L2 allows for faster and cheaper transactions. Not by a much, though. Since developers are thinking of adding even more layers.
Well, I think "not by much" is an understatement. The capacity increase is quite significant, both in "state channel" L2s and in "sidechain" L2s. "Third layers" are normally not really needed, only if we had to scale to billions of everyday users, or for interoperability solutions like certain swap and DeFi technologies.
Both have however their own bottlenecks:
- State channels like Lightning need to be initialized on-chain by every new user. This means that the block capacity limits the number of new users who can onboard. People who transact many times are those most benefitted by LN and similar solutions.
- Sidechains don't have this problem, new users can also be onboarded off chain, i.e. on the sidechain. The problem here is however that the sidechain itself will have a limited capacity. But the idea of sidechains is anyway to have multiple of them. So even if a single sidechain only increases capacity roughly from factor 2 to factor 100 for a big-block sidechain, if all the ~70 already existing sidechain projects would work as promised, a capacity increase of 5000 times the original (i.e. 30k+ tps) would not be a problem.
- Statechains and Ark are not chain-based and don't have the onboarding problem of Lightning, but they need a trusted or semi-trusted party. In Ark, afaik the idea is that the trusted party can't steal the funds so it may be one of the most interesting concepts.
Conversely it is also worth noting that while L2 is dependent on L1, L1 is not affected by anything that is happening on L2.
While this is normally true, there were some discussions some years ago in the developer community about risks for L1 coming from extreme scenarios, e.g. malfunctioning or attacked L2s. A popular sidechain for example, which was successfully attacked, could threaten the stability of the Bitcoin price.
There was also the fear that (merged-mined) sidechains could be a "covered" block size increase, and this could have also an effect on the L1: it could make the L1 as centralized as a big block coin. The logic of this argument is approximately: Miners will need to merge-mine all relevant sidechains because they otherwise aren't competitive anymore, so they will need to process tons of transactions. However, this does only affect miners, full nodes who want to only validate, would only need to process transactions and blocks of the sidechains the user who operates the node considers relevant. And normally an user should only need to use a single L2, plus the L1 for increased security.
I'd add an interesting advantage specifically for the "sidechain"-type L2 most people ignore: Almost any blockchain can work as a L2 for other blockchains, like Bitcoin. They only need a contract language expressive enough to handle the two-way peg. For example, Ethereum works as a Bitcoin L2 via the tBTC project which is relatively decentralized in comparison to most other solutions. In theory even Litecoin could work as a Bitcoin L2, but it would need an OP_RETURN-based additional protocol for custodian slashing.
An "already established" blockchain thus can constitute a quite secure sidechain or L2.
