Category Archives: News

Article note: Anyone else remember when in ~2012 PGI (Portland Group, who at the time were a subsidiary of STMicroelectronics), a long time builder of high-performance compiler tooling was showing a new round of platform-independent parallel accelerator tooling including teasing a credible independent CUDA implementation that could target non-Nvidia platforms, then Nvidia bought them in 2013 and it disappeared from the face of the earth in favor of "we're excited to be working on CUDA Support for FORTRAN"? ...Yeah. Nvidia seems well aware that the "There is existing code many users want to run written in CUDA, and you can only run CUDA code on an Nvidia part" situation is their competitive advantage. ATI/AMD's failure to settle on a stable GPGPU toolchain (CTM/THIN/Brook+, Stream, ROCm with OpenCL, HIP, and perpetually broken CUDA compat...) and OpenCL's ugly boilerplate gave them an opportunity to get that core set of lock-in software, and they're not giving it up without a fight.

Article note: The "Refurb, AOE" Chromebooks being dumped for cheap are _awesome_ toys. Selling "new," full price, no warning out of support ones is a different, terrible thing. The main appeal of Chromebooks is the idiot-proof management system (for individuals or institutions) so long support windows are a big deal. The question of why the support windows _are_ so short is still valid; new Chromebook hardware really isn't substantially higher spec in the last several years, and the fact that years of aged out ones can run mainline Linux _just fine_ indicates there is likely no major technical reason that ChromeOS-flavor Linux needs to be dropping 2016-era machines... which indicates "business reasons."

Enlarge (credit: Getty)

Google resisted pleas to extend the lifetime of Chromebooks set to expire as of this June and throughout the summer. Thirteen Chromebook models have met their death date since June 1 and won't receive security updates or new features from Google anymore. But that hasn't stopped the Chromebooks from being listed for sale on sites like Amazon for the same prices as before.

Take the Asus Chromebook Flip C302. It came out in 2018, and on June 1—about five years later—it reached its automatic update expiration (AUE) date. But right now, you can buy a "new," unused Flip C302 for $550 from Amazon or $820 via Walmart's Marketplace (providing links for illustrative purposes; please don't buy these unsupported laptops).

That's just one of eight Chromebooks that expired since June while still being readily available on Amazon. The listings don't notify shoppers that the devices won't receive updates from Google. The US Public Interest Research Group (PIRG) pointed this out in a press release Wednesday, sharing screenshots of the models:

Article note: ...because it didn't work for shit. It turns out "series of grams statistically similar to what a human would write" and "A human trying to write like they think is expected of them" are nigh-indistinguishable. Especially if you have multiple weighting nets available and/or reseeding - which we do. It spells doom for the "Write about nothing" paper as a non-cheesable assignment. I'm not entirely sure that's a problem, those usually didn't involve a lot of thinking anyway.

Article note: Majestic. 13 solenoids for output, photo-interrupters for input, all interacting with selector bails, a little micro between a bank of FETs and a level shifter to glue it together electrically and some 3D printed bits to glue it together mechanically. The anachronistic (and not hideously expensive) 2741 every good nerd thinks about when they look at a Selectric.

Article note: An incredibly obscure object, hacked deeply and described in great (software) detail.

Article note: Aw, he was such a character in our own cultural mythos as computer people. Pancreatic cancer at 59.

Article note: The least surprising. Follow prestige to find fraud.

Article note: Free space optical networking is always nifty. This is a _drastically_ more sophisticated/complicated protocol stack than most of the other recent interesting ones (802.11bb implies a whole ass 802.11 stack!) and it sounds like all the compliant implementations are rather expensive right now, but it's neat.

Enlarge / "I can't believe the quality of this 8K video tutorial! It's like you're really… Ted, could you, yeah, just shift over a little there? You're blocking… thanks." (credit: LiFi.Co)

Light is almost certainly the fastest thing around. So it makes sense that "light-based wireless communications," or LiFi, could blow the theoretical doors off existing radio-wave wireless standards, to the tune of a maximum 224GB per second.

So long as there's nothing blocking the space between your receiver and the lightbulb you've fashioned into a LiFi access point. Or you don't need to turn the bulb off entirely to sleep. And you're willing to add a dongle and keep it pointed the right way, at least for the moment.

But LiFi, or 802.11bb, isn't really meant to replace Wi-Fi, but complement it—a good thing for a technology theoretically nullified by a sheet of printer paper. In an announcement of the standard's certification by IEEE (spotted on PC Gamer) and on LiFiCO's FAQ page, the LED-based wireless standard is pitched as an alternative for certain use cases. LiFi could be useful when radio frequencies are inhibited or banned, when the security of the connection is paramount, or just whenever you want speed-of-light transfer at the cost of line-of-sight alignment.

Article note: All software preservation is under-managed, though games are particularly culturally relevant. And that is the point, copyright is to encourage things to enter and be available to the culture, and if the legal regime is not accomplishing that goal, it should be renegotiated.

Enlarge (credit: Aurich Lawson)

Anyone with a passing interest in retro games knows the bulk of classic video game history is effectively "out of print," with legitimate copies limited to defunct hardware platforms and secondhand physical copies (if you're lucky). Now, in a first-of-its-kind study, the Video Game History Foundation has determined the full extent of this issue, finding that a full 87 percent of games released in the US before 2010 are no longer commercially available.

This vast expanse of out-of-print games isn't exactly "lost," of course; libraries, archives, and even software pirates have helped ensure the games will continue to be accessible in some form. But the VGHF argues persuasively that the poor market availability of reissued games highlights how the game industry is not doing a sufficient job of preserving access to its own history.

"The industry has done a great job re-commercializing a wide catalog of [popular] titles, but for the vast majority of games, we can't rely on the commercial market to solve this," VGHF Library Director and study author Phil Salvador told Ars in a recent interview. "We need to give libraries and archives more tools to get the job done."

Article note: I've been saying RHEL is only relevant because it's an agreed-upon standard with an adequately slow rate of change. Entities want to be able to easily use code and documentation from elsewhere, and easily hire experts to support it (1). Some of that is inertia driven, but nothing else about it really matters. Thinking they can generate their own network effects is a really classic IBM blunder (2). Much of RH's historical relevance was basically "RedHat exists to launder Linux for the National Labs" and that basis is not a terribly deep moat - hell the big physics centers maintained their Scientific Linux fork for 16 years before fully shifting onto CentOS then getting rug-pulled... which is not the kind of relationship maintenance you want to do with the customers whose network effects create your value proposition. The death condition for RH is if the Alma/Rocky CentOS community successors and the Oracle type commercial clones all agreed to a different reference point for "Standard Enterprise Linux"(3), and the big producers of code people want to use (your large Physics centers and and NIH scientific compute projects, and/or the shared infrastructure like OpenHPC) tracked that other reference point, suddenly RH brand EL is mostly irrelevant. SuSE running their own EL fork is interesting because there was not an obvious successor reference point to coordinate the non-RH ELs and/or major users if they have trouble tracking RHEL, and now there kind of is - it'll be really interesting to see what happens if there is any divergence. (1) The support point is "why not Debian," they've never had the commercial support partners, certs, etc. Having that was a huge part Ubuntu/Canonical's proposition, but they never really got the enterprise/scientific compute/HPC penetration that RH has. (2) Ya'll remember PS/2 and MCA (Micro Channel Architecture)? IBM was gonna take back control of the PC industry by setting an new (much more license controlled) standard and... the PC cloner industry set up outside coordination points and end-ran them with EISA/VLB and eventually PCI. (3) It might not get called SEL for SuSE or Standard because that would be confusing/trademark problems with the adjacent-industry Schneider Electric. I half jokingly translate the "Enterprise Linux" terminology into "Srs Bsns Linux" some of the time anyway, so the name game will surely be funny.