The roads I take...

As mentioned in a previous post, I've been working with the Capacity Blockchain Solutions team on the Crypto stamp project, the first physical postage stamp with a unique digital twin, issued by the Austrian Postal Service (Österreichische Post AG). After a successful release of Crypto stamp 1, one of our core ideas for a second edition was to represent stamp albums (or stamp collections) in the digital world as well - and not just the stamps themselves.

We set off to find existing standards on Ethereum contracts for grouping NFTs (ERC-721 and potentially ERC-1155 tokens) together and we found that there are a few possibilities (like EIP-998) but those ares getting complicated very fast. We wanted a collection (a stamp album) to actually be the owner of those NFTs or "assets" but at the same time being owned by an Ethereum account and able to be transferred (or traded) as an NFT by itself. So, for the former (being the owner of assets), it needs to be an Ethereum account (in this case, a contract) and for the latter (being owned and traded) be a single ERC-721 NFT as well. The Ethereum account should not be shared with other collections so ownership of an asset is as transparent as people and (distributed) apps expect. Also, we wanted to be able to give names to collections (via ENS) so it would be easier to work with them for normal users - and that also requires every collection to have a distinct Ethereum account address (which the before-mentioned EIP-998 is unable to do, for example). That said, to be NFTs themselves, the collections need to be "indexed" by what we could call a "registry of Collections".

To achieve all that, we came up with a system that we think could be a model for future similar project as well and would ideally form the basis of a future standard itself.

At its core, a common "Collections" ERC-721 contract acts as the "registry" for all Crypto stamp collections, every individual collection is represented as an NFT in this "registry". Additionally, every time a new NFT for a collection is created, this core contract acts a "factory" and creates a new separate contract for the collection itself, connecting this new "Collection" contract with the newly created NFT. On that new contract, we set the requested ENS name for easier addressing of the Collection.
Now this Collection contract is the account that receives ERC-721 and ERC-1155 assets, and becomes their owner. It also does some bookkeeping so it can actually be queried for assets and has functionality so the owner of the Collection's own NFT (the actual owner of the Collection itself) and full control over those assets, including functions to safely transfer those away again or even call functions on other contracts in the name of the Collection (similar to what you would see on e.g. multisig wallets).
As the owner of the Collection's NFT in the "registry" contract ("Collections") is the one that has power over all functionality of this Collection contract (and therefore the assets it owns), just transferring ownership of that NFT via a normal ERC-721 transfer can give a different person control, and therefore a single trade can move a whole collection of assets to a new owner, just like handing a full album of stamps physically to a different person.

To go into more details, you can look up the code of our Collections contract on Etherscan. You'll find that it exposes an ERC-721 name of "Crypto stamp Collections" with a symbol of "CSC" for the NFTs. The collections are registered as NFTs in this contract, so there's also an Etherscan Token Tracker for it, and as of writing this post, over 1600 collections are listed there. The contract lets anyone create new collections, and optionally hand over a "notification contract" address and data for registering an ENS name. When doing that, a new Collection contract is deployed and an NFT minted - but the contract deployment is done with a twist: As deploying a lot of full contracts with a larger set of code is costly, an EIP-1167 minimal proxy contract is deployed instead, which is able to hold all the data for the specific collection while calling all its code via proxying to - in this case - our Collection Prototype contract. This makes creating a new Collection contract as cheap as possible in terms of gas cost while still giving the user a good amount of functionality. Thankfully, Etherscan for example has knowledge of those minimal proxy contracts and even can show you their "read/write contract" UI with the actually available functionality - and additionally they know ENS names as well, so you can go to e.g. etherscan.io/address/kairo.c.cryptostamp.eth and read the code and data of my own collection contract. For connecting that Collection contract with its own NFT, the Collections (CSC) contract could have a translation table between token IDs and contract addresses, but we even went a step further and just set the token ID to the integer value of the address itself - as an Ethereum address is a 40-byte hexadecimal value, this results in large integer numbers (like 675946817706768216998960957837194760936536071597 for mine) but as Ethereum uses 256-bit values by default anyhow, it works perfectly well and no translation table between IDs and addresses is needed. We still do have explicit functions on the main Collections (CSC) contract to get from token IDs to addresses and vice versa, though, even if in our case, it can be calculated directly in both ways.
Both the proxy contract pattern and the address-to-token-ID conversion scheme are optimizations we are using but if we were to standardize collections, those would not be in the core standard but instead to be recommended implementation practices instead.

Of course, users do not need to care about those details at all - they just go to crypto.post.at, click "Collections" and create their own collection for there (when logged in via MetaMask or a similar Ethereum browser module), and they also go the the website to look at its contents (e.g. crypto.post.at/collection/kairo). Ideally, they'll also be able to view and trade them on platforms like OpenSea - but the viewing needs specific support (which probably would need standardization to at least be in good progress), and the trading only works well if the platform can deal with NFTs that can change value while they are on auction or the trade market (and then any bids made before need to be invalidated or re-confirmed in some fashion). Because the latter needs a way to detect those value changes and OpenSea doesn't have that, they had to suspend trade for collections for now after someone exploited that missing support by transferring assets out of the collection while it was on auction. That said, there are ideas on how to get this back again the right way but it will need work on both the NFT creator side (us in the specific case of collections) and platforms that support trade, like OpenSea. Most importantly, the meta data of the NFT needs to contain some kind of "fingerprint" value that changes when any property changes that influences the value, and the trading platform needs to check for that and react properly to changes of that "fingerprint" so bids are only automatically processed as long as it doesn't change.

For showing contents or calculating such a "fingerprint", there needs to be a way to find out, which assets the collection actually owns. There are three ways to do that in theory: 1) Have a list of all assets you care about, and look up if the collection address is listed as their owner, 2) look at the complete event log on the blockchain since creation of the collection and filter all NFT Transfer events for ones going to the collection address or away from it, or 3) have some way of so the collection itself can record what assets it owns and allow enumeration of that. Option 1 is working well as long as your use case only covers a small amount of different NFT contracts, as e.g. the Crypto stamp website is doing right now. Option 2 gives general results and is actually pretty feasible with the functionality existing in the Ethereum tool set, but it requires a full node and is somewhat slow.
So, for allowing general usage with decent performance, we actually implemented everything needed for option 3 in the collections contract. Any "safe transfer" of ERC-721 or ERC-1155 tokens (e.g. via a call to the safeTransferFrom() function) - which is the normal way that those are transferred between owners - does actually test if the new owner is a simple account or a contract, and if it actually is a contract, it "asks" if that contract can receive tokens via a contract function call. The collection contract does use that function call to register any such transfer into the collection and puts such received assets into a list. As for transferring away an asset, you need to make a function call on the collection contract anyhow, removing from that list can be done there. So, this list can be made available for querying and will always be accurate - as long as "safe" transfers are used. Unfortunately, ERC-721 allows "unsafe" transfers via transferFrom() even though it warns that NFTs "MAY BE PERMANENTLY LOST" when that function is used. This was probably added into the standard mostly for compatibility with CryptoKitties, which predate this standard and only supported "unsafe" transfers. To deal with that, the collections contract has a function to "sync" ownership, which is given a contract address and token ID, and it adjusts it assets list accordingly by either adding or removing it from there. Note that there is a theoretical possibility to also lose an assets without being able to track it there, that's why both directions are supported there. (Note: OpenSea has used "unsafe" transfers in their "gift" functionality at least in the past, but that hopefully has been fixed by now.)
So, when using "safe" transfers or - when "unsafe" ones are used - "syncing" afterwards, we can query the collection for its owned assets and list those in a generic way, no matter which ERC-721 or ERC-1155 assets are sent to it. As usual, any additional data and meta data of those assets can then be retrieved via their NFT contracts and their meta data URLs.

I mentioned a "notification contract" before which can be specified at creation of a collection. When adding or removing an assets from the internal list in the collection, it also calls to that notification contract (if one is set) as a notification of this asset list change. Using that feature, it was possible to award achievements directly on the blockchain for e.g. collecting a certain number of NFTs of a specific type or one of each motif of Crypto stamps. Unfortunately, this additional contract call costs even more gas on Ethereum, as does tracking and awarding of achievements themselves, so rising gas costs forced us to remove that functionality and not set a notification contract for new collections as well as offer an "optimization" feature that would remove it from collections already created with one. This removal made transaction costs for using collections more bearable again for users, though I still believe that on-chain achievements were a great idea and probably a feature that was ahead of its time. We may come back to that idea when it can be done with an acceptably small impact on transaction cost.

One thing I also mentioned before is that the owner of a Collection can actually call functions in other contracts in the name of the Collection, similar to functionality that multisig wallets provide. This is done via an externalCall() function, to which the caller needs to hand over a contract address to call and an encoded payload (which can relatively easily be generated e.g. via the web3.js library). The result is that the Collection can e.g. call the function for Crypto stamps sold via the OnChain shop to have their physical versions sent to a postage address, which is a function that only the owner of a Crypto stamp can call - as the Collection is that owner and its own owner can call this "external" function, things like this can still be achieved.

To conclude, with Crypto stamp Collections we have created a simple but feature-rich solution to bring the experience of physical stamp albums to the digital world, and we see a good possibility to use the same concept generally for collecting NFTs and enabling a whole such collection of NFTs to be transferred or traded easily as one unit. And after all, NFT collectors would probably expect a collection of NFTs or a "stamp album" to have its own NFT, right? I hope we can push this concept to larger adoption in the future!