Verizon, Cloudflare, Microsoft, Facebook, Apple, and Twitter all have problems.
Last week, Verizon caused a major BGP routing error, causing most of the Internet's brands (including CDN Cloudflare) to be down for a day. This week, the rest of the Internet is clearly following the footsteps of Verizon.
Third-party users report a local failure on Office 2 on July 2 at downdetector.com
Facebook and its WhatsApp and Instagram suffered a wide range of failures related to image display most of the day. The problem seems to be that in some image tags, the wrong timestamp data is fed to the company's CDN. I looked at the bad images on my own Facebook timeline and found that different timestamp parameters were embedded in the same URL. Loading an image from fbcdn.net with the wrong “oh =” and “oe =” parameters (or no parameters at all) resulted in an HTTP 403 error message "wrong URL timestamp”.
If you delve into the code, Facebook references the same image URL, with different parameters in the anchor tag and image tag; in the corrupted post, the URL in the anchor tag works fine, but the URL in the image tag does not work. (On the plus side, everyone is now suddenly aware of the AI-driven hidden alt tag that the company has been embedding in images since 2017.)
Twitter's private messaging service has also encountered some unexplained failures today. The failures seem to occur primarily in the eastern United States and Europe, with little or no reporting failures on several third-party fault tracking sites in other regions.
Twitter’s private message service failed on July 3, such asHttps://outage.report/twitter-direct-messagesShown.
Apple's "System Status" page report shows that iCloud's calendars, contacts, and alerts are currently affected by ongoing disruptions, and some users may be affected as well.
At about 7 o'clock in the afternoon yesterday, many netizens posted Weibo saying that the editor's graphic could not be saved or the image failed to load. After that, the WeChat public account login interface cannot be opened.
There seems to be no common reason to correlate these failures, which is a terrible week for Internet engineers.