How is IPNS over pubsub faster than DHT?

Documentation says that IPNS over DHT can be slow, which I agree.

It has to do dht.get() (run a lookup) over a quorum and find out the record with highest sequence number. and maybe needs to do a dht.findProviders(cid)

But, I think pub-sub is no different. It has to do “topic discovery” anyway and connect peers to its mesh, then maybe a libp2p fetch!

isn’t topic discovery a lookup (dht.findProviders) with some prefix topic+foo/ ? (run a lookup again)?

Is my understanding correct?

@HarveySpecter : my guess is you’ll get more engagement on this topic in https://discuss.ipfs.io/