because it only works for nodes running behind Tor/I2P, which means that it improves obfuscation only to that percentage. Definitely better than nothing, but not complete heuristic breakage which Dandelion++ etc aim to achieve. I like that your autocorrect turns supernodes into supermodels though.
Discussion
That's not exactly true though, your node and the peer you anonymously connect to can both be clearnet for Tor. You just need a tor proxy available to your node and it can route it through an exit node.
Similar for I2P, except the peer must be behind I2P as well.
It breaks all heuristics if used today for any transactions sent with it. For dandelion, you need many peers running it to hide in an anonymity set.
Right but propagating exclusively via Tor/I2P even via proxy is not desirable for everyone due to reliability, but again its def an improvement. I don't think an anonymity set assumption makes sense for Dandelion++ as we're not hiding in a crowd, it's rather the idea of a sneakernet: since P2P traffic is encrypted now it's more the idea of passing messages out of band. You could likely observe who is running a stempool, but shouldn't be able to see much more beyond that
I don't think any new onion routing network bolted on to the Bitcoin network will be able to be more reliable than Tor or I2P with decades of research and development. Not to mention that Tor and I2P exist today and dandelion is vaporware that will never realistically be built.
I don't think it makes sense to compare the two. Tor and Dandelion serve completely different purposes. Let's agree to disagree.