Starting a new service and considering #ipv6? It's a good practice to go IPv6 for a number of reasons:1) IPv4 is exhausted, 2) IPv6 offers much better privacy thanks to the client address rotation, 3) IPv6 greatly simplifies P2P, 4) slightly better performance & latency
Ok, this makes perfect sense - looks like privacy extensions offer privacy protection equivalent to an ISP-scale NAT.
> by using throwaway addresses for each connection
This is technically possible. I've seen some services using this technique to bypass Google search query limits by switching their egress IPv6 address every 5 minutes or so. Of course this works only as long as Google doesn't enforce limits per subnet but switching is easy. Choosing a new IPv6 address *per connection* is also technically possible with /etc/gai.conf, the client software would just need to actually do it.
Regarding stable IPv6 addresses - they are pain to configure as you have to use DHCPv6 and DHCP simply makes little sense with SLAAC. I've spent significant amount of time configuring DHCPv6 only to be able to track traffic in my LAN as all my personal and kids devices had different IPv6 addr each time :)
At the end of the day, I'm just tracking their MAC addresses as these are stable and configured everything else to use privacy extensions.
This is not a feature of IPv6, it's how they've chosen to manage addressing, which could be applied to any addressing scheme.
A recent-ish paper showed that even /48 aggregation can still uniquely identify a single customer.
https://arxiv.org/pdf/1707.03900.pdf
If you're using your computer at work, you were probably getting more privacy from IPv4 NAT aggregation than you are from IPv6 address rotation, most enterprises will have fixed addressing to make network management easier.