By Mark Solomonik
With the goal of becoming a permanent archive of growing internet history, the InterPlanetary File System (IPFS) is starting to take hold among the growing fear of increasing web centralization. Web centralization is the idea that soon all internet users will be heavily dependent on only a few companies, such as Google or ISPs (Internet Service Providers). Reminiscent of Oct. 26, 2009, when Yahoo took down 38 million Geocities websites, the peer-to-peer file sharing technology IPFS aims to protect historical data and websites from being erased by the decision of only one company.
Although HTTP (Hypertext Transfer Protocol) has been the internet file transfer standard for the last 15 years, according to the IPFS white paper, HTTP struggles with new computing requirements that are starting to take hold, such as dealing with the sudden disappearance of important files, versioning large datasets and high-definition video streaming. IPFS attempts to solve these issues and replace the HTTP standard by using a peer-to-peer file transfer system, such as the popular file-sharing technology BitTorrent.
“It’s a cool idea. Peer to peer was really explored a lot in the ‘90s. And some of the people who really pushed for peer to peer got in trouble in a big way,” computer science teacher Debbie Frazier said. “There’s a trick to it. My friend gives me access to their movie that they bought. That’s peer to peer.”
A peer-to-peer network is when multiple computers can exchange information without having to access a third party server. IPFS’s peer-to-peer model helps decentralize the internet by preventing any single entity from deleting these files or having too much power over the network. According to Cloudflare, who recently created an IPFS gateway, IPFS also allows anyone to post files on the network without paying expensive hosting costs.
IPFS aims to take the concept of a simple peer-to-peer network to a larger scale by following the footsteps of BitTorrent, attempting to create an internet-like network of accessible files by coordinating members of the network into holding pieces of files and distributing them to each other when needed. IPFS mimics BitTorrent by using a “tit-for-tat” strategy to reward members who contribute in holding data and punish members who do not. IPFS also uses concepts from Git, a version control system, to model file changes on the network over time.
IPFS content-addressing and content signing also prevent DDos attacks, which involve overwhelming a target website with artificial traffic, by preventing untrusted nodes which do not contribute to the network from accessing data.
“We use content-addressing so content can be decoupled from origin servers, and instead, can be stored permanently. This means content can be stored and served very close to the user, perhaps even from a computer in the same room,” Juan Benet, founder of IPFS, said in an interview with Techcrunch. “Content-addressing allows us to verify the data too, because other hosts may be untrusted. And once the user’s device has the content, it can be cached indefinitely.”
IPFS is also free from using IP addresses, which are linked with security risks such as IP address spoofing, when someone uses another IP address and pretends to be that IP address from another host on the network. Another concern is reused IP addresses, when a user replacing an old user in a network does not receive a new IP address and instead receives the old one, potentially allowing them access to the old user’s files.
Using a peer-to-peer system on a larger scale is not without its issues, as computer science teacher David Greenstein points out.
“You’re actually adding to a reliability issue. What happens if one machine goes down, and they’re in the middle of a database?” Greenstein said. “You need a continuous piece of data, and you’re reading across server to server to server and then suddenly one server goes down, then you need redundancy. And then that means that the cost goes up.”
In order to reliably contain data on the network spread out throughout many different systems, IPFS uses redundancy and holds the same data in multiple places in case one node goes down. Although IPFS does track rarity of certain file pieces, or “objects,” and prioritises sending the rarest objects first in order to minimize load, the underlying redundancy of the network remains.
Although IPFS’s redundancy can be a strain on the network, according to a study from Oregon State University in 2009, video streaming efficiency could be improved using peer-to-peer. Using the peer-to-peer network model proposed in the study, they could save up to 60 percent of bandwidth over using traditional CDNs. One way they can accomplish this is by using proximity optimization, when the node geographically closest to the user is chosen to send the data. The model also uses a path diversity streaming protocol, sending parts of the file from multiple nodes if there is congestion. The concept of path diversity streaming protocol can be applied to IPFS, which according to Techcrunch, can be accessible despite sporadic internet service. As of 2014, Netflix was looking into researching these types of peer-to-peer technologies for streaming.
Although decentralization can make it easier and cheaper to publish files, make the internet more resilient and hold files permanently, Frazier acknowledges potential issues with the decentralized nature of IPFS.
“It’s kind of cool, but at the same time we are running that risk of not crediting people appropriately for their artistic work, so there’s that angle of peer to peer that’s kind of sketchy, that someone needs to figure out,” Frazier said. “Part of the issues that have come up with peer to peer is a lack of royalties and copyright acknowledgement of those artists.”
Despite these issues, Frazier believes in the potential IPFS has to offer in terms of decentralizing the internet and removing the reliance on individual organizations.
“I think peer to peer architecture is brilliant. I think anything that removes the authority for a single power, single mindset or a single goal minded body to decide what we should or shouldn’t see on the internet. I think that’s fantastic,” Frazier said. “Because I think the internet was really designed to be a community where anyone can share, anyone can post.”