If it's goal of that PR, then it failed miserably. IIRC arbitrary data on Taproot witness data only have 1/4 weight size compared with arbitrary data on OP_RETURN. It have implication spammer could pay lower fees by misusing Taproot and store bigger arbitrary data in a TX.
This is correct but the PR is also not worsening this situation here. The idea seems to be at least to put OP_RETURN on better terms related to the Taproot-based methods and fake-Pubkey methods (which have no witness discount but less data limitations). OP_RETURN with this PR would become cheaper than Stampchain-type methods.
Ideally it could be combined with some Taproot restriction, I however suspect this could have implication for other, more important Taproot projects. Currently Ordinals doesn't seem to be a problem, the fad has waned a lot, so a careful proceeding with such restrictions is a good idea I think.
Stamps are explicitly designed to be unprunable, despite cheaper, prunable approaches already existing.
And that's why I consider them an especially unfriendly method (for nodes and decentralization of the network in general). It's not possible to stop them but at least it should be prevented at all cost that this becomes the standard method to store data and NFTs on-chain. Here this PR comes into play.
Isn't that the point of a permissionless and decentralized network, that arbitrary decision makers can't decide what is illegal and remove it or prevent it from being embedded into it?
Here I must agree with the critics of Ordinals and similar stuff: The purpose of Bitcoin is not storing non-financial data. Developer decisions affecting this "functionality" thus aren't censorship.
As embedding data in fake public keys can't be prevented because you are free to use whatever public key you want, you then have to reccur to incentives to prevent at least this kind of method. Because if embedding things with unpruneable methods is viable and relatively cheap then people will use them more often even if they don't need the "unpruneable" aspect, with the negative consequences I already mentioned (cluttering UTXO set, unpruneable). If OP_RETURN, which is more node-friendly, becomes cheaper (at least without limits it would be cheaper than Stampchain) then at least some people should consider using it instead of the node-unfriendly methods.
@alani123: No, bigger blocks would lead to more centralization as it would be more difficult to run a full node. It would encourage spam even more, see gmaxwell's post.