On Mon, 23 Feb 2026, 7:06 pm D. J. Bernstein, <[email protected]> wrote:

> Writing off-list since the list is censored.
>

Writing on-list because that's the entire reason I'm a member.


> Jack Grigg writes:
> > Compute in the hybrid solution would predominantly be determined by
> ECDHE,
> > not ML-KEM. See e.g. [0] [1] for numbers.
>
> It's true that on big CPUs the _cycles_ for ML-KEM are lower than the
> _cycles_ for ECDHE.


I'm glad you agree that my statement was correct.

Cheers,
Jack

However, the dominant cost of ECDHE+ML-KEM by far is
> the _communication cost_ for ML-KEM. See
>
>     https://blog.cr.yp.to/20260219-obaa.html#cost
>
> for many numbers and supporting links.


> The myth that removing ECC from ECC+PQ produces an important speedup is
> the most effective part of the NSA/GCHQ anti-hybrid advertising. I
> disputed the NSA/GCHQ claims years ago in
>
>     https://blog.cr.yp.to/20240102-hybrid.html
>
> but of course they simply keep flooding the zone with misinformation.
>
> The draft-ietf-tls-mlkem motivation statement was circular for almost
> all of the draft's existence, but version -06 of draft-ietf-tls-mlkem
> a day before "last call" suddenly claimed "constrained environments
> where smaller key sizes or less computation are needed". Note the word
> "needed". The claim was then _edited_ but there was no _erratum_. The
> replacement text refers to "smaller key sizes or less computation" as a
> "use case"; this doesn't say explicitly that the savings are important
> but it still makes the reader _think_ they are, which isn't true.
>
> ---D. J. Bernstein
>
_______________________________________________
TLS mailing list -- [email protected]
To unsubscribe send an email to [email protected]

Reply via email to