Under the CMB method, it sounds like the calculation gives the same expansion rate everywhere. Under the Cepheid method, they get a different expansion rate, but it’s the same in every direction. Apparently, this isn’t the first time it’s been seen. What’s new here is that they did the calculation for 1000 Cepheid variable stars. So, they’ve confirmed an already known discrepancy isn’t down to something weird on the few they’ve looked at in the past.

So, the conflict here is likely down to our understanding of ether the CMB or Cepheid variables.

Except it’s not that they are finding the expansion rate is different in some directions. Instead they have two completely different ways of calculating the rate of expansion. One uses the cosmic microwave background radiation left over from the Big Bang. The other uses Cepheid stars.

The problem is that the Cepheid calculation is much higher than the CMB one. Both show the universe is expanding, but both give radically different number for that rate of expansion.

So, it’s not that the expansion’s not spherical. It’s that we fundamentally don’t understand something to be able to nail down what that expansion rate is.

And the article content posted is just an excerpt. The rest of the article focuses on how AI can improve the efficiency of workers, not replace them.

Ideally, you’ve got a learned individual using AI to process data more efficiently, but one that is smart enough to ignore or toss out the crap and knows to carefully review that output with a critical eye. I suspect the reality is that most of those individuals using AI will just pass it along uncritically.

I’m less worried about employees scared of AI and more worried about employees and employers embracing AI without any skepticism.

Thanks. Very interesting. I’m not sure I see such a stark contrast pre/post 9-11. However, the idea that the US public’s approach to the post-9-11 conflict would have an influence makes sense and isn’t something I’d ever have considered on my own.

Me too, but I’d put Usenet in there before Slashdot.

Because most people aren’t technical enough to understand there are alternatives, particularly if those alternatives involve removing a scary label telling you not to.

The South. Just below Indiana, the middle finger of the South. And I say this as a Hoosier for much of my life.

Nice job. Packet loss will definitely cause these issues. Now, you just need to find the source of the packet loss.

In your situation, I’d first try to figure out if it is ISP/Internet before looking inside either network. I wouldn’t expect it to be internal at these speeds. Though, did you get CPU/RAM readings on the network equipment during these tests? Maxing out either can result in packet loss.

I’d start with two pairs of packet captures when the issue happened: endpoint to endpoint and edge router to edge router. Figure out if the packet loss is only happening in one direction or not. That is, are all the UK packets reaching DE but not all the DE making it back? You should clearly be able to narrow into a TCP conversation with dropped packets. Dropped packets aren’t ones that a system never sent, they’re ones that a system never received. Find some of those and start figuring out where the drop happened.

Just curious if you’ve had the chance to dig into this and can report anything back?

If the bandwidth numbers you’ve described are accurate, I’d start looking at CPU and RAM usage on the network device. The Fortigates are going to be doing extra work to handle the VPN. I wouldn’t expect an IPSEC VPN on a Fortigate to top out at 10mbps, but if it’s doing a lot of other work, it’s possible. ACL’s on the Cisco devices? You run the potential of CPU/RAM exhaustion on those. Hopefully, you have remote monitoring on all network devices and you can just look at the history when these transfers are happening.

If nothing obvious there, then I’d try packet captures when this is happening, perhaps to start on the system doing the ssh and on one or two others experiencing issues. What are you seeing? Evidence of dropped packets? High latency? If dropped packets, start capturing the same traffic on the network devices it’s flowing through.

Kids these days with their containers and their pipelines and their devops. Back in my day…

Don’t get me started about the internal devs at work. You’ve already got me triggered.

And, I can just imagine the posts they’re making about how the internal IT slows them down and causes issues with the development cycle.

Yep. I’ve hosted my own mail server since the early oughts. One additional hurdle I’d add to you list is rDNS. If you can’t get that set up, you’ll have a hard time reaching many mail servers. Besides port blocking, that’s one of the many reason it’s a non-starter on consumer ISP.

I actually started on a static ISDN line when rDNS wasn’t an issue for running a mail server. Moved to business class dsl, and Ameritech actually delegated rDNS to me for my /29. When I moved to Comcast business, they wouldn’t delegate the rDNS for the IPv4. They did create rDNS entries for me, and they did delegate the rDNS for the IPv6 block. Though the way they deal with the /56 IPv6 block means only the first /64 is useable for rDNS.

But, everything you list has been things I’ve needed to deal with over the years.