I very rarely write about crypto and blockchain in general, but one of my pet peeves is people complaining about high transaction fees and how offering lower or free transactions is automatically a benefit. Not true at all.
Blockchains work by storing data in a single ledger. Blockchains achieve decentralization by having many people across the world validate data for the chain. Unfortunately, the more data is stored, the more expensive it is to run a node, the less people or entities validate transactions and the less decentralized it becomes. This is why bitcoin has a block size limit, why there was a "block size debate" in 2017. Ethereum works similarly, but has a variable gas limit chosen by miners. The gas limit is set to be conservative for a good reason - to keep decentralization and security high. Today, Ethereum could easily do much higher TPS and transaction fees would plummet as a result, but the chain will become quickly bloated to the point much fewer miners would run it, thus severely compromising on decentralization. Indeed, many researchers would argue that Ethereum's gas limit is already too high and is slowly but surely centralizing mining.
Some would argue that mining pools mean it's already centralized, but that's not quite true. Miners within mining pools vote on various changes, and can easily switch to a competing mining pool. If you're a DPoS fanboy, we'll have to agree to disagree.
The other reason transaction fees are high because Ethereum and Bitcoin provides the confidence, security, and value that a competitor like EOS simply doesn't. The markets are smart, and they have spoken that decentralization and security are critical, and they will pay higher transaction fees for it.
So, it's clearly been established that just blowing out block limits and bloating your blockchain is clearly not the solution. So how can we achieve scalability? Through new and efficient methods of organizing data.
The greatest innovation in this regard is rollups. You only need to store a proof of the transaction on the blockchain, rather than all data about the transaction in entirely. Think of this essentially as "compression" for blockchains. Today, on Ethereum, we have zk (zero knowledge) Rollups already active. Loopring is the first DEX to use this technology, offering 3,000 TPS on Eth1 without bloating up the chain. The fees for trades are dirt cheap. The crucial factor here is all of this is done while being backed by the very same security of Ethereum. Another rollup technology is Optimistic Rollup, which will be implemented by Synthetix later this year. Of course, there are more layer 2 solutions available, here's a great post summarizing the various solutions. We have seen Reddit's scaling bake-off, with a plethora of solutions available for scaling Ethereum today. Bitcoin has its own layer 2 solutions in the works, such as Lightning, but I'd like to see the bitcoin blockchain adopt some of these innovations.
But, of course, we also need to scale the base layer itself. Ethereum 2.0 brings sharding to the fore, essentially splitting up the blockchain into multiple parts. For example (because most new protocols are implementing sharding), at first, we'll see 64 shards on Ethereum 2.0, which will automatically expand its capacity. Again, the crucial thing to note here is that validators will only need to run a node with data from their shard, so this does not compromise on security. Ethereum 2.0's Phase 0 testnet, Medalla, is running with 25,000 validators today. It's not hard to imagine when Ethereum 2.0 goes live on mainnet, there'll be hundreds of thousands of validators. This is what true decentralization looks like, not 20 people and a few also-rans. Of course, as the demand increases, more shards can be added to the network.
Finally, in research phase is the concept of stateless clients - where validators need not run the entire blockchain to validate transactions. This would be a significant breakthrough, but we aren't quite there yet.
Putting it all together - high transaction fees are painful, but they are necessary given today's available technology. There's a reason why bitcoin and ethereum process thousands of times more value than all other blockchains combined. As we get next-generation technologies, we could see lower fees but without compromising on decentralization and security.