This past April, JPG launched its On-Chain Canon, a collection of works considered canonical within the on-chain NFT movement. Originally proposed by the artist and data scientist known as Chainleft, what followed was a passionate and highly technical discussion about what counts as “on-chain,” played out on JPG’s Discord with input from dozens of people across the NFT space. In the interest of sharing this important information in a way that might benefit non-technical NFT collectors, JPG organized a conversation with speakers from across the ideological spectrum. Here we present some of the key takeaways from that discussion.

‍William M. Peaster: Chainleft, how would you define an on-chain NFT to someone unfamiliar with the conversation? What is the starting point? 

‍Chainleft: Even today, it’s something that we are trying to define. I’ve been in this niche part of the NFT space for a long while, and my opinion can change through these ongoing conversations that we’re having. 

To me, it’s mostly about the permanence and persistence of an NFT. 

As we’ve already seen, some NFTs’ metadata have essentially vanished due to server failures or being altered, or sometimes even unpinned IPFS content. For someone who comes to me and asks, “what is an on-chain NFT?” I respond that it’s a type of NFT whose art isn’t possible to be lost. But there is also a lot of nuance there [because] you can say: “Oh, that’s an NFT that is stored fully as a file on the blockchain, which you can render directly on the chain.” Alternatively, even if the art might not be served from the chain directly, it is possible to reconstruct the art with the elements that are stored on the chain. At a minimum [an on-chain NFT] is the kind of NFT whose art elements are stored on the same blockchain as the NFT.

‍WMP: I know you put a big emphasis in your introduction to the JPG On-Chain Canon about what you call, “on-chain edge cases.” Tokenfox in the JPG Discord mentioned how some collections are on-chain, but because of HTML constraints, they have thumbnails that are served from IPFS, which wouldn’t satisfy some purists. What are the edge cases that you find most interesting? 

‍C: There are cases where collections store their metadata on the blockchain but the art or code that generates the art is not on the chain. Personally, I wouldn’t consider that “on-chain” because you can’t actually reconstruct the art from the elements that are inside the chain. 

Having the metadata on-chain is a good step forward, but I wouldn’t consider such projects necessarily part of the On-Chain Canon because you can’t reconstruct the art entirely from the chain.

Then there are cases where an NFT contract doesn’t have any art on-chain, but then someone else puts it on-chain via another contract. The most popular example here is CryptoPunks (2017), but there are other examples like CrypToadz. The [issue] there is that there is no link between the two contracts, which makes it hard to say whether it’s really on-chain or not. I’m in favor of being inclusive, so I usually prefer to say “yes, they are on-chain.” But people can have different opinions. That is why we have this ongoing conversation, including on Discord. 

Perhaps the most popular edge case is Art Blocks, which is one of the platforms that has popularized the concept of on-chain generative art. Indeed, if it wasn’t for Art Blocks, we wouldn’t have this space today. Art Blocks is also interesting because it stores the JavaScript that creates the artworks within the chain. The issue is that sometimes this JavaScript includes references to some off-chain libraries, such as p5.js. There are some arguments for and against here, and a lot of very deep and technical edge cases.

‍Chris King: For “on-chain” projects, there are often additional pieces of information that are only stored off-chain and are not reproducible in any way from the on-chain data. These are things like traits and rarities and other properties associated with the NFT, but sometimes they are even more meaningful, such as the entire set of metadata, including name and description or some of the media content which is not capable of being generated from on-chain code or JavaScript libraries on which the on-chain code relies. 

With almost every “on-chain” project, there are actually pieces of information that are not reproducible, and in many cases these pieces of data are stored on the marketplace or project website, not on IPFS (InterPlanetary File System). 

That explains why many people who look at their ClubNFT dashboards are like, “I thought this project was on-chain, but it says that these pieces are actually off-chain and can’t be backed up, what’s going on?”

‍Trent Elmore: I find these edge cases to be where the fun is in the conversation, because we’re dancing around the question of what and where the art actually is. Chainleft said that if the art is on-chain, it can never be lost, and we had another ClubNFT team member in the JPG Discord this week advocating that. IPFS might actually be a better storage network in the long term. We don’t know what’s going to happen to Ethereum, and maybe we can actually keep the art alive longer on IPFS than on Ethereum or another blockchain. 

‍CK: For Ethereum, Tezos, or another blockchain to work, there has to be some level of popularity. The security of these blockchains is protected by the participation of large numbers of people. 

I don’t know if Ethereum is going to be around in ten years. But if I really care about proving that a certain file is a certain piece of art that was created by a certain artist at a certain time, I do trust math. And that’s all that IPFS requires. 

IPFS isn’t a storage network and it doesn’t actually help you to store data. Instead, it uses mathematics to identify and find a piece of data, whether it’s a JPEG or a movie. If I have a picture that I love and I know its IPFS identifier, then I know that the identifier will only ever point to that file. I also know that if I care about that file, I could store it in 100 different places around the world on 100 different hard drives and on Google Cloud, Amazon, or elsewhere. Either way, I know that I’ll be able to prove a million years from now that the file I’ve been protecting is the one that was created by the artist. On that basis, I do regard IPFS as a more permanent way of identifying and signing a piece of art. 

I also think that, conceptually speaking, there’s a huge amount of value to art being strictly on-chain because it imposes some serious creative constraints. The blockchain is expensive to store large files on, which creates constraints that one needs to minimize in order to get the art you want to produce to look the way you want it to look. There’s a whole class of art that deserves to be evaluated separately from off-chain art because of those constraints. 

‍WMP: Are there any threats to long-term on-chain storage? I’m not very well versed in EIP-4444 but, as I understand it, some sort of data pruning might need to happen in the future on Ethereum. Is there a risk there for on-chain NFTs? 

‍CK: EIP-4444 is really interesting. There are many ways that you could theoretically store data on the blockchain. But because blockchains weren’t originally designed for storage, they’re not really charging accurately. When you exploit a system like that, you can expect that avenue to be closed off eventually because it doesn’t make economic sense for node operators to be storing all of that data. It also bogs down the network. 

I assume that Ethereum is going to be around for a long time. But, with IPFS, the data will be around as long as you keep it around; it’s entirely within your control. People store data and servers store data — IPFS is just a way of finding it and identifying it. If I had something super valuable, or if I had what I thought was an important piece of art that ought to be preserved for many years, I might trust myself to protect that file more than I would necessarily trust a particular blockchain. 

TE: One of the things that strikes me about this conversation is that we’ve really been divorcing the token from the art — treating the art as a media file that is either on-chain or off-chain. But it seems to me that you can’t really divorce the two, or at least it’s very murky. 

If you’ve got an image on IPFS and you’ve got the token, does that token represent a receipt or is it part of a larger networked experience? One of the things that got me into NFTs was an old Pak drop called The Title (2021) which involved many different tokens that all sold for different prices but which nevertheless linked to a single IPFS hash. The idea that the art is actually embedded in the token clarifies the value proposition. They’re one and the same, there is no real difference. The token is the art. 

‍CK: I also don’t consider IPFS to be a network. There is no network that I need to trust. There is a piece of software that I could download and run on my computer that executes some math in order to determine a hash. Let’s say you mint an NFT, and it uses IPFS to “store the files,” what you’re actually storing on the blockchain is the hash. The cool thing about IPFS is that there is a two-way link between the on-chain and the off-chain. Somewhere there is a file whose hash comes to this magic number. And then, on the blockchain, there’s a token that says: “I am the token for that magic number.” Neither side can be forged. 

IPFS is the perfect match for the blockchain — especially the NFT as a use case — because you don’t need to trust anybody else. You don’t need to trust the network to exist, you don’t need to trust Pinata or any other service. As long as you have a file, it will continue to hash to the thing you own on the blockchain. That makes it a powerful tool.