notes
this was a dense and often redundant read with a lot of the classic, academic need to spend a lot of time on (re)naming things.
to summarize … tunnels are the new internet. you might not necessarily realize/recognize what’s going on, but the focus of today’s internet is no longer end-to-end connectivity and addressing, etc. it’s what you can accomplish with tunnels.
this is true, pretty much everything that you’re going to do in internet and enterprise engineering for the remainders of our life time is going to be in the service of some form of tunneling. we can spice this up all we want with things like SRv6 but this is just tunneling and operations at different layers or interface points in the network.
tunnels are here. make your peace with them.
it’s only taken a little over 10 years for someone to write this down and put a new nomenclature around it.
this is focused on “architecture”, but i wish it would have provided a bit more solid discussion around the need for control plane validation and correctness. there’s precious little discussion given to the challenges we’ve had over the past 20+ years in terms of routing security and the contortions we’re currently engaged in with things like RPKI and its ilk in order to retrofit something resembling address announcement validation and other control plane hardening that’s required.
this is, i suspect, less engaging than talking about spiffy new architectures using tunnels and making metaphors about subduction, etc. which isn’t to say this isn’t a useful read, just that it kind of scratches at the service itches as emergent elements that need taxonomizin’ and describin'.