I absolutely have a selective mempool, I monitor it. To be clear, we are talking only about unconfirmed transactions. Nothing can be done about transaction validation without a standards change.
As for unlimited mempool (unconf txns) what do I care? It's your memory and storage.
The valid blocks is not the debate currently. That is the red herring being bandied about. This is about control options of your own mempool rules within your own machine.
It isn't a red herring, they are creating a misdirection so we don't even consider actual block size in the argument.
At this point I'm realizing that the end result of all of this is to effectively remove the op_return field altogether. I can start buying 1g op_returns tomorrow now that I know I was lied to about the limits. How many do I need to push before the entire network is pruned? If 100% of nodes are pruned who cares what is in mempool? If 100% of nodes are pruned who will buy op_returns?
If 100% of nodes are pruned we have destroyed the history of satoshis op_return in the Genesis block.
