He reaches the conclusion that this means that it's impossible to build such a system. No. It's impossible to establish a universal clock. It's not impossible to voluntarily participate in a system which coordinates information that allows this

youtu.be/nfRouGH0oMg

Our industry has a lot to unlearn before it can start solving these issues. It's unfortunate that it made these shortcomings into its very identity.

The need to braid multiple relative timelines together to achieve timelines which can label increasingly distant information from each other is such an obvious concept and so will necessarily underlie our data services eventually.

Calibrating streams of independent information is not some niche curiosity. It's the minimal semantics required for allowing independent progress but providing a unified view for observers. You simply can't do simpler

Different CPUs talking to each other will always be slower than the speed at which the CPU can progress. We currently waste these orders of magnitude differences mostly because our theory is bad at managing these

I don't think I'm being unreasonable when I say this shouldn't be controversial. Yet, we've been stuck with working with the "single computer model". Ironically, these stop working for single servers also for the same reason: the need to scale out.

So I presume you would agree that ideally we should address the root cause of the root cause, which is centralisation

bsky.app/profile/omni...

To be clear: I'm not suggesting anyone is lying. I'm saying that it's hard to convince a manager that they should invest in a new tech and an engineer to learn new science if they have nothing to gain from it.

So beyond credentialism, another thing enabling this is lack of competition. If providers would face stiff competition, innovation would be the only option. No market- or industry-pressure means engineers gaslighting everyone else.

The fact is that DNS is a distributed coaching problem and as such, is basically solved. The problem isn't science. The problem is the amount of money invested by different IT providers in bad distributed caching implementations and how much it would cost to fix it all.

They are hoping to eventually work in FAANG, so are being sycophantic to get on the good side of the cool kids. They aren't putting their own companies first by demanding better solutions, they would much rather join in singing "DNS, amirite?" with the others.

This is all to say that I do understand the rationale. But please, spare me the pearl-clutching "it could happen to you"-s. No, it couldn't. And it wouldn't need to happen to them either.

But most importantly, it wouldn't keep happening to the customers if there was no oligopoly

are a choice. It's one the entire industry makes by maintaining and carrying their existing investments forward. If even one of them decided to move forward with a better setup, all of them would need to move and they wouldn't make a penny more, so why would they?

Now, I didn't speak to nearly everyone in the space and few actually understand what this is all about, but there are a good few who do, in detail. So I can say with good confidence that there are important people at major cloud providers who do understand that outages like these

however was the candour with which so many in business would talk about the realities of the situation. I'm genuinely thankful to these people, even if their message was mostly that innovation is not something they're looking to do as long as they have a choice.

problems still considered unsolvable by most.

The second (industry) was much more of a mixed bag. I can't really be surprised by the fact that incumbents are not interested in innovation, I accept that it was very naive of me to presume otherwise initially. What I didn't expect