I am interested to find out more about ethereum scaling solutions. So far I've read about Raiden Network, Plasma / Plasma Cash, Striim (officially launched this week), Sharding and a bit about Casper.
With Raiden I am not sure why initially it was in the official Ethereum roadmap as "the lightning network of ethereum" or something like that, and now is a separate project. Nobody said anything about that but I think it's an important detail. Anyone knows? Looks like devs are still active on it.
About Plasma the issue with the exponential scalability resides in users having to download/authenticate each plasma block. So with Plasma Cash there would be an extra plasma coin with unique id created for every user every time they, say, deposit on an exchange and whatnot thus compared to Plasma have the only requisite to pay attention to the blocks that contain coins they want to track.
Then there's Striim by Hubii which is gonna bring an ethereum scaling solution and protocol but not too much info on it yet, other than the launch will be in a day or two and the striim tokens will be gradually airdropped to hbt token holders. Positive is their partnership with telenor which is gonna jumpstart the use of it for micropayments.
Differently from off-chain solutions like Striim and Raiden, Sharding is an on-chain scaling solution, i.e. it maintains all transactions on the original blockchain. Sharding basically allows for operations to run not in a linear way but simultaneously which results in higher TPS. In short with sharding the ethereum network is divided in multpile nodes (aka shards), which process all the transactions that occur within that group.
To me it seems sharding will be the most impactful but also will take the longest to implement; but being onchain it is also intrinsecally different from raiden and other offchain solutions which are easier to "plug" in the network do absorb dust transactions right out of the box.
Will offchain solutions continue to exist and be utilised even once solid onchain mechanisms are implemented?
I believe so. Scalability/security/decentralization is not only a trilemma, but really a cat-and-mouse game where you also have to compete against other aspiring networks and protocols which try to boost higher and higher tps (some of which aren't even a blockchain.. cough hashgraph cough..).
I agree with you, namely with your statement
"To me it seems like sharding will be the most impactful, but also on the inside, and on the other hand, it's also intrinsically different from raiden and other offchain solutions that are easier to plug in the network. of the box. "
It is these actions that should be carried out now.