The main threat of lock-in isn't price increases, it's business termination.
Obscured / redundant infrastructure providers as a first defense against attacks on network disconnects.
I wonder if this will go all the way to ISP network-level attacks on right-affiliated data centers.
@StevenFan Anything a freedom-loving developer involved in the Linux kernel network stack could do to help increase robustness against such attacks? Asking for a friend...
Practically, start having conversations about digital rights and defend them in a principled and knowledgeable way, particularly at any organization you belong to.
You uniquely have access to technical networks and you are credible. Taking a perspective of having long term, in depth conversations and understanding how you can influence your extended network will prepare you to contribute.
@StevenFan oh, I'm already doing _that_ (e.g. in #Linuxgate I spoke out against the new CoC). I was asking more about _technical_ measures: what kinds of features in the networking subsystem would be useful to a RW datacentre fending off such attacks? Better support for multihoming? MPTCP? A 'reverse load balancer' (maybe XDP based) to shunt flows around between different ISP uplinks?
The future belongs to those who show up to build it, so what should I _build_ to get the future we want?
@soundnfury
The attacks I'm thinking about are all termination of end user services. The tech you are talking about is towards the bottom of the dependency graph.
I don't know because I don't know the technology graph or the service provider graph.
I'd ask these questions:
Who is your user?
What are their needs?
With an understanding of the components of the landscape, what can you do to create value for your user?
I'd use wardley mapping for this.
@soundnfury thinking outloud as a layman, it seems like improving connectivity within networks of (self-hosted) data centers are a place you could look into.
@soundnfury I'm interested in learning about this tech. Can you teach or point at resources that contextualize it?
@StevenFan Probably the best thing to learn about is #XDP & #eBPF.
XDP: start here https://www.iovisor.org/technology/xdp
eBPF is the enabling technology for XDP (and much more besides), https://ebpf.io/
With XDP, the network becomes programmable at the kernel layer. E.g. Facebook's load-balancer uses XDP for the fast path https://engineering.fb.com/2018/05/22/open-source/open-sourcing-katran-a-scalable-network-load-balancer/ — easy to see how this kind of multiple point-of-presence could be useful to a self-hosted DC network facing political attacks on its edge connectivity…
@PstafarianPrice remembering our conversation about ideologically aligned economic-partners as a defendable position to have opinions in.