A collection of post-mortems
github.comMost of the config errors seem related to networking.
Why haven't we been able to simplify networking, reduce the impact of any errors or even increase security? Why do we still put up with 40-year old protocols and even concepts?
Anyone working on disrupting networking?
A lot of changes in networking have been happening recently; in particular, software defined networking is (slowly) replacing dedicated routers and hardware devices.
The downside of this is that it's much easier to change configuration than it was before, and so the pace of such changes is increasing. Given that most errors can be traced back to a human error at the start of the chain, he problem isn't really in the networking stack but the meat at the end of the buttons.
Realistically you need to address that end of the problem, not the networking protocols themselves.
Those were my thoughts at first, but then again why? Why any innovation is just automating the same cmdlets and concepts we had for decades?
Why can't we just make something that requires less changes, less human intervention, ... ?
I guess other HNer is right when she/he says the problem is to get everyone on-board (IPv6 I'm looking at you), but is there anyone at least trying?
Most of those config errors are router configurations, it's hard to disrupt that because it is tied to the hardware. As a networking guy, I'm not surprised either. Cisco's interface is really awful and they have no interest in making it better. They make tons of money selling expensive courses and certifications. There is very little sanity checking in the interface which makes it easy to shoot yourself in the foot.
Probably because networking protocols are slow to update since everybody has to get on board.