You're both right from where I'm looking.

I've been saying that DNS is the perfect mechanism to assert NIP-05 identifiers (and Lud-16 lnurls, and even BTC addresses) for quite some time.

In the case of NIP-05/LUD-16 The .well-known/URI scheme is limiting - and I would also argue, less secure than, say, a DNSSEC signed zone.

While true that DNS is "centralized" ultimately at the root level, there is no content or participant level censorship there - the DNS tree is somewhat like a federated namespace AND there will in time be more decentralized namespaces outside of the IANA tree within which one could pin their NIP-05 or LUD-16 ids.

None of these arguments have convinced the people it needs to (*cough* nostr:npub180cvv07tjdrrgpa0j7j7tmnyl2yr6yr7l8j4s3evf6u64th6gkwsyjh6w6 *cough*) - and to his counter-points, I do somewhat get it).

It is possible that somebody, someday could propose another NIP entirely for the assertion of NIP-05-like id's (they would look exactly the same) via DNS TXT recs.

It is not obvious or probable that such a NIP proposal would succeed.

At the risk of sounding repetitive, I'll refer to my previous article on this sort of thing here (and there are a few threads on Github somewhere):