If I understand it correctly, this DANE-without-root provides a TOFU model as an alternative to the normal case of verifying up all the parts of the domain levels, right?

Indeed. (This part is not in any RFC or DANE standard, it is my idea to improve the system further). If you want to check the current implementation, you can search for ‘tofu’ string in this file: https://git.sr.ht/~yukikoo/dane_without_root/tree/master/item/src/core/dane.rs

and key rollover

This is a case that I didn’t cover, good idea. Probably something like ‘if successfully validated with tofu, then add all the valid dnskey to the valid tofu key list and remove the old ones’. It would still be problematic if a user doesn’t connect for a very long time to a website with tofu support and miss the key rollover, but that still a good improvement over the current tofu implementation. Will do that later.

maybe that could be nice to extend with other methods for OOB verification

Something more specific in mind?

Yes, this part is intentional to raise questions and remark :)

The certificate on my website is not valid using the WebPKI standard, but is valid using the DANE standard. It is related to my comment for this project https://sr.ht/~yukikoo/dane_without_root/ .

My issue with the WebPKI model is that any government or big company on the planet could do a MITM on your connection, generate a certificate valid for any website, and get a read/write access to all your webpki TLS communications. The DANE model is an improvement over webpki because instead of the "anyone (every ca / intermediate certificate) can generate a certificate valid for anyone" model, it bring a hierarchical trust structure.