A Hands-on Introduction to IPFS

The Interplanetary File System is going to be a big deal.

Jonas BostoenBlockedUnblockFollowFollowing

May 23, 2018

This article will be divided into two parts:

• A theoretical introduction to IPFS
• A practical guide on using IPFS

The first part aims to show you how IPFS works, while the second part will walk you through using IPFS and interacting with the already existing network. The second part assumes some technical knowledge and a *nix based system.

Theoretical Introduction

HTTP, or the HyperText Transfer Protocol, is the peer-to-peer communication protocol that governs the internet now. It’s how you access websites, watch video’s, download files. There are some problems with it however, a lot of it stemming from the fact that the current model is largely centralized.

What’s wrong with the Internet of today?

When you want to visit a website today, your browser (client) sends a request to the servers (host) that “serve” up that website, even when those servers are across the globe from your current location. This is location-based addressing, and it uses IP-addresses to show your location. That process eats bandwidth, and thus costs us a lot of money and time. On top of that, HTTP downloads a file from a single server at a time, which is way worse than getting multiple pieces of that file from multiple computers. It also allows for powerful entities to block access to certain locations, like Turkey did with the Wikipedia servers in 2017.

Juan Benet, with IPFS, plans to solve a lot of the aforementioned problems. For one, IPFS doesn’t have a single point of failure. It is a peer-to-peer and distributed file system that can literally replace HTTP and decentralize the Internet once again. Internet censorship will largely be impossible, and published information won’t suddenly disappear at the whims of a service provider or hosting network.

Security also increases — DDos attacks, for example, wouldn’t work, since they rely on attacking a central distribution system, which IPFS doesn’t have. Speed is another factor that increases. In a distributed web, every node that requests something, requests it to the node closest to him, instead of to a single, central location.

How does it work?

IPFS works by connecting all devices on the network to the same file structure. This file structure is a Merkle DAG, which combines Merkle trees (used in blockchains to ensure immutability), and Directed Acyclic Graphs (used in Git version control, which also allows users to see the versions of content on IPFS). Think of it as a large BitTorrent swarm. Imagine you want to read the IPFS whitepaper. What you would normally do is type in a URL, which can be resolved to an IP address, which provides information about the location of the file (would be the IPFS servers, if they had those), which then allows your client to make a connection with the host and get the file. There are numerous ways in which this can go wrong, a lot of which have been discussed above.

Now, imagine accessing it from the IPFS network. The file, and all of its blocks, are identified by a unique cryptographic hash of the content itself. The whole system is based around a key-value data store. This is what allows for the content addressing: anyone can host the key no matter the origin of the information. So, you would connect to the swarm and request to the network that file. It would first look to the peers closest to you, because chances are they have a copy of that file. If they don’t, however, you will connect with the node that originally uploaded the file, since he’s the one that hosts it.

You then download that file, and become a host yourself. This means you’re basically the host and the client at the same time. It also means that you only host files you are interested in. Let’s now go to the IPFS whitepaper on IPFS itself. Since chances are you don’t have IPFS installed, we’ll use a gateway. The URL looks like this: https://gateway.ipfs.io/ipfs/QmV9tSDx9UiPeWExXEeH6aoDvmihvx6jD5eLb4jbTaKGps. It starts with the IPFS gateway URL, and then includes the unique hash identifier of the content. The gateway will connect to the closest peer that has the file (which is probably itself), and then serve that file to you.

You might think downloading files from untrusted nodes is dangerous, and that would be a good remark, but since cryptographic hashes are tamper-proof, and the hash is of the content itself, you can be sure you’ve got the right file.

We make websites and web apps have no central origin server, they can be distributed just like the Bitcoin network is distributed. — Juan Benet

IPFS serves content, but content can be much wider than just a pdf-file. Application logic is content as well. In combination with other decentralized projects like Ethereum, IPFS is extremely powerful as well. A blockchain like Ethereum can then serve as the back-end of your App, while the front-end is served by IPFS. This makes for a completely decentralized application.

To really drive the point home, let’s end with another example. Imagine a room full of people that are collaborating on a Google doc at the same time. With HTTP, every time a change is made, the information is sent back to the server, and then from the server back to every other student. You can imagine the bandwidth that is wasted and the unnecessary latency that is produced, especially for people on the same local network.

With IPFS, however, the information wouldn’t have to make all those unnecessary travels, it would just be shared real-time across all those people on the same network. Even if the Google servers were down, since every node is now being its own host. This would also significantly reduce latency and bandwidth.

This might also be a good time to explain why it’s called the Interplanetary File System. Think of it this way, if the group of people we just talked about were on Mars, they would be able to get parts of the internet usable on Mars like this: one person requests Google docs, or any other file, from any node on the IPFS network back on Earth. This would take a while (about an hour), but once the file has arrived (and remember, this could be anything, Wikipedia broken down are also just files), the person that originally requested it becomes a “seeder” in the network from which other nodes can get the file. If this process goes on long enough, eventually most parts of the internet will be available on Mars without having to wait hours after sending a request.

IPFS is still in alpha, and a ton of developments are still underway. One very cool one is Filecoin. Filecoin is a cryptocurrency that incentivizes nodes on a network to store (seed) content with economic rewards, so that the accessibility of content doesn’t depend on the whims of untrusted nodes.

Below is a video in which Juan Benet explains how IPFS and Filecoin would work together: