Re: This is what I keep saying
I suspect it is not, you just haven't realised yet.
7164 publicly visible posts • joined 16 Jun 2009
Backlit keys are great.
A dim glow with some colour coding for mode and special keys is a wonderful thing, especially when working in the dark.
Backlight control software is horrific, because out of the box it cannot do any of the above.
On the bright side, now that backlit standard keyboards are so common, the actual keys themselves are far cheaper so building custom backlit keyboards with custom legends is affordable.
When I was regularly onsite with a small laptop my USB numeric keypad was an essential bit of kit.
For numeric data entry it absolutely cannot be beaten - but woe betide the user of a numeric keypad where the decimal point key does not match the numeric entry locale!
Make paying off ransomware a criminal offence with no defence whatsoever.
In some cases it already is, so prosecute those to the maximum extent permitted. Publicly and noisily, por encouragement les autres.
At the moment it's just money, so the board and senior management try to guess the cost of paying and the cost of not paying. The criminals will always try to make it appear cheaper to pay - though this is usually a lie.
If the board risk personal direct consequences, they won't authorise payment, and ransomware dies within a year. The criminals aren't stupid, they just want the cash.
The criminals will still try to break in, but they'll be trying to steal data to sell on. So they'll try not to disrupt the business as that gets them found very quickly.
Yes, but what does that mean?
Cell sizes vary quite a lot, and combined with signal strength it can be a rather skinny donut.
The visible WiFi SSIDs, MAC and signal strength can locate you down to within a couple of metres, so one hopes they're not doing that.
IIRC, the idea was that you'd have some capacity on-prem, then use EC2 to handle surge using all your existing tools.
So your on-prem estate can be sized for the normal workload, and you'd rent and provision some VMs in EC2 during product launch or the runup to Christmas etc, then shut them down afterwards.
Which is perfectly reasonable, and a good fit for "cloud". But can only exist while VMware allowed you to use your own licenses or rent them via Amazon.
Amazon were of course expecting that in reality you'd end up slowly moving much more there, as when your on-prem servers need replacing the beancounters would push towards opex instead of capex.
Start with "Here's six pictures, choose one you like the look of". Make it clear that they can easily change their mind afterwards.
Step two is "download this and run it".
The vast majority of people only need a browser, LibreOffice and Steam for games.
They don't care what it's called, so don't make that front and centre. Just a few pictures so they can see it looks basically the same as Win XP, 7 or 10 - whichever they prefer.
Then install, migrate their browser details over and there you go.
Done.
They don't need to know that Steam uses Proton which is a fork of WINE, or that their distro is an Ubuntu or Debian or whatever. Or even that systemd exists. They can find that out later if they care - which they likely don't.
We ship products based on both Windows IoT and Linux. Most of our customers don't even notice, let alone care which. All they see is that it does the things they want.
If industrial electricity in (eg) Scotland became significantly cheaper than inside the M25, a lot of these bitbarns would suddenly realise that they don't need to be near London.
Then we wouldn't need to build suchbmassive interconnects either.
Transporting power is orders of magnitude more expensive than the associated data, and for the vast majority of purposes (246% of AI) the change in latency is irrelevant.
Section 106 money is notoriously difficult to usefully spend as it ends up with weird conditions about exactly what can be done with it.
Community Infrastructure Levy is a new way of doing it that supposedly is more useful, but I don't think many places have actually done anything with it yet.
You're making the unwarranted assumption that anyone above middle management at the Post Office is smart. Or capable of thought. Or indeed has any memory whatsoever.
Apparently none of the CxOs remember any of their time there at all. One wonders why such forgetful people were ever employed.
Yes, it was bad then. It's far worse now.
The privatised rail infrastructure directly caused multiple deaths, and was renationalised.
The privatised rail operators are currently being paid by the Government - all ticket revenue goes to the government, and the operators are paid to run the trains. Their profit is entirely a subsidy from the taxpayer.
Put simply, the current situation is the worst of all worlds.
For water, the private companies have invested almost nothing. Instead their assets have been stripped and they've been loaded with massive debts - in many cases borrowed from the company that owned them.
Privatised monopolies never work.
The President is not permitted to do this.
It is illegal.
Some of what Trump is doing are things that only Congress can do.
Some are things that only the judiciary can do.
Some of what he's doing is straight-up illegal and nobody at all can do without a Constitutional Amendment, ratified by the States.
Imagine if Obama had banned all private firearms, no exceptions. You'd rightly expect the courts to reverse that decision immediately - because that would require a constitutional amendment.
If I hacked your leg off, I'd be arrested and prosecuted - even if I claimed it was "necessary" because you had gangrene.
Because I'm not a doctor, and you didn't consent.
Interesting article.
Almost completely and utterly wrong, but interesting. Explains a lot about Rust.
Things like ECS and messages are orthogonal to OOP.
Though you pretty much need OOP to implement either of them in a reasonable fashion - otherwise, how does one extend or replace parts of an existing component or message?
There's a reason why almost every large C library has objects, despite it not being part of the language.
Though you don't need OOP to use them - declarative is very powerful.
They'll need to learn some new skills, like swapping the right dead drive, and keep up to date on the VM host management software they chose.
On the other hand they don't need to continually relearn where the cloud provider moved things so they'll have time to do the above.
All their application stuff will be the same, no matter where it's hosted.
While kind of true, that's misleading.
Linux can install updates to huge parts of the system without rebooting, but the applications themselves have to be restarted to actually start using the patched version.
Installing generally takes a lot longer than restarting, so the application downtime can be kept very low - often such a short time that nobody even notices.
Same with the kernel. You can indeed install module updates and even kernel updates, but they will not actually be used until the updated part is restarted.
Hence rebooting is safest, because then you know that everything is using the new stuff.
Not C++. That's the thing. They're different languages.
There are several design decisions in Rust that mean it cannot ever safely replace C++.
Rust can however replace C for small standalone projects - and if they get around to fixing the glaring holes before another language-du-jour does, then it may eventually handle larger C projects.
If you're looking for a C++ replacement, it might well be Swift - though realistically it's far more likely to be modern C++.
The main reason people get annoyed is because Rust cannot safely replace C++.
Several core design concepts of C++ don't exist in Rust. You can fake them, but to do that you wrap the entire codebase in "unsafe".
In reality, the design of Rust is such that it's only suited to small, monolithic projects that have little to no UI.
Rust could be very good for replacing small self-contained C utility libraries, and probably for some embedded. But poor to disastrous for anything large, modular, or long-running - and basically impossible for anything with a GUI.
Though given that the default fault behaviour is to terminate the entire process, and library code must never terminate the process, maybe it's not good for that either.
Most likely because there's an infinite variety of flavours of Markdown, none of which are compatible.
The compatible subset is basically UTF-8, headline, bold and italic. Usually one layer of bullet points, but even that is pushing it.
The IETF have even gone so far as to say "there is no such thing as "invalid" Markdown". So in many ways it's worse than early Word.
"Reject" is required to be inconsequential.
Any site that breaks if you reject is breaking the law. That is, as they say, the entire 'ing point.
I have seen a few sites that break the law - oddly, they all seem to have been the same parent company. I've reported them to the competent authority, and blocked them.
Eventually they will get fined, but it is likely they'll fail due to zero visitors first.
The trouble is that almost all development tools require local admin.
Some because things like "attach a debugger to a process" fundamentally require privileged access, but most because the teams making development tools have local admin and don't test at lower privileges.
The majority of commercial SDKs can only be installed by running a privileged installer - for no technical reason whatsoever, as they're just some precompiled binaries and the header files.
And don't get me started on the weird build systems that insist on downloading all kinds of stuff from random places. Some of them can be redirected to an internally-auditable location, but many cannot.
Docker is of course the cause of (and occasionally solution to) many of these problems.
Virtual reality devices like Microsoft’s HoloLens and Apple’s Vision Pro are also targets.
HoloLens is completely dead, even the military project is over - and it won't even run with Windows 11.
Vision Pro theoretically still exists, but has no software other than the ports Apple paid for, and likely won't exist next year as you'd be crazy to spend 3500 when the 500 units are as good, possibly better and have far better software support.
The idea is that the biometric or PIN never leaves your device. It's only used to unlock the local keystore that has automatically generated keys. The user never sees or interacts with the keys.
The real problem is - if all the keys are on your phone, WTF do you do when someone steals it?
You no longer have any of your keys. Hopefully the thief cannot get the keys out.
But all the keys are gone.
It's going to take a few days to weeks before my insurance replaces the phone, so do I lose access to anything for that whole time?
How do I regain access?
How do I revoke all the keys in that stolen device?
What prevents a miscreant from doing that takeover?
Ah yes, and when Microsoft breaks Windows 11 passkeys again (it's done that several times to my work laptop), what do I do?
Except that there's still the CAPEX to buy the "thin client". Which also needs support and licensing.
So it is even more expensive per user than it appears at first sight.
The alternative is to spend that same capex and lower licence fees for a local machine that supports an infinite number of non-concurrent users.
In terms of overall costings and assuming DOGE's figures are correct, DOGE has cost the US at least $10-20 billion a year.
So the same or more money spent per year, for far less actual stuff - an abject failure.
However, we already know their figures are wildly incorrect, and a lot of the things cut were actually bringing in a lot of revenue - directly and indirectly. So in reality, DOGE is worse than the estimates.
What it has succeeded in doing is trashing all the departments who were investigating Musk's various business ventures. I'm sure that is entirely a coincidence.
Sentencing guidance for attempted murder is 3-40 years at the moment.
Reckless endangerment might cover it too, which is theoretically up to life.
Perverting the course of justice has a precedent, so is easier for the prosecution.
Obviously Category 1 in every case where police attended, so 9 months to 7 years depending on culpability.
Yes, however allowable expenses still aren't free.
At the "small profits" rate of 19%, assuming a normal profit of £10,000:
Tax paid is £1900, income after tax is £8100
With an extra £500 expense, pre-tax profit is £9,500, tax paid is £1805, income after tax is £7695.
So that £500 expense cost £405.
Real taxes are a lot more complicated because the rates for various expenses vary wildly for no apparent reason, which is of course why you pay a tax accountant. The maths is easy, but the rules are deliberately obtuse.
If your job involves a significant amount of Microsoft Word, then you'll be starting it immediately after logging on - not waiting ten minutes.
If it doesn't, then you probably aren't going to for a few hours, if at all, because you're doing your normal job.
So most days this will simply burn energy loading something into RAM, then unloading it again. And some days it will load into RAM, unload, then load again when you actually want it.
All while slowing down the things you're actually using.
It's worse than that sometimes.
I recall someone refusing point blank to accept that temporarily changing the global locale of the entire process from within a library was a bad idea - despite the language documentation explicitly saying not to do that.
They insisted it was fine because they "put it back" before returning - and it wasn't an issue in their proof-of-concept single thread application. They couldn't comprehend that multi threaded applications exist.
Now imagine if that was a closed source library?
JS is still entirely single-thread of course, and Python is ... a bit weird. So there is a lot of pain coming because the "free lunch" of ever increasing single thread performance is over.