Open eberle1080 opened 7 years ago
id say including the network stack (internet included) would be awesome. between googles endpoint and what is returned is going to be impossible to include. only the general stuff can be said like how http works.
Not impossible, but you'd have to be a bit more hypothetical than usual. You could easily make up a few fake companies (MegaISP, Transit Provider Co, Peers R US, etc) and as long as you include a nice diagram with an explicit note that these are imaginary entities, I think you could get by. Obviously you wouldn't want to dive in any further than, say, the AS level (no need to talk about every single router along the way). Nonetheless this would enable you to talk about concepts like "hot potato routing" vs "cold potato routing" while still remaining agnostic to any specific AS or network topology.
I think that too much speculation misses the point. That you have a working network stack and internet connection is already a big enough speculation :innocent:
Quite the opposite actually. Too much specificity runs afoul of disclosing confidential information. Personally I know a LOT about the border network of a few major giant corps, but if I got specific I'd have a few very motivated lawyers chasing after me (with a pretty good case). So I have to make up imaginary entities and omit certain critical details to protect my ass.
. That you have a working network stack and internet connection is already a big enough speculation
I strongly disagree. There is no "internet" just as there is no "cloud", it's all just other people's data centers and networking equipment. You haven't reduced speculation, you've just neatly packed it inside of the word "internet" -- an abstraction which I'm trying to crack open without getting the corporations too upset. Hope that makes sense.
This article is a good start, but it's very lacking from the perspective of (a) networking (routing, BGP, ISPs, last mile technologies, ECMP, fiber optics) and (b) backend infrastructure (transit centers, data centers, clusters, racks, TOR, servers). It focuses a lot on the client's perspective, but it glosses over what happens when the packet leaves the client's machine. And since we're talking specifically about google.com, we can definitely make certain assumptions about the complexity / infrastructure on the server side.