Peace,
On Wed, 21 Sept 2022, 12:42 am Khaled Omar, <eng.khaled.omar@xxxxxxxxxxx> wrote:
There is no dependence on end users any more, they don’t have to configure a new protocol on end hosts, network and security devices only should have IPv4/IPv6 routing enabled.
The end hosts are entirely not an issue anymore. They were — in the initial period of IPv6 deployment — but we're far from there already.
A today's end hosts — mobile phones, laptops, Linux and Windows servers — have updates turned on by default. Basically, assuming agreement across vendors, you could apply a change to the end host behaviour globally, and in couple years' time, like, 70-80% endpoints would be upgraded.
Moreover, most of the endpoints were produced in last 10 years and supported IPv6 straight from the factory.
Endpoints are not an issue. Network and security devices, on the contrary, are, as well as a myriad of stealthy middleboxes and a myriad of ISPs reluctant to change. Even comparatively simple and completely valuable features like extended BGP communities take decades for vendors to implement and then more decades for ISPs to apply in their existing hardware.
This is understood very well in the IETF community. ISP and middlebox vendor ignorance was the sole reason QUIC runs over UDP and not over it's own IP protocol number, uselessly wasting a few octets along the way. There just was no way around the issue that ISPs and middleboxes would never be properly updated to address the fact that there's now a fourth "proper" IP proto number along with 1, 6 and 17.
Hence, it is not worth discussing a network layer protocol which requires a change in network and security devices. That would just never happen.
--
Töma
