Gosh, that wasn't what I observed. When 1c and 2c coins were removed in 1988, prices were rounded to the nearest 5c. Initially I thought this could be gamed, since the consumer decides what to buy and can ensure it rounds in their favour - but since the whole situation arose because 2c wasn't worth people's time, this was short lived.

It's true that the long run effect has meant fewer $1.99 items and more $2 items, and it's true that implies a one-time inflation hit, but it also seems to benefit both consumers and businesses to get rid of those goofy prices anyway.

If the frustrations and anger of 2024 are different, what are the new areas to reboot and reimagine?

(I miss the revolutionary days and can't believe there's nothing out there to overthrow, while everything today feels so ... settled.)

It was published as an MFC sample, then samples were published under the MIT license. I kept the pre-ribbon one. https://ancillary-proxy.atarimworker.io?url=https%3A%2F%2Fgithub.com%2Fmalxau%2Fword...

Note the data here are per-student. I've always been suspicious of per-student numbers, because there's another narrative that colleges are accepting more students than ever (since the tuition fees are so lucrative!) then complaining that state funding isn't sharply increasing in response. Note that's also neatly compatible with the claims in this article of huge capital expenditure (if you want more students, you need more buildings.)

Looking at Washington (my home state), the fiscal service reports spending $6.78bn on higher ed in 2014, rising to $11.99bn in 2023. This is roughly in line with overall state spending (which went from $38.6bn to $74.5bn in the same timeframe.)

Do you have good nationwide data on absolute spending?

Well, Windows does detect common 16 bit installer stubs and automatically invokes the 32 bit installer, even if it can't natively execute the 16 bit code.

It was a curious choice to not emulate 16 bit code on 64 bit Windows though. Remember all of the RISC versions included software emulation of 16 bit Intel code.

In 2001, it was feasible (but herculean) to build a new alternative browser core.

Today, I really wonder how this could be done. A Chromium fork is possible in theory, but the community size needed to maintain it just doesn't seem to exist. Any fork driven commercially is going to have the same incentives as the current one.

Thanks for the thoughtful post.

What I'm puzzled by is even if hardware or software is old, if it was previously able to perform this recalculation promptly, something else changed.

My uninformed guess is this is the result of efficiency improvements: the closer an airline gets to a fully booked crew and fleet, the more follow on effects occur from any given change. In the extreme, a perfectly efficient airline implies any change to anything must recalculate every future movement thereafter.

If this uninformed guess is accurate, a software change is unable to really fix it, and a hardware change can only make linear improvements. It's a traveling salesman problem that quickly becomes unsolvable, so rescheduling things promptly implies not attempting to fully solve it (which creates another pile of headaches later.)

Important context for anyone trying to reason about those numbers: https://ancillary-proxy.atarimworker.io?url=https%3A%2F%2Faaron-margosis.medium....

Well, I'm a Microsoftie, just not part of Windows anymore. I wrote http://www.malsmith.net/edit/ for those SSH moments. That said, I agree that there's value in having something that you know is built in, since plenty of downloadable options exist.

Well, it did. I think what bustinbrains is getting at is edit was a DOS program, so it was removed when moving to a 64 bit native OS, and Windows 11 has no 32 bit native edition, so edit doesn't exist in any version of Windows 11. But Windows Server hasn't had a native 32 bit edition since the Vista-based WIndows Server 2008, so hasn't had edit for a long time either.

I think the problem is the CDC is the intersection of science and politics. Science can quantify the decrease in probability of infection caused by wearing a mask in different conditions. Deciding that it's worth people's time and energy to actually wear masks requires evaluating the probabilities against inconvenience and public opinion, which is intrinsically political. Blurring these lines is dangerous, because purporting that science unambiguously resolves a trade-off is at best misrepresenting things, allowing for very trivial counter-arguments to be presented.

This isn't new for the CDC either. The CDC have long maintained travel advisories for different countries. Measuring what illnesses occur in which countries is largely answerable by research. Telling travelers what precautions are worth taking depends on all kinds of non-science factors. Given a particular disease, the advice might differ if a precaution is $5 at a local pharmacy or a $50k experimental treatment.

Do you have a more recent link? That link appears to describe changes in 2018 with adjustments in 2020. I don't think it describes what's happening in 2021.

According to the Register article (and the text of 1605.3(v)(5)), devices need to consume (depending on expandability) 75 kWh/yr or less. That's a 25% reduction on the previous value and doesn't appear to refer to the PSU. There's a pile of adders based on components including a discrete GPU, storage device, Ethernet controller, and RAM size. Knee-jerk reactions to regulation aside, this particular set of regulations seems to fairly administratively burdensome for an assembler to figure out for each configuration whether it is still compliant; any time a user changes a configuration on a website, the allowable power budget and consumption is changing.

I was at the talk. His main point was that each source file needs to have a copyright indicating a license, not just a license attribute of the project, which he believed did not provide sufficient protection for the license. He mentioned the risk of code being taken from one project to another and being re-licensed accidentally in ways the author of that code had not intended, because the source file failed to explicitly state its own license. I presume he was looking for ways that GitHub could encourage this type of behavior.

I'm sure he's also not happy about not requiring projects to explicitly specify a license. Obviously he's intimately familiar with software copyright and licensing, and mentioned in the talk about how without a license copyright means nobody could copy that code for almost any purpose, including to execute it. But I don't recall him explicitly commenting on GitHub not forcing people to choose a license, despite clearly seeing how nonsensical it is to fail to specify one.

Note that Office was doing this in Office 95, before each of these.

rundll32.exe pnpclean.dll,RunDLL_PnpClean /DRIVERS /MAXCLEAN

/MAXCLEAN means remove every driver not used by a currently installed device. Without it, the system will purge drivers that are obsolete (replaced by a newer driver) but will keep drivers for devices that may not be plugged in, because it doesn't know if you're about to plug them in five minutes later. That's the behavior you'll get from the cleanup wizard.