You're probably being downvoted since your comments appear to have nothing to do with the article or any ongoing discussion in the comments section. The use case the authors had was deduplication/change detection of large art asset files. Just how are reversible hash functions for small fixed size inputs relevant to that?
Also, to the GP whose posts are getting downvoted to the point of being flagged, in case it's helpful in the future:
(1) The idea of using a bijection as a building block of hashing is not new, and it's not new to the creators of MeowHash. You point out that there exists a function from 512 bits to 512 bits that's guaranteed to be unique.
Another such function that also has good avalanche properties and is hard to determine the original input from if you don't know the key: The AES mixing function.
MeowHash is based, in fact, on transforming individual input blocks with a single round of AES, and then merging streams of AES-mixed blocks. The AES part is straightforward -- but provides much better practical hash properties than your proposal (for example, I can use the low-order 128 or 64 bits of the output of MeowHash with the same general collision behavior). The merging step is more important from the perspective of collision-freedom, and you omit that fairly critical part.
(2) The contribution of meowhash is its _speed_ and good hashing behavior. It's a very elegant use of highly-optimized on-CPU AES mixing (a feature used by previous hash functions), with a nice block size and stream mixing design that lets it keep a lot of blocks in flight for high throughput.
(Meta) You're engaged in a pattern that HN readers generally respond poorly to: Get downvoted -> Complain about downvotes in a way that implies the downvoters are idiots -> get downvoted -> double down.
It's pretty easy to rephrase your comments in a way that doesn't have that tone of "you're all idiots for not seeing the connection here".
don't you sometimes think you'd want a platform where the users are in control of the platform rules? of course we'd all be held to the standards to which we hold others...
