On Wed, Jun 25, 2025, at 18:10, John Mattsson wrote: > The 3GPP requirement (as I understand them) would be to change > certificates during a long-lived mutually authenticated “association”, > where the association may or may not be a TLS connection. That requires > sending the new certificates. With ever shrinking certificate lifetimes > this becomes more important. > > - Setting up a new TLS 1.3 connection is one good way to solve this, > but might require changes to other layers in the protocol stack, which > may or may not be problematic.
At least for signaling, none of the state should be bound to the connection, so a new connection is better in almost every way. The question becomes then whether there is state that ends up being bound to the connection for long enough that this is a real issue. This is especially true if you consider the need to have redundancy. Active load sharing across multiple network paths is almost necessary for uptime reasons. Efficient fault recovery, too. With those in place, regular rotation of connections only improves your ability to rely on recovery being functional when it is needed. Rotation schedules on the order of days would still be far faster than certificates cycle out, I'd guess, even under the most aggressive lifetime shrinking. (Yes, I really am questioning the requirement.) _______________________________________________ TLS mailing list -- tls@ietf.org To unsubscribe send an email to tls-le...@ietf.org
