At least in regions like the EU where they are taking EV adoption seriously. Provide additional tax relief or a grant for any car that supports V2H with battery tech that supports the charge / discharge cycle. It means if you have a power cut you can plug the car in to supply the power or even use the car to power the house and lessen the cost (& strain on the grid) during peak periods.

The way I see it is that many websites would be better off with a self signed cert than no cert at all. At least the site traffic is encrypted and a plugin like SSL lighthouse could alert the user if the cert changes.

And a site wants to add trust to their cert then they'd be better off to have their partners, customers, competitors sign their cert than some random CA that nobody has ever heard of. And it wouldn't stop CAs from signing the cert too and perhaps to gain maximum trust in a browser that might still be necessary.

Browsers could show some kind of score / checklist of what the cert says about the site based on who signed it or not. It would still be a better system than the shakedown racket that exists today.

Erm, yes it is and yes it does. At least if you use systemd. And you say "perfectly possible" and yet sysvinit and other systems don't do it.

Your sentence is gibberish. Assuming you mean to say "systemd does not preclude somebody adding security by other means", then no it does not.

But does any sysvinit based dist do it? Devuan, the white knight of the dump-systemd "movement" certainly doesn't. Service sandboxing? Nope. Cgroup process tracking? Nope. Credential handling? Nope. Log tamper proofing with FSS? Nope - rsyslog has some tamper evident features but the capabilities are way easier to set up in systemd. Disk encryption via TPM? Nope. Least privilege? Nope - its all LSB spaghetti code running as a user and no control beyond that.

Devuan admits how terrible it is by offering OpenRC during installation. OpenRC supports cgroups, capability dropping and supervision. But then it runs OpenRC over the same LSB scripts in compatibility mode, so none of that is used. And even if it was used it would still be limited compared to systemd. And it tacitly admit that sysvinit sucks when that was its whole schtick in the first place.

And I don't think people raging about systemd were ever motivated by security, or common sense for that matter. You demonstrated that with your Poettering remark. There are reasonable points of debate / concern about systemd, about overreach, modularity, packaging and suchlike but you're not interested in any of them. Just some kind of irrational fear of change even when it is progressive and beneficial. Almost as though you see a change to a subsystem in an operating system as a threat to your world view. It's very weird IMO.

Fuck all to do with security except all the ways that I stated and that are easy to verify.

Systemd improves security in a number of tangible ways - small commands which do specific & least privilege operations, tamper resistant / journal logging, cgroup process limits, per service sandboxing, disk encryption, credential storage, security auditing tools and more. The sandboxing is probably as close as you would get to OpenBSD's pledge/unveil functionality in a Linux.

Where systemd may not be considered secure is it's certainly not minimalist. It has a lot of commands in that one package. But many of those commands exist to avoid dragging in even larger packages for the sake of using small bits of them (e.g. to set the time without ntpclient), but there have been flaws found over the years.

Linux was NEVER like this. From the early days it was kitchensink mentality - dists would install stuff whether you needed it or not. Dists were especially bad for installing Telnet, FTP, SSH, NTP etc.

It was only when security was taken more seriously that dists started being more selective about what they installed. Most modern dists still have a couple of ports open, mostly locahost only but won't even run SSH unless somebody explicitly chose to install it. There are cut down Linux dists like Alpine which run with almost nothing additional at all and would be suitable for being firewalls, routers and net filters etc.

I think if I were ultraparanoid about security I might still consider OpenBSD, but it has its own limitations in terms of supported hardware & drivers, networking throughput, package support that may prevent it from being viable regardless of how secure it may be.

All that really matters is the content you're trying to play, the DAC and the output impedance. Anything else is basically snakeoil. If Sony is trying to sell this thing based on "oxygen-free copper", or any other weasel marketing terms then it's snake oil. I'm sure that it does have a decent DAC and audio specs, but so do most mid/high tier phones. Anyone paying $2000+ for this is a fool.

My memory is that I had joined Apple the first time right after Jobs stepped down as CEO, when he remained as Chairman... right before he passed away. Then the Steve Jobs interview was released and I saw it at the Aquarius Theater in Stanford and Cringely was there in person at the showing. Stylish as ever.

To me he represented a whole era of computing, when it was exciting me, before I had really started making a living with computers yet. Back when I really did learn a lot by reading physical magazines.

Except it doesn't produce complex code if you tell it not to and review to ensure understanding. And no I'm not "special". I'm just capable of using the tool in a way that mitigates against its excesses and still makes it useful for some but not all tasks I would otherwise have to do by hand.

I'm a realist and I haven't "drunk the kool aid" I can see the absolute shit AI can produce and its limitations. It needs supervision and knowledge and it needs to be used in small, iterative, reviewable steps on tasks it is trained and capable of doing.

The problem with AI generated code, whether it is Claude or not, is a person vibe coding it. If they're ignorant of coding and just prompting AI until it does what they want then the code underneath is 100% guaranteed to be unreadable garbage and probably very brittle, unpredictable, insecure and unmaintainable outside of what it appears to do.

Conversely if it is driven by somebody technically proficient where code is generated, reviewed and adequately tested to specific requirements and standards, then the result is likely as good as handwritten code, better in some ways worse than others. As a developer myself, my assumption is that AI gets dumber the further away it gets from good training data. So if I ask an AI to generate unit tests, or generate code against some popular language or framework then it works fantastically well. It also works better with good context - existing code, rules & guidelines. But if I ask it to generate something for which there is scant information then it will definitely generate garbage and need a lot of hand holding.

I live in B.C. and applaud losing the time change, but think the choice of time zones going forward wasn't the best.

I'm expecting major buyer's remorse this winter when people realize that the sun sets at 5 in December and January, and doesn't rise until 9.

...laura

Keep the time constant the year around and if there are reasons that certain workers or schools should start an hour earlier in summer then change their times by an hour.

Altman is hardly likely to come proclaim AI is a shell game with enormous potential to ruin lives. Because he's got his bank account and personal enrichment to worry about. The grift will perpetuated by his ilk until the bubble collapses or they and their company becomes criminally and financially liable for the consequences.