[conspire] (forw) Re: [skeptic] A butterfly flapped its wings and:

[Rick Moen]

----- Forwarded message from Rick Moen <rick at linuxmafia.com> -----

Date: Wed, 9 Jun 2021 16:18:39 -0700
From: Rick Moen <rick at linuxmafia.com>
To: skeptic at linuxmafia.com
Subject: Re: [skeptic] A butterfly flapped its wings and:

Quoting Laurie Forbes (laforbes at telus.net):

> *A major internet blackout that hit many high-profile websites on
> Tuesday has been blamed on a software bug.*
> 
> Fastly, the cloud-computing company responsible for the issues, said
> the bug had been triggered when one of its customers had changed
> their settings.

As the old saying here in the open source world goes, 'There is no 
"the cloud"; there's just someone else's computer.'

> https://www.bbc.com/news/technology-57413224


"Fastly" is a Content Delivery Network (CDN) company, essentially a
place you pay to handle incoming requests for your Web traffic and serve
them cached copies, because otherwise you expect to be overwhelmed,
either because some parties hate you and are trying to take you down
with Denial of Service (DoS) attacks or because you've just become
insanely popular with legitimate traffic and need help handling the
load.  Typically, CDNs have data centres widely distributed around the
world so that they can use GeoIP techniques to serve up cached content
from a network-wise close location to the customer, hence with good
responsiveness.

Major CDNs (that may sound familiar) include Cloudfare and Akamai.

Here is a story about Fastly's embarrassing problem of yesterday, that
is less uselessly vague than was Auntie Beeb's:
https://gizmodo.com/how-one-fastly-customer-broke-the-internet-1847061755
And a VP's blog post:
https://www.fastly.com/blog/summary-of-june-8-outage

In order for a CDN's services to be usable by paying customers, they 
must be able to communicate over the Web to inform the CDN network about
what to cache, under what circumstances to refresh the content from the
customer's own Web server, 

The description of the specific error cascade is vague.  Some customer 
changed his/her local IP address or DNS configuration, and through
triggering an unspecified bug in the software connection to Fastly's CDN
software, which in turn cause massive software failure within Fastly's 
CDN management software, taking out 85% of their capacity for an hour.

As should be obvious, this isn't supposed to be possible, and indirectly
points to one hell of a serious bug.

----- End forwarded message -----