I was always wondering why statistics never correctly showed information for my billing cycle. Admittedly, I wasn't the account holder until recently, but I never understood that it was a carrier settings. Are there other carrier-specific features that you know about?
By the way, your parent comment hypothesizing about http/3 and Apple services likely being routed via Private Relay was also very insightful. The http/3 part seems quite speculative, but I could easily see the Apple services being routed over Private Relay as part of any standard "dogfooding" roadmap.
There is not a lot of difference between http3 and a VPN protocol that can bundle and transport multiple bidirectional channels asynchronously, especially when considering the restriction “only uses http protocols and none other” — a quality that many Apple services share with Private Relay.
It’s absolutely speculative to consider http3 a viable replacement for ”a VPN”, as many consider Private Relay to be^, but that’s not due to any technical limitation. We could benchmark Squid-over-http3 versus Wireguard VPN in various latency and packet loss scenarios today (if implemented), and get productive and interesting results. We could do the same with varnish / nginx versus openvpn / pptpd. Whether or not it supports CONNECT is worth noting, but is no obstacle for “let’s assume that everything is HTTP”, as appears widely applicable to both Apple’s services to their customers, and a significant majority of consumer VPN traffic, today.
Have those benchmarks been performed yet? Are we all doing it wrong by using wireguard et al. to Mullvad et al. for 99% HTTP traffic and we just don’t realize it yet? I look forward someday to finding out the answer :)
ps. iOS carrier APIs and functionality are NDA’d so I don’t have any other information or useful examples to offer, as my experience with carriers as their customer is very limited. Perhaps others will know; apologies.
^ I haven’t seen an FP yet titled “If it can’t carry SSH traffic, it’s not a VPN”, but certainly there’ll be one someday. I can’t predict when, though, anymore than I can guess when someone else will realize to seriously ask, supported by benchmarks and encapsulation overhead charts, “Should tcp be deprecated in favor of http?”. (I have no opinion on what the correct answers are at this time.)
By the way, your parent comment hypothesizing about http/3 and Apple services likely being routed via Private Relay was also very insightful. The http/3 part seems quite speculative, but I could easily see the Apple services being routed over Private Relay as part of any standard "dogfooding" roadmap.