Interesting, I use the 300MB default maxmempool and have never been harmed by spam. Sounds like FUD to me.

It depends on what you find harmful In this post, Chris Guida outlines why he thinks spam is particularly harmful:

X (formerly Twitter)

Chris Guida | ⚡🪢 BIP110 (@cguida6) on X

Several core supporters have pointed me to Pieter Wuille’s recent StackExchange post[0] explaining core’s rationale for raising the default opr...

What about the consequences of slower verification of blocks for the whole network of nodes, due to Knots nodes not being able to start validation because they dismissed economically relevant valid transactions? Longer validation can cause competing blocks more likely. The fee estimatiom problem is also a well-known predictable effect of purging mempools this way. Yet another serious issue is to burden all nodes with substantially more validation _for ever_ with inscriptions in pubkeys. In contrast, op_return can be ignored in validation. I think these side effects dwarf your concerns about temporary RAM benefits. In addition, the CSAM arguments have existed for long as well and, though not a lawyer and all, but Knots nodes will store these as well and relay them as valid blocks to other nodes, so there is no benefit I see here either. I am almost sure you knew these arguments already. Failing to steelman the other side is a mistake. Do you want to get closer to truth or justify an outcome? The only "moral" of bitcoin is economic incentives, strictly speaking. People will argue on subjective morals and semantics for ever. What we strive to achieve with money is as close to objectiveness as possible though. Wishful thinking with ignorance towards reality caused the fiat system in the first place. Camouflaging this with morality and a veneer of certainty instead of focusing on error-correction is just an appeal to authority and a path leading back to a fiat.

If you store "spam" in RAM until it gets mined you avoid having to redownload and cryptographically verify the "spam" tx once it gets mined into a block or it is broadcasted to you by another node. That saves both bandwidth for downloading and CPU for verifying. I would not agree that disk space is abundant. You can get by with just 300MB RAM for an unfiltered Mempool and don't have to upgrade RAM however many noderunners will soon have to upgrade from a 1TB SSD to a 2TB one which is a cost factor and also implies manual effort and downtime.