Cloudflare Global Network experiencing issues

Are people really this confused?

Every.Single.Time

[1] https://x.com/dok2001/status/1990791419653484646

Your power is provided by a power utility company. They usually serve an entire state, if not more than one (there are smaller ones too). That's "centralization" in that it's one company, and if they "go down", so do a lot of businesses. But actually it's not "centralized", in that 1) there are actually many different companies across the country/world, and 2) each company "decentralizes" most of its infrastructure to prevent massive outages.

And yes, power utilities have outages. But usually they are limited in scope and short-lived. They're so limited that most people don't notice when they happen, unless it's a giant weather system. Then if it's a (rare) large enough impact, people will say "we need to reform the power grid!". But later when they've calmed down, they realize that would be difficult to do without making things worse, and this event isn't common.

Large internet service providers like AWS, Cloudflare, etc, are basically internet utilities. Yes they are large, like power utilities. Yes they have outages, like power utilities. But the fact that a lot of the country uses them, isn't any worse than a lot of the country using a particular power company. And unlike the power companies, we're not really that dependent on internet service providers. You can't really change your power company; you can change an internet service provider.

Power didn't used to be as reliable as it is. Everything we have is incredibly new and modern. And as time has passed, we have learned how to deal with failures. Safety and reliability has increased throughout critical industries as we have learned to adapt to failures. But that doesn't mean there won't be failures, or that we can avoid them all.

We also have the freedom to architect our technology to work around outages. All the outages you have heard about recently could be worked around, if the people who built on them had tried:

- CDN goes down? Most people don't absolutely need a CDN. Point your DNS at your origins until the CDN comes back. (And obviously, your DNS provider shouldn't be the same as your CDN...)

- The control plane goes down on dynamic cloud APIs? Enable a "limp mode" that persists existing infrastructure to serve your core needs. You should be able to service most (if not all) of your business needs without constantly calling a control plane.

- An AZ or region goes down? Use your disaster recovery plan: deploy infrastructure-as-code into another region or AZ. Destroy it when the az/region comes back.

...and all of that just to avoid a few hours of downtime per year? It's likely cheaper to just take the downtime. But that doesn't stop people from piling on when things go wrong, questioning whether the existence of a utility is a good idea.