Discuss impact of PQClean wind-down

[dstebila]

As per PQClean/PQClean#604 (comment), the PQClean project will be winding down over the next 6 months. We rely on PQClean's implementations for:

• Classic McEliece
• HQC: soon to be replaced with another implementation Hqc update reference implementation  #2328
• Kyber: only for aarch64
• Falcon: eventually to be replaced with FN-DSA
• SPHINCS+: soon to be removed Remove SPHINCS+ #2339

[baentsch]

Thanks for the summary - so it seems OQS may want to check with users whether keeping McEliece is desired. Keeping an architecture-specific Kyber doesn't look necessary as long as a reference implementation is still around -- if there is a need to keep Kyber (is there such a demand?). And the rest will be gone soon.

Concrete proposal: Announce upcoming removal of McEliece and Kyber support in the next liboqs release to see whether anyone complains.

[dstebila]

Thanks for the summary - so it seems OQS may want to check with users whether keeping McEliece is desired. Keeping an architecture-specific Kyber doesn't look necessary as long as a reference implementation is still around -- if there is a need to keep Kyber (is there such a demand?). And the rest will be gone soon.

Concrete proposal: Announce upcoming removal of McEliece and Kyber support in the next liboqs release to see whether anyone complains.

I've revived the discussion in #1989 about Kyber support to see where things stand on usage of Kyber.

Regarding Classic McEliece, as far as I know it is still on track for ISO standardization, and so would be in scope for OQS to continue to include.

[baentsch]

Regarding Classic McEliece, as far as I know it is still on track for ISO standardization, and so would be in scope for OQS to continue to include.

In that case, does that mean that OQS will effectively (have to) become maintainer of this code base (resolve security issues, follow required changes during the standardization process, etc.) when the upstream's gone? Allow me to highlight again #2064 to capture such dependencies/obligations on OQS and/or upstreams (where available).

[thomwiggers]

It's probably possible to vendor https://lib.mceliece.org/, but you'd need to do some work to adapt it to the OQS build system and so on.

[Guiliano99]

Hello,

I’m interested in the McEliece algorithm, as I’m currently using it for the hybrid KEM Chempat. If supporting McEliece wouldn’t mean additional work (or only minimal effort), I would appreciate it if it could be included.

[RodriM11]

It would require additional work, but as @dstebila noted, Classic McEliece is on track for ISO standardization so IMO it should continue to be supported.

[Guiliano99]

Thanks.