Add NIST names as further alias for ML-DSA

[falko-strenzke]

Currently in src/lib/pubkey/dilithium/dilithium_common/dilithium.cpp ML-DSA is mapped only to the algorithm strings with the pattern ML-DSA-4x4. This should be amended by the official NIST naming pattern ML-DSA-44.

@reneme @randombit

Any thoughts about this? I would make the PR then.

[falko-strenzke]

In my view, the to_string() method should also return the NIST names, but as that will be a breaking change, I am not sure about this.

[reneme]

With "official naming pattern" I figure you refer to the names that are used in the FIPS 204 standard document -- particularly in the "Parameter Sets" in Section 4. Indeed, for SLH-DSA and ML-KEM we did stick to those names byte-by-byte and the discrepancy in ML-DSA is unfortunate.

Other places in the public API this shows up are:

• the enum entries DilithiumMode::ML_DSA_4x4
• the OIDs (oids.txt)

For the mapping from string to "ML-DSA mode" we could be ambiguous and just support both, but for the reverse mapping and in particular for the OIDs we can't. If we decide to adopt the NIST nomenclature, I think, we should go all-in and drop support for our current nomenclature (with the "x") entirely. That can only happen with Botan4, and we could re-consider it then.

Just adding partial support now makes behavior inconsistent across the public API surface for people that use the NIST nomenclature and personally I find that worse than requiring people to map the strings in their application code.

[falko-strenzke]

@reneme
So should we define Botan 4 as Milestone and file this for consideration there? Or how should we proceed else?

[reneme]

Or how should we proceed else?

I made this a sub-issue of #4666 which is the generic tracking issue for all things Botan4.