Inside the pods it makes no sense, but I do enable it on some memory-constrained worker nodes. Note that the kubelet by default refuses to start if the machine has any swap at all.
I wrote this guy a few years back. It's lock free for both consumers and producers. Blocking variants are also available, but with significantly poorer performance.
In contrast, all of Europe has demonetized its old currencies (Deutschmark, French Francs, Italian Lira, etc.) when switching to Euros, and a few times before that. I've just looked it up and one Napoleon Franc is currently worth ~20€.
Yeah, I know of one example where idea was actually created and pitched to media outlets.
They did not like the idea. They very much more like the idea of owning the customer and keeping the customer paying subscriptions.
I also know of a media outlet which incentivizes customer to go from credit card payment to automated bank transfers, since credit cards expire every x years but the bank transfer goes on forever.
When you're paying for RHEL you're not paying for the software that you are installing. All of that can be downloaded.
You are paying for:
- A reproducible target. You know EXACTLY what code you are running. If you manage more than three installations this is the only way you can diagnose and fix whatever issues your installation has.
- Support. The very few times I used RHEL support I always got timely and thorough assistance. Even when chasing a hardware bug or issues with third party device drivers.
- Backwards and FORWARDS compatibility. Red Hat systematically backports kernel bug fixes and support for new hardware to old kernels. We ran 2.6 kernels on Intel hardware released long after the 2.6 series were EOL.
- Device drivers. No, not for your five dollar mouse, but for hardware that costs the same as a small SUV.
If you're avoiding RHEL due to cost, have a look at their SKU list and talk to your local sales org, they have a wide range of options.
(Not affiliated with Red Hat or IBM, but RHCE since 2004)
> To reduce transmission overhead, the Kermit protocol uses a simple, but often surprisingly effective, compression technique: repeated byte values are represented by a count+byte combination.
> Analysis of large volumes of both textual and binary data shows an average compression of 15-20%.
And it relates to Windows and Linux only, and using the TPM.
My guess is that unreliable TPMs made it risky to have this enabled by default.