Facebook's Downtime And Why Protocols Are More Resilient Than Centralized Platforms

As you know by now, much of the tech news cycle yesterday was dominated by the fact that Facebook appeared to erase itself from the internet via a botched BGP configuration. Hilarity ensued -- including my favorite bit about how Facebook's office badges weren't working because they relied on connecting to a Facebook server that could no longer be found (also, how in borking their own BGP, Facebook basically knocked out their own ability to fix it until they could get the right people who knew what to do to have physical access to the routers).

But in talking to people who were upset about being cut off from Facebook, Instagram, WhatsApp, or Facebook Messenger, it was a good point to remind people that another benefit of a protocols, not platforms approach to these things is that it's way more resilient. If you're using Messenger and it's down, but can easily swap in a different tool and continue to communicate that's a much better, more resilient solution than relying on Facebook not to mess up. And that's on top of all the other benefits I laid out in my paper.

In fact, a protocols approach also creates more incentives for better uptime from services, since continually screwing up for extended periods of times doesn't just mean losing ad revenue for a few hours, but it is much more likely to lead people to permanently switch to an alternative provider.

Indeed, a key part of the value of the internet, originally, was in its resiliency of being highly distributed, rather than centralized, and how it could continue to work well if one part fell off the network. The increasing centralization/silo-ization of the internet has taken away much of that benefit. So, if anything, yesterday's mess should be seen as another reason to look more closely at a protocols-based approach to building new internet services.