On Mon, Mar 16, 2026 at 7:20 PM Viktor Dukhovni <[email protected]>
wrote:

> On Mon, Mar 16, 2026 at 03:41:59PM +0000, Salz, Rich wrote:
>
> > I also could see folks trying to avoid the HRR altogether and rip the
> > X25519 out of the hybrid key share  and use immediately.  That's not a
> > "reuse", I suppose, but still seems a bad idea.
> >
> > Can you say why?  My inclination would be to codify it and say that
> > any hybrid keyshare could be used for its constituent parts unless the
> > definition of the hybrid says otherwise.
>
> The server decomposing a hybrid is of course not viable, the actual
> scenario that makes sense (and MUST NOT be considered reuse) is for
> the client to use the same public key in multiple parallel keyshares:
>
> An extreme example of this could be:
>
>     keyshare 1: X25519MLKEM768: <X25519 pk> + <MLKEM 768 pk>
>     keyshare 2: SecP256r1MLKEM768: <P-256 pk> + <same MLKEM 768 pk>
>     keyshare 3: X25519: <same X25519 pk>
>     keyshare 4: P-256: <same P-256 pk>
>
> This sends 4 keyshares based on just 3, rather than 6 underlying
> generated keys.  There's no actual "reuse" here, each key is used at
> most once for just whichever keyshare the server selects.


This text would have no impact on this practice, as it says for
"multiple connections", which this is not.

-Ekr
_______________________________________________
TLS mailing list -- [email protected]
To unsubscribe send an email to [email protected]

Reply via email to