The good RAID controllers had a battery-backed writeback cache, which would make transactional workloads (lots of fsync) much faster by reordering across write barriers. It's an order of magnitude faster in many cases.

You can't do that without battery-backed memory to replay the transactions after a power loss. Memory-only cards did exist, but they were rare and proprietary. In practice 99% of people who needed this just used a hardware RAID card.

150% correct.

My Firefox on Linux has a system title bar which I can double-click to windowshade (just have the titlebar visible, no actual window.) I can click the downarrow button to get the menu that lets me pin the window to all workspaces, or keep it above all other windows.

It has a File / Edit / History / Bookmarks / Profiles / Tools / Help menu below the title bar.

It has square cornered tabs, positioned above the browser window, not above the URL bar.

It has a URL box and a search box - the search box lets you select which search engine to use.

It pays attention to the system theme, so the scrollbars are wide and permanent, and the edge is more than 0.01 pixels wide so I can actually grab and resize the window. (Anybody who likes tiny edges and disappearing UI elements has never had to use these things on remote connections with rescaled windows - it's all too possible for a screen edge to get scaled away completely, or there not being a trigger point to unhide the magic missing scrollbar.)

I'm running MX with XFCE, have the compositor disabled and the GTK NOCSD library installed - that banishes _most_ of the absurd "I wanna make my own window decorations that don't do what _you_ expect!" BS from Firefox and other moronic software UI developers.

Every few FF updates, I have to go get new User Chrome to put the tab bar back where I want it.

All in all, my Firefox window looks and works a lot like Netscape did, which is fine.

At least there's still a bunch of mechanisms in FF to make it work the way I want it to.

- Old man yelling at Clods.

When you find out that the $4500 you would have spent for a brand new 8-bay NAS plus 8 pieces of 8TB HDD (at today's absurdly high prices, would have been like half that two years ago) is far less than a years' worth of data storage charges, and absolutely peanuts compared to the legal fees, court costs, third party specialists, and on and on and on.

Zero excuse not to have several local copies of such a small amount of data - especially today, when you can buy single drives with 26TB of storage.

My home NAS holds 50TB - that's 8 x 8TB in Raid 5, striped for 7+1. Before the AI nightmare sent prices through the roof, the whole thing cost under two grand, and that's with dual SSDs for read/write buffering. Let's see what they might have been spending for cloud storage:

Google charges $ 0.02 / GB / Month - so 50TB is a grand a month. By the end of a year, they could have had a shelf full of NAS units, a couple in the CEO's garage, whatever. Even if their provider was charging half that, they'd _still_ have a new NAS every 4 months.

Everybody seems to think "The Cloud" is free...

The button hole camera doesn't stream everything to Meta, so they can perform facial recognition on everyone who walks by. It's a lone creeper vs mass surveillance.

'Remember, Pilots - The Ground is "Down". Now, on my mark, everybody land... Go!'

They have different color schemes, background art, fonts, spacing on the digits, etc. Image recognition can easily do this without having to OCR the text.

System architects:
Statically linked executables are a problem, we should break them up into on-disk libraries to save RAM/disk space, ease upgrading, and enhance security. This means devs won't have to constantly release upgrades when library functions are updated!

Reality:
Libraries constantly evolve, changing APIs and behavior. Developers need to modify their programs to track library changes they didn't need or ask for, and to keep manifests of what versions of each library a given version of their executable works with. Linux is capable of having many versions of libraries available, but that feels cluttered to the distribution packagers, so basically no older executables run on Linux.

Developers:
This is confusing and hard. I will put my entire laptop into a Docker container and ship that.

System architects:
Docker containers are gigantic and unwieldy and nobody knows what they contain or of they can be safely modified. We'll make a file that contains the executable and just the libraries that it works with - Flatpak! Snap! Appimage!

Computer scientists:
You mean you've invented a file format that ships all the pieces to build what is essentially a statically linked binary that gets assembled at runtime?

System Architects:
No, you college educated person, it's immeasurably different than that in ways which I am currently incapable of articulating, but you can be sure it is so.

Correct answer:
Just go back to shipping statically linked binaries, FFS.

If companies had to pay for your commuting time instead of just stealing it from you, they'd be way more interested in having as many employees as possible work from elsewhere.

This is not news for nerds. This is not stuff that matters.

I know that ship sailed long ago, and slashdot is kinda mid-grade engagement bait these days, but come on, there is no pretense of substance in this one. It's monkeys flinging poo on the front page. Please try to be better than this.

Happens I know both SCO's Expert and IBM's Expert. They both independently analyzed the codebases, each developing a similar set of analysis tools (think of a modified version of grep which searches for similarities in code patterns that remain after changing variable names and re-factoring equations to do the same math in more or fewer lines.)

Both researchers reached the same conclusion - the code which was the same in both systems was free to use - open source and MIT licensed (primarily code that was taken from the BSD source tree.) SCO's Expert went to the legal team and said that there was no stolen code, but that IBM likely had violated their obligations to the UNIX trademark by diluting its value when they put their weight behind Linux. The SCO legal team didn't feel that was a path forward, and continued to pretend there was stolen code.

Really have no complaints with battery life so far - knock wood!

2017 Tesla Model S P100D - 9 years on the road, 202155 miles (purchased at 50K miles), range 86% of nominal. 270/315 miles

2015 Tesla Model S 85D - 11 years on the road, 125251 miles (purchased at 16K miles), range 90% of nominal. 244/270 miles

Both vehicles have free Supercharging, and that gets used a fair amount. Both now have home charging, but the 2015 was almost exclusively supercharged for about 3 years.

These battery packs have active cooling and heating, and an exceptionally good level of cell balancing, which doubtless helps with the longevity.

Also, in the 250K+ miles we've driven, we've had one unscheduled stop - a damaged tire. Biggest maintenance item so far was a failed cabin heater and leaking coolant valve on the P100D, that required them to drop the battery to fix it, the rest were several minor warranty repairs, a cracked suspension bushing on the 2015, replacing a rock-damaged windshield, and between the two vehicles there were some door handles and a trunk cinch actuator repaired or replaced. Most of the service work was done by "Tesla Rangers" in the driveway.

The P100D looks and drives like a two-or-three-year-old vehicle which has had a couple of curbside parking dings, the 85D looks and drives even better. I remember my first automobile - a 1970 Plymouth Duster, slant 6 engine, which had the engine's top end rebuilt at 75K miles and died utterly at 150K miles. Obviously a different class of vehicle, but it was thought of at the time as a vehicle which was so reliable it would outlast the owner - in comparison, it seemed to have been manufactured from empty soup cans and bailing wire - even the paint was mostly gone by 150K miles.

36.1% pass would be worrying if this was a qualification test of things it needs to be able to do. It's not. This is a benchmark, and it SHOULD have a low pass rate. That's how you know if you're making improvements.

We could quite easily create a different benchmark where it passes 99.9%. That wouldn't mean the device being tested is good. It would just mean we have a useless benchmark.

I have no opinion on whether AI is good or bad for this use case. I just hate when statistics are used to mislead people.

https://f6ffb3fa-34ce-43c1-939d-77e64deb3c0c.atarimworker.io/prefs => Discussions => uncheck "Disable Sigs"

Your sig is currently '"Even if you don't have a Tesla, your children still walk in the footpath." - Lukasz Krupski (ex-Tesla employee)'. You can change it on the User tab in prefs.