Wxfisch

What you want is a reverse proxy, something like Trafik or Nginx Proxy Manager makes it really easy to setup and even then lets yous apply a lets encrypt certificate for HTTPS support getting rid of the browser warnings.

In a non-malicious way it can all be helpful to websites to know the capabilities of your device to allow it to change what the site delivers/how it renders. Knowing your GPU allows it to know if your device supports WebGL/DirectX/Vulkan etc., knowing light or dark mode allows for it to set the site the same as your system, if the tab is not active it can pause content, and if you have a low battery the site can try to be less power hungry by perhaps not asking to render a ton of active content. Knowing if your on a mobile device can allow the site to deliver a mobile optimized layout, or if you have touch capability to render buttons larger.

The fact that advertisers and data brokers use this to fingerprint you as a user is just a non-intended use of good intention features. In reality, if you do hide this information (which you often can using developer tools in many browsers) you’ll find the some sites will just not work or will act wonky and data brokers will still fingerprint you using things like tracking pixels, your IP, or user agent string info that you can’t really hide without fully breaking the web. You only need three or four individual pieces of information to pinpoint specific individuals in most cases so they don’t really need all of it, it’s just easier and more accurate the more information they have.

This is comparing apples and oranges though. Automotive cooling systems are designed for a very different problem set than datacenter cooling systems. The temperature gradients are much larger in ICE systems, they need to be small, light, and portable, and they cool something that generates much more variable heat loads.

A data center creates a consistent heat load, is stationary, with access to a source of water that is functionally limitless to the operators, cools a much smaller gradient and needs to do so in the most economical way possible to be as profitable as it can be to the owners. Evaporative coolers are dead simple, very effective, and scale very easily which is why they are used.

It evaporates, that’s how it cools. The water is sprayed over a heat exchanger and gets turned to essentially steam and then new water is pumped in and thus the water is “gone”. It will fall as rain somewhere but likely not near where it was taken from.

A closed loop system could be used but they are more expensive and require more maintenance so large data centers don’t usually use them unless required to.

It really depends on what you want to accomplish, your priorities, the amount of time and effort you are willing/able to put into it, and your risk appetite (not just privacy but also availability of your mail server).

It is for sure one of the more challenging services to self-host, and IMO doesn’t offer a huge improvement over a hosted solution with your own domain from an actual security and privacy standards point since email is inherently insecure and non-privacy protecting without adding additional not-always-standard layers on top like PGP/GPG, SMIME, one-time passcode escrow systems, etc. that all have their own huge trade offs.

Your self-hosted server will have downtime as well, some planned but also some unplanned. If your server is down, it can’t accept or send mail obviously which can be an issue (many services will try to deliver again after a back off period, but won’t try forever). Enterprises work around this with load balanced servers and running different services on fault tolerant infrastructure. That increases complexity quickly though and isn't what most self hosters do AFAIK.

It’s a combination of conservative designs, robust training, and a zero tolerance safety stance where even minor misses that have any relationship to the reactor or power systems get throughly investigated through a formal process that seeks to understand and learn from mistakes rather than assign blame.

If anyone is curious, the Naval Nuclear Propulsion Program (NNPP) publishes the Gray Book with some history of the Program, the various arms that are involved to make it successful, and how the Program is managed including training, suppliers, labs, and fleet operations and maintenance.

Turn the Ship Around is a leadership book that also touches on safety and operations of a nuclear sub and is just a good read overall if your looking for a different way to think about bringing a leader in an organization.

Email is de facto not private/secure without adding additional layers to it, so using services like ProtonMail or Tuta are putting lipstick on a pig. They give a false sense of security and privacy that just doesn’t exist without a ton of additional overhead and opsec. Unless you plan to only email other Proton accounts, or use janky one-time password secure messages, your email isn’t E2EE, it’s just encrypted on Protons servers which is table stakes for most paid mail services. They are marketing something that just can’t fully work in the real world. You also then make a ton of trade offs like very limited client support (especially on mobile), and can’t even use S/MIME for compatibility with enterprise secure mail solutions.

To be clear, I think it makes sense to pay for something as critical as email so you aren’t the product, as well is using your own domain for portability. But I don’t recommend folks buy into the false security and privacy promises of services like Proton for email/calendaring.

I use SimpleLogin with my own domain and have no problems 99% of the time. Every once in a while I’ll find a site that doesn’t like my .net domain, but those typically won’t take anything that isn’t a major provider. I use that as a sign I probably don’t need to use that service.

There are two things in my house I don’t “play” with: internet connectivity and core home functions (lights, locks, garage doors, etc). That doesn’t mean I don’t self host anything or then, but I always start from a mindset of “must work”.

I run HA on a Yellow (functionally an RPi 5 with radios and storage interface built in). My lights are either Hue running as plain Zigbee devices, or Zigbee switches. I don’t necessarily want more customization with home automation, I want stable, extensible, and easy to use day today. HA checks all those boxes easily. I’ve not done much looking into OpenHAB, but I would caution against going with something for home automation just because it’s more customizable. Sure, it’s great to have an automation routine that turns on your lights when you get home, it’s less great to have an integration that misbehaves and now you cannot turn off a light, or lock your door, or turn down the volume on your music, etc. Be sure to know what you want to accomplish before you buy devices, build automations, and always build things with a manual backup operation option.

There is, I think, a few things that contribute here.

1. The US has a very stupid “bigger is better” mentality. So if you go out you expect a large portion because that translates to better (and more value). This is of course not true, but culturally it’s very embedded.
2. almost everyone I know takes home some portion of their meal from a restaurant. So that single portion is really two, or maybe three.
3. IME people don’t usually have giant portions at home, they sometimes do of course, but things tend to be more sane for home cooked meals for your family. They also tend to be a lot more balanced, with more veg and grain.
4. what you see on TV is often sensationalized, and not fully indicative of normal here.

I would hazard to guess they are investigating how the use of AI was missed in their editorial process, how they missed the incorrect quotes, and who violated their journalistic standards by using an AI to directly write article text since it’s a coauthored piece.

In typical Ars fashion, the editorial team appears to be looking into what happened and are being fairly open about at things: https://arstechnica.com/civis/threads/journalistic-standards.1511650/

I will be very disappointed if this was BenJ or ~~Dan~~ [edit: I had messed this up, it wasn’t Dan but Kyle Orland that coauthored it] Kyle using AI to write their article since both have had really good pieces in the past, but it doesn’t sound like this is some Ars wide shift at this point. Like all things, it makes sense that it will take time for them to investigate this, Aurich (the Ars community lead and graphic designer) was clear that with this happening on a Friday afternoon and a US holiday on Monday, it’s likely to be into next week before they have anything they can share.