What is IPFS?
A fully decentralized, distributed and encrypted file distribution platform.
>why would one use it
* Distributed and decentralized: no single point of failure, censorship-resistant.
* You can have a mutable address (i.e. always points to the latest version of a site), or a static address (points to a specific file). Yes, you can host sites over IPFS.
* Peers are found fast for new downloads. You don't need to wait that much to start a download.
* You can watch your animu while it downloads, I watched few episodes that way and it didn't even buffer.
Presentation of ipfs by its lead designer:
https://ipfs.io/ipfs/QmSd1buim52M
tQP1SH2XhN7FC8c1K
dFrFKBNmgdUKFdss8/Stanford%20Seminar%20-%20Juan%20Benet%20of%20Protocol%20Labs.webm
(over ipfs)
>how to upload a single file
$ ipfs add ./$file
Access it at localhost:8080/ipfs/$outputted-hash
>how to upload a dir
$ ipfs add -r ./$dir
Access it at localhost:8080/ipfs/$last-outputted-hash
>how to make the thing mutable
$ ipfs name publish ./$file-or-dir-hash
Access it at localhost:8080/ipns/$output-hash-aka-peerid (it's ipNs not ipFs)
To update, publish another hash and it will be available at the same IPNS address.
>gateways (how to access IPFS if you don't have it installed)
https://gateway.ipfs.io/
https://gateway.glop.me/
https://ipfs.borg.moe/
Add an ipfs url at the end to use them.
>I2P and Tor support coming soon™. We need that thing anonymous so pls halp.
https://github.com/ipfs/go-ipfs/issues/1118
https://github.com/ipfs/notes/issues/37
https://github.com/jbenet/go-multiaddr-net/issues/7
>Sites with content on them
/ipns/gindex.dynu.com/ site index
/ipns/QmUqBf56JeGUvuf2SiJNJahAqaVhFSHS6r9gYk5FbS4TAn anime tracker
threadly reminder to pin files that you care about.
>BOTNET
fucking dropped
>>51848018
why
also, to ccd0(gut who makes 4chan-x,i know you lurk here), how about making ipfs/ipns hashes auto link?
>>51847987
Isn't this basically 9p servers?
There were a lot of GNUnet threads, now it's IPFS, what happened? Is IPFS better?
>>51847987
How can I get on the ipfs train?
>>51848190
delete system 32
>>51847987
how does IPNS (the nameserver) work?
>>51848156
I used to be in all those GNUnet threads and the main consensus among those who left for IPFS threads:
>GUI
>no Windows binaries
>slow downloading at first
>default speed capped way too fucking low
>why the fuck do I have 2 peers after all this fucking time
>it's THE CURRENT YEAR why do things take so long why are things buggy who uses cli etc.etc.
>>51848670
At this point, IPNS just represents an IPFS node.
ipfs name publish <anything>
will always return the same hash for your node.
Botnet
>>51849979
funny way of spelling awsome
>>51848670
not very well
how do you think this will play with the anime torrenting community?
>>51850298
Not well, it doesn't even try to provide privacy/anonymity
>>51850355
when would you need that? among private trackers? those things aren't private for this. among the regular stuff like nyaa, what do you think?
>can't use it on the most popular operating system
dropped
>>51850380
I think people will stick to torrents in that case, but if anything were to happen that would prompt a mass shift from torrents, e.g. huge gumment crackdown on copyright infringement, people might switch to IPFS, but it won't be very big or for very long
>>51850298
could possibly be an alternative to torrents
>>51850355
neither do torrents
>>51850436
>implying
>>51850436
go to install then to "see more builds"
>>51850501
If only GNUnet weren't such a broken pos with apathetic developers, then we might have had GNUnet be a real successor, since it takes privacy (and complete anonymity) very seriously
>>51850529
pretty sure the idea is that if you want anonymity just pair ipfs with tor instead of reinventing the wheel
>>51850902
>use tor with ipfs
The tor developers have repeatedly stated that you shouldn't use bittorrent/other P2P traffic over it
>>51850902
>instead of reinventing the wheel
Why don't we just stick with HTTP if we want to avoid reinventing the wheel?
>>51850982
ipfs is supposed to be a http replacement, not a tor/ipfs/gnunet/freenet/etc replacement
>>51850967
ok
>>51851045
Right, so why are we bothering reinventing the wheel (i.e. HTTP)?
>>51848670
It's a signed IPFS object that is distributed among your immediate peers. It's not content addressed, but rather the key of the object is the hash of the public key used to sign it. The object also has a TTL, so it needs to be republished regularly (every 48 hours, IIRC). The daemon should handle the republishing for you, but it might not do that if it's hanged.
That's what I've gathered, at least. I should probably read the spec a bit closer.
>>51851092
>The InterPlanetary File System (IPFS) is a new hypermedia distribution protocol, addressed by content and identities. IPFS enables the creation of completely distributed applications. It aims to make the web faster, safer, and more open.
>https://www.youtube.com/watch?v=skMTdSEaCtA
I guess the main goals would be to replace http with something resistant to censorship, with no single point of failure, and better scalability.
>>51850967
Why would HTTP be ok, but not IPFS?
>>51851235
Most people aren't getting huge amounts of data over HTTP. Since bittorrent is by design for huge amounts of data, it shouldn't be used over tor. IPFS would be fine with small webpages and such, but anything big such as that which you'd normally torrent, it bogs down the network
I can access all IPNS links on gindex except the one for the Totoro tracker. I talked to someone with the same problem yesterday.
Totoro guy, if you're reading this, please restart your daemon and/or republish the site.
>>51851499
But you wouldn't run Bittorrent over Tor either and IPFS is primarily a HTTP replacement. The advice should be "don't download large files over Tor".
>>51851565
Considering the first post in this reply chain was about how the anime torrenting community might be interested, that would be why they might not be hugely jumping to it
>>51851596
Oh, right. Missed that, my mistake. Yeah, this is not suitable for an anonymous bittorrent replacement, although it supports any underlying transport protocol if something better suited/less important than Tor comes along. It can be used as a non-anonymous replacement today, though.
>>51851636
An ideal future of the web would be IPFS replacing HTTP and GNUnet replacing torrents (if the fucking devs ever get off they're asses and make it fucking work right)
>>51851666
anything GNU is destined for stagnation and eventual death
>>51851666
I mostly agree, but there are use cases for torrents where anonymity is not a priority, e.g. downloading distro isos. And we already use HTTP for some high-bandwidth things that ought to have used torrents from the beginning (like Youtube). Both of these should be handled with IPFS. But then you have the problem with GNUnet only being used for "illegitimate" communication, which is a problem, since governments could ban their usage without much backlash (and smear opponents by claiming they support pedo terrorism or whatever).
IMHO, the right solution is I2P with some sort of anonymity layer used by consumers, but not necessarily by businesses (that might prioritize high bandwidth over anonymity).
>>51852056
>right solution is I2P with some sort of anonymity layer
Err, I mean IPFS, of course. The anonymity could be I2P.
>>51852056
People all along should be using GNUnet and tor sites and I2P sites etc. for legit purposes instead of settling for "well, I've got nothing to hide, so I don't care about my privacy"
Why must you be hiding something bad if you're using a service such as GNUnet or tor? Why does your ISP have to know who you're talking to? Why isn't privacy the default?
>>51852146
the problem is that the anonymity makes shit slow so nobody wants to use it for normal shit
>>51852146
I agree, but in order to do that, you either have to sacrifice speed or store/tunnel someone else's traffic in order to participate in the network. This is something that privacy conscious individuals might be okay with, but not regular users, and certainly not businesses.
I agree with Sunde's position that the Internet (not just the Web) is fundamentally broken. Maybe we can fix it when the bubble bursts and it's time for IPv7 to roll out, but for now, IPFS+optional anonymity is a huge improvement.
>>51852274
What didn't help the speed was the default speed cap being set to 5k/s down and 1k/s up, which is the devs fucking up defaults and not giving easy ways to set the options when you set up the first time. The anonymity had less to do with it than that.
>>51852406
How is IPFS + tor better than IPFS + GNUnet (assuming GNUnet stops being so buggy eventually)?
>>51852490
>How is IPFS + tor better than IPFS + GNUnet (assuming GNUnet stops being so buggy eventually)?
Tor is more mature and has more users, but other than that probably nothing.
>>51848134
You can already do
https://ipfs.io/ipns/gindex.dynu.com/
or
https://ipfs.io
/ipns/gindex.dynu.com/
>>51852490
With IPFS+Tor, everyone uses the same address space and two users can communicate with each other with only one of them choosing to be anonymous, and the other one being totally oblivious. This prevents fragmenting the network into two parts, and also helps avoid the "only bad users need anonymity" mentality.
>>51852534
GNUnet could have had a chance to boom with those threads a while back, with /g's/ interest perhaps there could have been a surge in development and available content for it, but the developers failed to capitalise on it and do anything useful before it waned.
The lead developer even considers a huge, compatibility-breaking change for 0.10.2, which has been behind schedule for more than a year and a half. His response to concerns that everything /g/ published in the spike of interest would need to be republished, which would probably kill whatever interest was left, was to "get 0.10.2 out as soon as possible," since that's obviously going to happen.
>>51852588
I still think GNUnet would be better suited to providing complete privacy than IPFS+tor, but it's entirely moot since GNUnet's never taking off anyway.
The stigma against anonymity won't go away until people start doing normal things anonymously. Nearly nothing else will change that.
>>51852727
i really would like to see both survive, although gnunet is really not polished or for the most part functional.
>>51852779
>i really would like to see both survive
Same here, actually, now that I think about it. The monoculture that we've seen with HTTP lately is worrying.
>>51852779
GNUnet worked, and still works, if you stick to the CLI like normal, but the GUI has no excuse for being so God-awful, and I have no idea how the fuck to get more peers or why peers won't connect with me. So yes, it works, but it doesn't work nearly enough to warrant hope anymore, unless /g/ can manage to contribute (lol) more than the logos its made.
Actually, maybe we could start with the fuckhuge icons in the GUI. At least it'd be something.
someone should make an IPFS imageboard
>>51852885
what do you mean monoculture? are you reffering to the http protocol being the most and pretty much only used protocol in its class?
>>51852943
Or an ipfs youtube
>>51852943
How? IPFS is pretty much only static shit. You'd probably have to use something else to actually store the data.
>>51852922
even the cli is a little borked on linux. trying to upload something leads me to endless library errors and crashing. i would like to work on it but i have only just begun learning how to code, so i am useless
>>51852953
Yep. That in itself might not that much of a problem, but it promotes every application moving into the browser as a JS blob, and ultimately moving to remote servers along with your files. I miss real programs.
>>51853002
>the internet is becoming more centralized
is this true?
>>51852997
images are static shit. keep the messages and threading on http, keep the content on ipfs.
>>51852999
What distro? It appears that Arch/Parabola users (myself included) had less troubles than Debian/Ubuntu/Trisquel users did
>>51853034
So, you would use HTTP to upload images/data to a server, use HTTP to get the IPFS hashes of images/data, and IPFS to actually download the images and data?
Sounds convoluted.
>>51853038
it was a minimal arch install with all my stuff on it. it was throwing errors for libraries installed. i cant remember which ones but i can look back into it soon
>>51853034
That would be an insanely good idea. You could also serve the threads themselves on IPFS (every post is a small IPFS object, a thread is a collection of objects; alternatively, a post is an object whose parents are the posts it's quoting). The only things that need to be served over HTTP is a small site that loads the IPFS content over IPFS (using JS) and a simple HTML form for posting.
>>51853078
Did you use pacman or compiled it from source? Using just pacman worked for me, though my Parabola install is hardly minimal, at the moment anyway
>>51853076
It sounds convoluted, but that's probably the best way to use IPFS. Combine it with HTTP to cater to both protocols' strengths.
>>51853103
i cant remember if it was from the repos or the aur. it was probably in the repos though.
>>51853113
Why not just use an IPNS? I'm not 100% on how they work, but would that remove the drawbacks of HTTP in case the main server goes down?
>>51853088
but i'm assuming you'd want to have some oversight on the posts, in case you want to ban illegal content and crapflooders. the content would still be there, but the "brand" of the board is protected from evildoers.
>>51853161
if you were using ipns to control them, the singular ipfs posts would always exist but they could easily be removed from the ipns index.
>>51853225
ok. i've just read up on it a little bit, but it's intriguing.
does anyone want to take a first crack at it? sounds like it could solve a lot of weak points in how things are run today.
>>51853161
True, but if you have a central HTTP server handing out IPFS/IPNS objects, it can pick and choose what posts the current thread object points to. If a post is deleted, you only need to remove it from your list of posts, and rehash the list. Here is a description of how to do it with directory-like objects, in case you're interested: http://127.0.0.1:8080/ipns/QmZhFdzvpes5XsQB9g6R5
qKYsdt32VSnr6vUnjifV
PXNxn/posts/2015-11-27-directory-listing-IPFS.html
I imagine the thread objects could be similar. Maybe a flat "thread is a directory of posts" structure is the simplest implementation, since the CLI tools already support them, but it's possible to create arbitrarily complicated DAGs of arbitrary objects if you want.
>>51853257
i would volunteer but i would rather not run a public image board of my ip
>>51853284
you can set it up to withhold posts until you approve them
>>51853265
that link doesnt work for me :^(
>>51853333
they would still have my ip and i dont like manually sorting through cp
>>51853349
You can just use tor :^)
>>51853335
Oh, damn. Here's the IPFS link: /ipfs/QmPGrRGRX7paZ6
NhS3vgs5jSyaS7Bio78s
15KkWpnkyTxA/posts/2015-11-27-directory-listing-IPFS.html
I'll try republishing it.
>>51853356
that is true anon. then give me a months time and an unlimited supply of jelly beans.
in all seriousness i need something todo so i will try atleast
is there a way to bootstrap an imageboard in written in javascript from ipfs/ipns?
>>51853373
Whatever you do, MAKE SURE YOU GIVE IT THE PROPER LICENSE
>>51853393
g-g-gpl?
no no no, m-mpl?
bsd?
dont hurt me senpai
>>51853393
wtfpl?
>>51853257
>>51853370
I'd be really interested in helping out, but I'm clinically burnt out and have ten other projects that I need to be working on.
>>51853367
Republished. I also set up a cronjob to do it regularly. Apparently, just running the daemon wasn't enough (though I remember reading that it should be, somewhere).
>>51853393
apache or maybe beer license? is idgaf good?
>>51853427
>>51853436
>>51853461
Use a FSF approved license and LibreJS.
old /g/ would have had a logo made up by now
>>51853473
no thanks
>>51853488
ipfs/QmV4JU8mw3TR7kDtw
kkevXm6KyQFAbSJWZKoto8NXe8iGR
Some ideas:
In the flat directory structure, every post could be a directory containing "files" as follows:name_hash name
email_hash email
image_hash image
date_hash date (required)
content_hash content (required)
Where *_hash is the IPFS hash of the corresponding object. You could also combine the name, email and dates into a single metadata object. Having the content be a separate object might help with deduplication, though (every instance of a copypasta will point to the same object).
Every such directory has an IPFS name h_n. The thread would be a directory:h_1 1
h_2 2
...
h_n n
where the numbers are the global post numbers. To add a post, you create the post object, pin it and add it to the thread object with ipfs object patch. You then do the same to add the thread to the board index. To delete a post, you unpin it recursively, modify the thread and index with ipfs object patch. When a thread 404s, you unpin the thread object recursively and patch the index.
You can't manually unpin images, since other posts might contain the same image. I'm not sure if recursive unpin takes this into account. Otherwise, you'd have to do the recursion yourself and check if it's referenced by any other object before unpinning it.
This looks pretty doable, actually.
>>51853794
You could publish the board index to IPNS, but since IPNS is kind of slow and you're going to rely on a central HTTP server anyway, it's better to let the client request the hash of the current index via HTTP. The JS could either include the full JS implementation of IPFS that the developers are making, or it could do HTTP requests to localhost (we should avoid using gateways, since they are run by volunteers that can't afford the bandwidth).
>>51853891
The JS should display the IPFS hash of the current state of the thread, and of every post and image, on the web page to make it easy to for users to pin things.
/ipfs/QmYbmoDHDNgqoDbLC
4vMyoMH5dYQdrxL1wh63x9rhSC6Zf
martian: /ipfs/QmP2H6CsEAa8uEeoYrah
wJ3FaustNaDoSiVu1ysptW2a8f
maritan: /ipfs/QmR1UMeznud29fB5
5AWpU6MJvaPXw9FUR4J2Pj1Jz9q7Tm
This is gay.
>>51851235
>>51851499
It's not about the amount of data
Http is server client architecture
Tor can anonymize server client connections
Ipfs is peer to peer
For CANNOT anonymize peer to peer connections sufficiently OR securely enough to be of any use at all.
Side note: bitcoin is very vulnerable when used with tor, there is a paper about it.
>>51853794
Any progress on this overnight /g/?
>>51848018
I'm running ipfs daemon in firejail.
It actually doesn't need much. it just needs access to some storage and obviously some ports to listen on.
>>51847987
>localhost
>gateways
that post should mention browser addons that register the ipfs:// protocol
>>51852997
>IPFS is pretty much only static shit.
ipns gives you the dynamic part. think of an image board as a tree of posts, whenever new posts arrive they form the new root of the tree, have ipns point at the root. readers just re-resolve the ipns every N seconds to update.
>>51853373
almost trivial: in a semi-centralized manner where one trust anchor is responsible for accepting new posts and refreshing the ipns entry.
that means everyone hosts the data, but only one person (key-holder) does the updating
non-trivial: some sort of blockchain-y approach of reaching consensus which posts are part of a thread.
>>51858100
Turning it into a blockchain would make it impossible for the mods to outright delete posts, so they'd have to also publish a "post banned, do not download" list that the client checks before getting an object. The gets then have to be non-recursive. This might be a better option, because it will be clear where in the history a specific pin of a thread is situated.
>>51858218
if you want mods, who are a form of trust roots, you don't need a blockchain.
I just read the paper.
Tell me about IPFS, why even use it? I still don't get it. Seems like a huge overhead to me. And if I spam random hashes, will the system hold up? It seems like there's a hundred and one places to have big exploits for you to deal with, that can crash the entire file system with no survivors.
>>51858718
ipfs uses a DHT and 'spamming' a DHT is pretty ... useless?
Also: /ipfs/Qmb9sTy6LiUbD
VKCexz4L2QbUuSbBCq7UaJXQHQvB2Bytr
all glory to ipfs!
>>51858824
Okay okay okay.
Just help me understand what actually goes on when I access an ipfs link
>>51858871
ipfs consists of 'object': Eg: If you add a file, the client will:
* split up the file into smaller parts and add it in your local storage (so that your node can seed the data)
* Calculate a hash of the contents
The hash can then be used to access this file from some other node as it globally identifies the contents.
So: How does the 2nd client get the data?
That's where the DHT is used:
Your node (= the one you used to publish the data) announced itself to the DHT for this hash.
A client looking for the data will walk the same path in the DHT and find the IP of a node which hosts the content and will therefore be able to download (and later seed itself) the contents.
If you never worked with DHTs, i'd recommend to read the kademlia paper:
https://pdos.csail.mit.edu/~petar/papers/maymounkov-kademlia-lncs.pdf
>>51859094
that really helps.
Thanks mate.
one question is where do you expect all the bandwidth and disk space to come from on an all-volunteer system?
you can't expect web browsing nodes to stick around to seed. would the theoretical "rest of the network" be able to bear the burden of storing and serving something like this?
>>51847987
Here are lots of mod files (Amiga music):
/ipfs/QmXudYwpCqUk77HissMhowbAatLUQyxNAHKWzyFhBfkGaX
>>51859308
>you can't expect web browsing nodes to stick around to seed.
Why not?
>>51859308
A node who 'published' something will never delete the object (unless the user unpinned it).
Every other node accessing it, will keep it in its own cache (10gb per default).
So popular files will automatically get lots of bandwidth (as they are distributed across multiple hosts) and higher availability.
>>51857997
i2p can, though.
>>51858145
S M A R T
M
A
R
T
C O N T R A C T S
O
N
TR
AC
TS
Reminder, the totoro tracker is down.
>>51859360
assuming a bright future of ipfs being integrated with browsers at some point, unless the browser is kept open, there's no seeding potential.
mobile also poses a unique problem, the least of which is data caps.
IPFS is the future. Switch now or die like the dinosaur you are.
what about WebTorrent?
>>51862681
>crippled version of bittorrent
>doesn't bring anything new to the table
why would I want less than what I already have?
>>51858824
Thank you for umaru-chan
>>51855491
>>51856519
how do i stream it?
>>51864510
mpv [url]
Ok to make sure i understand this properly. if i upload a video and someone else have uploaded the exact same video, will the hash be the same? and would this mean when someone over ipfs protocol is resolving the hash i dont even need to have my PC on if someone else is "seeding" the same hash?
>>51865022
Yes and yes
is there a way to git a hash without doing "add" first and seeing the result?
>>51859574
If that ever takes off I can see router vendors jamming cheapo hard drives or SSDs in to consumer stuff for local IPFS cache. "This router makes IPFSBook load faster!!!!!1!111"
>>51865493
The -n option to "IPFS add" does what you want.
in b4 interplanetarychan..
>>51866563
We need to agree on a smart contract framework first, otherwise we can't have a fully distributed, self-updating website and need to rely on a centralized authority.
>>51847987
how do i browse those sites?
i have the deamon running already
>>51867167
localhost:8080/[ipfs url] in your browser. For example, localhost:8080/ipns/gindex.dynu.com
>>51866990
nobody will want to use a website that's fully distributed. you need centralized authorities (note the plural) to link to the content, otherwise it'll be overrun with illegal shit and spam.
even 4chan had to give in and add a captcha (one that's operated by Datamining Central, no less..)
>>51868505
If you don't know shit about distributed website technology, don't talk about it.
>>51868611
and if you don't know shit about running websites that people actually visit, don't talk about it.
>>51868635
Oh the irony.
Can we get back on topic now?
>>51868719
>Can we get back on topic now?
fuck off, i2p is better
>>51869927
>tripfag
Check
>retarded
Check
>>51869955
lel so mad at chu
>>51858076
All the IPFS integration addons just redirect them to localhost 8080 instead.
>>51870468
The IPFS team has been approached by an unnamed major browser vendor to offer ipfs integration (real one, not just localhost redirect) Soon (tm), however.
>>51870534
>The IPFS team has been approached by an unnamed major browser vendor to offer ipfs integration
So almost definitely Mozilla?
>>51870576
Actually I'd think google, since mozilla is full-on chrome-copying nowadays and wouldn't want to add any feature that doesn't involve tracking their users or integrating even more ads in the browser.
>>51870595
I suppose you're right. I usually don't think of Google as likely to adopt such new things into their well established products.
But I guess Mozilla more fits that bill nowadays.
