InterPlanetary File System IPFS

[GreyGeek]

Out of concern for increasing levels of Internet censorship I have been re-re-evaluating P2P distributed web technologies.

There have been several in the past but none have taken hold. Currently, there are two contenders: IPFS and ZeroNet.

In reading about IPFS I've come to an opinion that it is still underdeveloped and too complicated for the average user to use. Like Btrfs, IPFS (and Dat) use a Merkle Tree, in which nodes have encrypted hashes.

I came across an article by Tom MacWright which expresses fairly well my understandings and concerns about IPFS. He first compares Dat with IPFS and then launches into his IPFS analysis.

The IPFS home page is here.

Here is a video. It won't take you long to realize why IPFS isn't going anywhere soon, for two main reasons: It's dependency on the command line, and the inability to determine from the long hash of a destination what its content is.

With extra long alphanumeric public encryption keys representing websites what IPFS needs is a form of local DNS where a human readable name of a website is linked to it's public key so that the user view its list of human readable website names just as if it were a search engine listing, click on a URL and have the associated key used silently in the background to call down the site's webpage.

ZeroNet, as you would see if you ran the demo, suffers from the same problem. The public Merkel hash keys get in the way of the user. And, very few users will resort to browsing the web using a terminal. Even Lynx, the old terminal web browser, is better and easier to use than IPFS or ZeroNet.

Sadly, even with my increasing concern or corporate destruction of citizens political rights, I am forced to conclude that at the present, IMO, IPFS is not ready for prime time.

[#]P2P[/#]

[vinnywright]

maybe not ready for prime time ,,,,but it looks interesting and quite cool ,,,,as it acts like a local file system ,,,,sort of

but this may introduce security concerns as well ,,,,,he made a statement ,,,to the effect of it running whatever you clicked on locally ?

VINNY

[GreyGeek]

maybe not ready for prime time ,,,,but it looks interesting and quite cool ,,,,as it acts like a local file system ,,,,sort of

but this may introduce security concerns as well ,,,,,he made a statement ,,,to the effect of it running whatever you clicked on locally ?

VINNY

A local file system indeed. With either protocol when you browse a remote website you are actually downloading that site and hosting it. That's how it becomes P2P and how you can browse a web site when the Internet is down, IF you've already visited it once. Depending on your Internet connection your first browse of a site could be very slow if your bandwidth is being eaten up by downloading the site. When you first install these protocols on your box you are given a set of encrypted private and public keys. Only you or folks you give your private key to can change your own local web site, if you made one. If you visited another website it becomes resident on your HD and the P2P protocol checks for changes and updates automatically take place, verifying that the downloaded file changes for that site are encrypted by the same key as the original download was.

IPFS has what it calls IPNS, their form of DNS, but the IPNS is local for the sites you visited, not some distant server containing all known IPFS domain names accessed by all users. If you don't visit a site you don't know about it except through

I'm currently enjoying a 60+ Mbps bandwidth the allows an actual browsing speed of about 30Mbps on average, because of the way Spectrum spools up a connection, and an actual file download speed of about 8Mbps down and 6Mbps up. In a month or two I'll be switching to a fiber optic connection that is rated at 100Mbps down or up for all traffic with no data cap. I have 6Gb of RAM and 4 cores with 8 threads. I have about 1Tb of free HD space. I could be a node in an IPFS or ZeroNet P2P network without breaking a sweat, but a lot of Internet users would not find their connection usable. 4Gb of RAM and/or 2 core with a 20Mbps or less connection speed wouldn't find either P2P protocol very friendly on their hardware.

IF you visit a LOT of other sites created using IPFS or ZeroNet, but not sites you access directly using IP quad addresses, you can eat up 10-50Gb of disk space or more "hosting" those sites, depending on exactly how many other P2P sites you visit. A person whose Kubuntu or Neon installation is on a 100Gb HD, or less, couldn't begin to run IPFS or ZeroNet without running out of room, especially if they are using Btrfs as their fs. You are hosting your own web pages on your own computer after you publish them using IPFS. When other people browse your web pages on your machine they upload your pages to their computer as they are browsing yours. Other people may browse your website on those other machines. If someone takes down your Internet connection those who visited your site will still be hosting your pages. When I was trying out ZeroNet I noticed that after I had visited a few sites I had over 90 "peers" connected to my laptop, browsing the pages I had downloaded in the background (not by any deliberate action by me other than just visiting the remote sites), while I was busy visiting yet other sites. Before I closed my ZeroNet connection I was "hosting" over a dozen sites and had close to two hundred "peers" visiting those pages on my computer. With IPFS not so many because their aren't that many remote IPFS sites except for the sites that MacWright mentioned. The blue sphere with rods sticking out of it is a graphical representation of peers attached to your IPFS connection. I never had more than 6-12 and most of those looked like developers.

As I said, IMO, neither IPFS or ZeroNet are ready for Joe and Sally Sixpack. And, my brief into to the Dat P2P system suggests that neither is it. MacWright was right. Even though IPFS is burning through over $50 in cash from venture capitalists it is still 2-5 years away from going live to the general public. That's an entire computer generation or longer.

[GreyGeek]

Several other P2P technologies have appeared and one called Mastodon has taken off in Japan.
These technologies are listed and discussed in a paper on the subject from MIT graduate students.
http://www.ethanzuckerman.com/blog/2...uncomfortable/

Our team at the MIT Media Lab – Chelsea Barabas, Neha Narula and myself – are releasing a new report today on distributed publishing, titled “Back to the Future: the Decentralized Web”
http://dci.mit.edu/decentralizedweb

We end up speculating that the main barriers to adoption of decentralized platforms aren’t technical, but around usability. Most distributed publishing tools are simply too complex for most users to adopt. Mastodon may have overcome that problem, borrowing design ideas from a successful commercial product. But the example of lolicon may challenge our theories in two directions. One, if you’re unable to share content on the sites you’re used to using – Twitter, in this case – you may be more willing to adopt a new tool, even if its interface is initially unfamiliar. Second, an additional barrier to adoption for decentralized publishing may be that its first large userbase is a population that cannot use centralized social networks. Any stigma associated with this community may make it harder for users with other interests to adopt these new tools.

Mastodon is big in Japan… at least, in one subculture. Whether that bodes well or ill for widespread adoption of the platform more globally is something we’ll be watching closely as we work to understand the future of distributed publishing.

https://mnm.social/

[GreyGeek]

Well, still hoping and still playing with IPFS.

I started the daemon and played with it a while.
I opened a browser to play with
https://ipfs.io/ipfs/Qm.....
and
http://localhost:8080/ipfs/Qm.....

I called up the IPFS main webpage and clicked on the connections link, to show the number of peers I was connected to. The number started at a dozen or so and then began building, reaching a maximum of 569 peers before falling back to 420+. All in the space of 20 minutes. Here is a snapshot of my browser showing the peers represented as blue spikes coming out of the globe of the Earth at their locations. On the right side of the globe is the listing of the peers. If you click on one of the items in the list in pops down the details and gives the public key of that peer. You have to copy that key and then paste it in front of https://ipfs.io/ipfs/ or http://localhost:8080/ipfs/ or htpp://localhost:5001/ipfs/... But, you have no idea what the content of that site is. I still have to find out about how to use the IPDN (the IPFS domain name server). Joe or Sally Sixpack could never get IPFS running in its present state.


Finding all the information necessary to run IPFS isn't at any one particular location,despite the use of the term "Documentation" at the IPFS home site. For example, what ports must you have open to run IPFS? These:

# Ports for Swarm TCP, Swarm uTP, API, Gateway, Swarm Websockets
EXPOSE 4001
EXPOSE 4002/udp
EXPOSE 5001
EXPOSE 8080
EXPOSE 8081

If your router doesn't have a firewall then you'll have to use gufw to open these ports. If you have an SPI firewall on your router you will have to open these ports through it and leave ufw off.

[GreyGeek]

I can't get off the page with the globe on it. Trying to connect to one of the peers gives:

https://github.com/ipfs/go-ipfs/issues/4128

bitswap: couldnt open sender again after SendMsg

So, there I sit. I can only see files on my local machine.