about all the progress being made
rehelmin has quit [Quit: Leaving.]
<whyrusleeping>
lol
<rand_me>
I get 403 forbidden in webui for requests like "http://localhost:5001/api/v0/swarm/peers?stream-channels=true" while it works fine (200 and response) with curl or pasted to a new browser tab address bar. What could be the reason / where should I report that?
martinkl_ has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
<whyrusleeping>
rand_me: how are you loading the webui?
<dignifiedquire>
krl: true, but the globe has the big problem for me that I have to “scroll” it to see another part, this way it gives a nice overview over where people are
<krl>
yep
<krl>
also, i'm thinking about if we would need more space to the left, for custom app names etc
<krl>
or maybe an app provided icon is enough?
<dignifiedquire>
I’m not a 100% sure about the sidebar, I might make it so that you can expand it to show labels, and contract to only show icons
<dignifiedquire>
how many custom apps do you imagine to be there?
<krl>
more than will fit on a screen, hopefully!
kytv_ has quit [Remote host closed the connection]
<david_ipfs>
can anybody help me with this error, i'm using ipfs-go api - prog.go:7:2: cannot find package "github.com/ipfs/go-ipfs/core" in any of: /usr/local/go/src/github.com/ipfs/go-ipfs/core (from $GOROOT) /go/src/github.com/ipfs/go-ipfs/core (from $GOPATH)
tsenart has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
tsenart has joined #ipfs
timgws has joined #ipfs
<rand_me>
deltab: I just restarted the daemon twice, after the second time it now works. May it be related to having multiple network ifaces?
<ReactorScram>
I am having ridiculous trouble installing the Firefox gateway redirect addon. I get it from addons.mozilla.org, right?
tsenart has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
<ReactorScram>
I think something's wrong with my internet
<deltab>
rand_me: with different IP addresses? could be
<rand_me>
swarm binds to all of them (not sure if relevant)
gordonb has quit [Quit: gordonb]
<deltab>
are they all in cfg.CORSOpts.AllowedOrigins though?
<rand_me>
how can I check that?
<ReactorScram>
after countless reconnections I was able to download the 27 KB addon from mozilla's website
<rand_me>
however I all the time used localhost and it always pointed to the same ip
gordonb has joined #ipfs
tsenart has joined #ipfs
<deltab>
rand_me: is it in ~/.ipfs/config ?
<rand_me>
no
wowaname has quit [Remote host closed the connection]
<rand_me>
deltab: there's definitely something random here. If I keep stopping and starting the daemon, the issue randomly comes and goes.
<rand_me>
is there a switch for verbose output?
wowaname has joined #ipfs
thelinuxkid has quit [Ping timeout: 255 seconds]
tsenart has quit [Quit: My Mac has gone to sleep. ZZZzzz…]
captain_morgan has quit [Ping timeout: 246 seconds]
<rand_me>
found it: "ipfs log level core debug". Now trying to reproduce
ReactorScram has quit [Ping timeout: 256 seconds]
ReactorScram has joined #ipfs
david_ipfs has quit [Ping timeout: 246 seconds]
thelinuxkid has joined #ipfs
eater has quit [Ping timeout: 240 seconds]
eater has joined #ipfs
<rand_me>
how can I enable logging to /.ipfs/logs/ ? and how to change (persistently) log level?
rendar has quit [Quit: Leaving]
<Zuardi>
anyone knows if there is any current attempts to use ipfs to host a bitcoin block explorer website?
josephdelong has quit [Quit: josephdelong]
r04r is now known as zz_r04r
captain_morgan has joined #ipfs
NeoTeo has quit [Quit: My Mac has gone to sleep. ZZZzzz…]