<snip> > > In those years sufficient data has been collected to conclude that > > /64 is not the "be all and end all". > > Alternatively, it could show that people are applying IPv4 practices > they're familiar with to IPv6, even when it isn't necessary. They're > missing out on operational savings from simplicity. >Have you read rfc1925 section 2.4? My operational experience is why I value simplicity, consistency and less variation, as I've worked on and troubleshooted networks that have been unnecessarily complex. I think RFC1925 2.12 is more important. > You are grossly overstating these 'operational savings'. I've seen 4x/24s on a router interface because renumbering hosts into a single /22 was operationally too expensive. Those hosts also were being left with default "class B" subnet masks because that was allowing their broadcast based discovery to work. Fortunately their router was performing proxy ARP which then allowed them to access other off-link destinations within the Class B address space. I've seen a router with 25 subnets on its interface because they were collapsed down from many PCs acting as routers, and renumbering the hosts into a single large enough prefix was too operationally expensive. That also meant the router was sending two RIP packets for each update because the limit is 25 routes in a RIP packet. If you see those sorts of things, and have experience with other protocols like IPX where subnets are always large enough for any number of devices the layer 2 technology can support, you value the simplicity of /64s everywhere, even on links with two addresses. >Also, for the >sake of this dialogue I'll apply Hanlon's razor and operate under the >assumption that you are not familiar with my work environment. >I am amazed your reaction to my sharing of real-world data is the >equivalent of "well, i think you are doing it wrong". >A lack of empathy >is exhibited in that you do not wonder _why_ things are what they are. I know quite well why IPv4 has been operated the way it has. Your data doesn't measure motivation. If I see a set of prefix lengths in IPv6 that vary like IPv4 ones would, which is what your data set showed, then I think the motivation is that IPv4 addressing methods and thinking have been applied to IPv6. If your mix of prefix lengths was only /64s, or only /64s and /127s, then I think that shows an understanding of IPv6 addressing models and intents. A lot of IPv4's addressing models and methods are a consequence of IPv4 never being designed to be used to build the Internet we have today. Here's what Vint Cerf has said about IPv4's address size and design context. http://mailman.nanog.org/pipermail/nanog/2010-April/020488.html IPv4 was designed for a research network that escaped and became a global network. IPv6 is really the first protocol designed by the IETF to solve a global internetwork problem. Let's leave behind unnecessary practices that have been used to extend IPv4's life, and that make things unnecessarily complicated and more costly to operate and troubleshoot. Regards, Mark.