Yes, storing data, but seems one of the biggest issues is what the data is.
Discussion
having consensus code that discriminates different types of data would be a consensus nightmare (chaotic hardfork generator)
So consensus code that limits the size of the data would be better?
yes of course, that would prevent large op_returns, but it would push people into storing large amounts of data in worse ways, like multisig outputs, witness data, etc.
I like eriks analogy here:
“Look, no one wants garbage on the timechain, but people are gonna litter.
So, we can either make garbage cans for garbage (i.e. op_return) and people can put their garbage there, or people will put their garbage on the streets (UTXOs) b/c there are no garbage cans.”