On Wed, Jul 25, 2018 at 3:34 PM, Christopher Wood < [email protected]> wrote:
> On Wed, Jul 25, 2018 at 8:05 AM Tommy Pauly <[email protected]> wrote: > > > > Yes, I'm inclined to agree—I think that having a unique address per > connection shouldn't be the default; an application could certainly always > set it if it wants to decrease linkability? > > Applications should not have to opt-in to better privacy. Decreased > linkability should be the default. So if doing this per-connection is > shown (not just believed) to be infeasible, then per-application is a > reasonable compromise. > The problem is that it might not be a dichotomy between "feasible" and "infeasible". What if the tradeoff is increased connection latency from acquiring a new address (either by DHCP or SLAAC/DAD)? What if it increases network load from neighbor discovery so much that the LAN visibly slows down? I would argue that the default should be BCP for the network architecture in question, whatever that is, and leave the determination of BCP to those SMEs. Kyle
_______________________________________________ Taps mailing list [email protected] https://www.ietf.org/mailman/listinfo/taps
