On Fri, Sep 14, 2018 at 8:15 AM Ard Biesheuvel <ard.biesheuvel@xxxxxxxxxx> wrote: > OK, so given random.c's future dependency on Zinc (for ChaCha20), and > the fact that Zinc is one monolithic piece of code, all versions of > all algorithms will always be statically linked into the kernel > proper. I'm not sure that is acceptable. v4 already addresses that issue, actually. I'll post it shortly. > BTW you haven't answered my question yet about what happens when the > WireGuard protocol version changes: will we need a flag day and switch > all deployments over at the same time? No, that won't be necessary, necessarily. Peers are individually versioned and the protocol is fairly flexible in this regard.
