Posts by Flocke Kroes 4557 publicly visible posts • joined 19 Oct 2007
The origin really is "Couldn't hit the broad side of a barn".
Harmondsworth great barn is 58.55 x 11.3 x 11.9 metres, so its broad side is about 7x10^30 barns. TheRegister already has a unit of area, the nanoWales.
Getting a PC without Windows has been getting steadily easier. It has reached the point where major distributors are openly selling blank hardware for less than the same hardware with Windows installed. The final nail in the coffin was banking trojans hidden in the crapware. The effort of searching for trojans exceeded the bribes for installing the rubbish in the first place. Although retailers could include Linux as standard, the first thing any penguin is going to do is wipe the machine and do a clean install, so I can understand them not making the effort.
W10 is Microsoft moving to an ad supported revenue model. The telemetry is part of the cost of the OS. If the cost is too high for you, don't buy it. You have choices that do not involve the old Microsoft tax. Forced updates and bundling of cloud services are more costs known in advance to anyone vaguely computer literate. Pay the costs or leave Microsoft to their thoroughly lock-in victims.
There is a whole chain of blame for lack of Android updates: The telco won't allow it, the manufacturer can't be bothered, the chip vendor will not maintain closed source drivers for old chips. You could wait a decade for the European court to reach a verdict, and another for effective enforcement. If you want updates, check out which makes get regular updates before you buy.
If you think Google are manipulating search results, jump straight to page 2. They are not the only search engine. I am more likely to search Wakipedia or Amazon than use a generic search. Java has an off switch. So does javascript. The web looks so much better without either.
Voting with your feet is far more effective than waiting for the EU to do something useful.
Laser beams and light sails are easy: You put a big fan (laser) in you (space) port, and a big (light) sail on your ship. The fan pushes air in one direction and the Earth in the other, but as the Earth is so big, no-one notices. The air hits the ship. The air slows down (light bounces back) and the ship speeds up.
The first good thing about the EmDrive is that the fan and the sails are both on the ship. The other clever bit is a trick to reduce the size of the fan: You put sails at both ends of the ship, and the fan in the middle. You change the direction the fan spins back and forth really quickly, in time with the frequency that air bounces back and forth between the sails so the air speed builds up like small pushes on swing.
Because it is symmetrical, there is no net force. The trick to getting money out of investors is to make the sails at each end of the ship different sizes, and put a truncated cone around the ship to match the wind to each sail.
I was wondering why this is a problem.
The most obvious answer is that the database contains so many errors that employees cannot rely on it to fill out their travel expense forms. If this became public knowledge, people might question whether this enormous database was worth all the tax payers' money needed to funnel garbage into it. I can see how this would be considered utterly intolerable, and that staff should be thoroughly discouraged from looking themselves up.
The key concept is that linking GPL code with code licensed with an incompatible license results in program you cannot distribute. Back when the GPL was written, linking to a library meant creating a single binary file containing its own copy of some of the code from the library file. The binary would run fine even if you deleted the library file. Later, dynamic linking allowed programmers to create programs that contained no library code, but several programs could share the same library (one copy on disk and up to one in memory no matter how many programs linked to it). The general consensus is that this still constitutes linking as referenced by the GPL.
Before compiling, the kernel must be configured. Most of the configuration options select which parts of the kernel are missed out, which get compiled into a large file and which get compiled into individual modules. Kernel modules can be loaded at any time and connect to the bulk of the kernel (and to each other) in pretty much the same way that a user land dynamic library connects to an executable. According to Linus, loading a kernel module is not the same as linking as referenced by the GPL. Linus is the copyright holder for much of the kernel, and is entitled to decide what is and isn't allowed when distributing derived works of his project. He can certainly decide not to sue even if Stallman thinks he should.
Linus is not the only contributor. Other people have contributed portions of the kernel, and they retain copyright over their contributions. Contributions are only accepted if they are licensed under the GPL or a compatible licence, that way the entire kernel including all the contributions can be distributed as a whole without all the distributors having to reach separate licensing agreements with all the contributors.
Not all contributors agree with Linus's opinion about kernel modules not counting as linking. To allow for this contributors can mark symbols in their own code with EXPORT_SYMBOL (any module can get the value of the symbol), or EXPORT_SYMBOL_GPL (only GPL licensed modules can get the value of the symbol).
The idea is, if you create a non-GPL module (which you can distribute independently of the kernel), and try to load it into the kernel, the kernel will check every symbol reference and will fail to load the module if it tries to use a symbol defined in code written by someone with a strict interpretation of the GPL. (I think compilation will fail, so you cannot not accidently create a module other people cannot load). You can see more of the details here.
Stallman is welcome to limit how derived works of his projects are distributed. So is Linus, but Linus lets contributors select their own interpretation of the GPL without screaming at others that they have to see things his way.
M(v)=M(0)/sqrt(1-v^2/c^2)
M(99% c) is about 7 times M(0). A big difference for the few electrons going fast. Electrons contribute less than 1 part in 4000 to the mass of metals, so you are not going to notice the wire putting on weight until you have most of them going at 99.99999%c. (If real life about 98% of the electrons in platinum are not going anywhere).
The automount demon requires CAP_SYS_ADMIN to mount devices. Back when I was a PFY, automountd ran as root, and did not run fschk before mounting a block device. A defective file system could crash the kernel, and a maliciously crafted one would be able to run arbitrary code as the kernel.
My information may be really out of date as I always disable automountd when commissioning a new system. (I do not use a file manager and I find it irritating to press <ALT><F4> every time I plug in a USB storage device).
It looks like a flash device, but the software in it pretends to be a USB hub with a flash device and a keyboard attached. When it thinks you are not watching, the software pretends another flash device has been plugged in then types the required command to run the malware on this hidden flash device.
Of course, the OS can prevent this from happening by not trusting any USB keyboards, and all the user has to do is type "Trust me" to tell the OS which keyboard to trust.
So, this web interface in going to find out the USB device is attached to a router running openwrt on a MIPS CPU. It determines the correct version of the kernel, applies the required patches and configuration file, modifies the driver code to match the kernel version, downloads a cross compiler and compiles the module for the USB device. It then magically pulls the correct system definition for QEMU, emulates my router in software, tests the driver, fixes some bugs, re-compiles and repeats until the tests pass before sending the driver to my web-browser - except my router does not have a browser. It does not have the X client libraries, so the only browsers it could run would be things like lynx, links and w3m.
Imagine how much easier it would be if manufacturers documented their hardware, and programmers contributed and maintained an free software driver in the mainline kernel. The driver could then be installed with distribution's standard tools and loaded when required without user intervention. Although there are many thousands of USB devices that already work this way, there are a few that don't. This is easy dealt with: no source code, no sale.
I have no problem with a child avoiding tracking (I would be a bit surprised that one made the effort). A way around the porn filter is a significant failure, but hopefully they can fix it. The thing that really caught my attention was that trying to access lukol by https caused certificate not trusted, certificate expired and a redirect to http. The form tag's action attribute was http.
If you use lukol, your ISP will be able/required to log the address and search key and hand the details to your local Stasi.
Oh that poor little Apple corporation outsmarted by the big bad Telcos.
It seems to me that Apple's lawyers and negotiators did not spot a €48.5m hole in the deal.
The Telcos took advantage of Apple's reputation for negotiating horrendous terms with their business partners. Ummm Newsflash: business works best when both sides have something to gain from the deal. No-one gets it right all the time (it would be nice if AMD succeeded a little more often).
No-one forced Apple to try to evade their responsibility for fixing defective kit. They could have set up the required infrastructure to repair/replace/reimburse devices that failed under warranty just like all the other manufacturers who trade in the EU.
I can imagine this working with Chrome OS. The boot loader only loads a signed kernel. The kernel mounts the root partition and the first time each time each block is fetched, its signature is checked before passing it on to the file system layer. It is an effective way to test for a trusted image in the root partition without a big delay on boot. The source code has been available for years. The downside is the effort required to delete the vendor's keys and install your own so only your signed kernels and root file systems can boot may require hunting down an exploit in the supplies OS.
(Do not bother if the device has AMT - unless Intel suddenly document it sufficiently for you to audit it properly.)
I am a Penguin, so my knowledge of Windows is thin and Macs even thinner. I keep seeing Windows PCs replaced because of software issues - the required software only runs on new versions of the OS, the new version of the OS will not work with old hardware. The hardware was fine. A quick Linux install, and ancient machines are useful. Resistance to Windows 10 may well push up the service life of Windows boxes.
From a Penguin's point of view, a desktop does not have a single age. The case might well be a relic from the nineties. There was a time for a Motherboard+CPU+RAM upgrades for speed. Later, a similar upgrade for power reduction or silence. Floppy disks barely got used because of network cables, and were retired in working order because of USB flash. Pre-SATA DVD players were retired when legacy PATA disappeared. After about a decade, the laser in a DVD player becomes too dim to play reliably. Keyboards and mice have a life time proportion to the cleanliness of the user. Disks got regular upgrades depending on bravery and the cost of restoring from backups, but SSDs are big and reliable these days. Properly specced power supplies last for ages (white box shippers liked to fit something with under 1W to spare so it would fry when you add a second hard disk). Monitors lasted until extra desk space was required, and I have not had a reason to refresh them since.
I am not sure how to compare the life time of a Penguin flavoured desktop PC to a Mac. Five years ago, I would have said replace the CPU/Motherboard/memory/SSD every five years, mostly to cut power consumption, and expect the rest to last a decade. The comparison has become irrelevant anyway. My final desktop was working fine when it was retired over two years ago because it's work could be done by £90/10W fist full of ARM (approximately equal to a Pi3 + SATA + GB ethernet + memory bandwidth). By the time that dies, a Raspbery Pi 6 will blend, and if I need something bigger, I will take a closer look at this.
Top of the list: Technical support cost. If it is £5/minute then service faults are a revenue stream and I will go elsewhere.
Next is cancellation: This must only depend on me doing things, like say fill in form, add cheque, post and cancel instructions to the bank. If it requires waiting for the ISP to do something then I will not sign up.
Terms and conditions are the ones on the sheet of paper I signed, not the ones on the web site that change with the wind.
My ISP got bonus points for defining contention ratio and telling what it would be. They even knew the difference between megabytes per second and a millibit.
Data caps are way down the list. If I could buy evidence of purchase for a film and download it then I would take an interest, but the only thing on offer is streaming. In the mean time, the post office is far more cost effective for high bandwidth using sneakernet.
One plan is to split the workload among servers with different operating systems in the hope that a single exploit will not work on both Linux and BSD.
A really quick look for market share shows nearly 2% of http servers running a BSD based OS. There is reason to hope that crackers will focus on the mass market.
Temperature is the average energy per degree of freedom (number of particles multiplied by the number of ways they can move). In a complete vacuum, there are zero particles, and the sum of their energy is zero. Ask your C compiler, and it will tell you 0./0. is not a number.
In real life, space is not a complete vacuum. Depending on where you are, the few particles could have a large or small energy. If you had a really good thermometer, and plenty of time for it to reach equilibrium with its surroundings it could read anything from 2.7 to several million Kelvin depending on where in space you put it.
Phrases like 'Icy cold of spaaaace' have been winding me up for years, and today I snapped because I had hoped for better science from The Register even in a really cool first of April article.
The Lawyer daft enough to agree to represent The SCO Group all the way to the supreme court in return for some shares in TSG is called David Boies. I have heard is does have a brain, even though he could not win a legal argument against a gardener. He must have caught on quite quickly because he shuffled the case off to some minions very early on. The minions have done really well considering they had no evidence to work with.
TSG is run by a bankruptcy trustee. The only way I know that Boies Schiller Flexner can escape before appealing to SCOTUS is to convince the trustee to let them off. Clearly it is cheaper to continue to fart about in Judge Nuffer's court. They should get some legal advice about splitting BSF into the profitable parts and the bit stuck with the TSG/IBM litigation, then bankrupt the useless bit. Perhaps, for reasons of judicial efficiency, they could get Judge Gross to assign the same trustee.
1) Starting with a factory acceptance test, and guarantying payment only after something passes said test should weed out all the suppliers expecting to piss away £24M before anyone asks them what they have achieved.
2) No source code, no money. I had hoped that after over three decades Microsoft's customers would understand the phrase 'lock in', but so far I have been disappointed.
The desktop is not dominant now, and non-gamers are noticing that it is more than they need. Smart phones out-number desktops. A modern smart phone is more than powerful enough for a word processor and the majority of business software. When connecting a phone to a monitor and keyboard become convenient, another chunk will vanish from the desktop market.
Microsoft saw this coming. They spent loads on a smart phone OS, a billion advertising it, over a billion propping up Nokia and more buying out the burned out shell. All because Microsoft knew that being dominant on the desktop will be the equivalent of being the dominant supplier of abacuses next decade. After their abysmal phone failure (35% market share -> 2%), Microsoft have switched to porting their products to Linux.
Next target: servers. The Google and Amazon account for a hefty chunk of the server market. They are designing their own silicon. I have seen plenty of rumours of an Amazon ARM or a Google ARM or Power. For now, I think these are big sticks to shake at Intel. Such sticks can be really effective: Itanium killed Alpha well before there was a functional Itanium CPU. ARM architecture could have made big changes to the server market that we do not see because we do not know the details of negotiations between Intel and Google/Amazon/Facebook.
Mid air retrieval has worked before. I could not find the mass of the film bucket from the Corona missions, but judging from the size, I would guess under 100kg. The plan for Vulcan is to separate the engines from the rocket (explosive bolts?) and just catch the engines. I could not find the mass of the engines, but about 2000kg each is a reasonable guess.
A Soyuz re-entry module is 2500kg, so a parachute is feasible. A CH-54B helicopter can carry 9000kg, so carrying both engines at once should be fine. Early designs of SpaceShipOne (1200kg) included mid-air recovery but that was considered too risky.
I would call mid air recovery of a pair of BE-4 engines at least as impressive as landing a Falcon 9 on an ocean barge, but not impossible.
The purpose of the watch is to prove that they have money to burn. It may sound beyond stupid to us, but some people genuinely need to prove they have money to burn. It makes a real difference to how they are treated. If the watch is useless after a few years then all the better, because wearing a new Apple watch shows they have money to burn regularly.
From a business point of few selling 10,000 hundred dollar watches for $1000 is just as good as selling 1,000,000 ninety dollar watches for $99. The only down side is a boat load of high margin products can be turned into landfill if one of the senior executives does a Ratner.
The watch business could wither and die and you would not notice the difference to Apple's profits. What matters is that about 20% of phone owners are happy to pay double for a phone if it has the Apple logo on it. With a determined effort, Wozniak could change that. I thought he knew better.
In short, Ex-Microsoftie Elop put a clause in his Nokia contract that got him a US$25million bonus if Microsoft bought Nokia. The problem was Nokia had twice the smart phone profits of Apple and Samsung combined. It took a combined Osborne and Ratner to trash Nokia to the point where Microsoft could afford it. The bonus was for a Microsoft buyout, not a sale to the highest bidder. Hence the switch to Windows Phone (Nokia factories could not manufacture the required hardware) and Elop's refusal to sell the award winning (iPhone beating!)N9 in large markets even though Nokia's idle factories could build it.
Elop sent Nokia into the biggest death spiral in business history, but still will not take credit for his outstanding achievement.
At first, I wondered what on Earth robot cars had to do with the budget. If the government want robot cars, then they can change to regulations to make them legal, and Google will turn up. If they are threatening to spend money on robot car software then I can see the project completing in the UK shortly after the NHS goes paperless - and at a similar cost.
Perhaps if the government throw enough money at robot cars then they will have to cancel some other over-priced software project like storing everyone's browser history.
In other news, DARPA is offering money to people who can turn consumer electronics into improvised weapons.
... doesn't matter to Microsoft.
Google has more market share than Microsoft, and Apple has more revenue. Google works giving things away for free and selling adds. Apple sells an expensive walled garden. Microsoft have decided to copy Google and Apple. bazza and a million other previously content Windows users are now third in the queue at best (behind advertisers and commercial licensing) when it comes to what Windows N+1 will do.
Ranting isn't going to fix anything. What you can do is make purchase searches on Amazon (already common) and information searches on Wakipedia (probably the first useful link in search results anyway). If you take the users from Google/Bing, then the advertising revenue will wander off and Microsoft will have to rethink where their money will come from.
If you have a problem with LibreOffice, then fix it - you have the source code and a license to use it. Personally, I use reportlab+python, but each to his own.
Install:
CLI: write script, debug script, document script.
GUI: click on things until it works. Attempt to document what you remember of the correct sequence of clicks missing out the miss-steps made while learning the GUI interface.
Rebuild after disaster:
CLI: type one word and press enter.
GUI: One of the miss-steps you didn't document is required to get the right option to appear on the GUI. Someone re-organised the GUI to make it more intuitive and all your notes are now useless. Fart about for ten minutes until you think you got it right then find out next week what you got wrong.
With practice, you might restore one service with a dozen clicks on a GUI. A dozen services means 144 clicks - or a script with twelve words in it. There are tasks that are better handled by a GUI - like copping one image. As soon as you have to do something repeatedly, (or if someone else may have to do it after you leave) the CLI is the right choice and anyone who says GUI can suffer the death of a thousand clicks.
What is better: 1,000 software patents rubber stamped per day or one software patent permanently rejected per month?
I applaud Battistelli and the union leaders finally doing something useful about the vast number of troll patents approved each day. Lets hope they can keep this up for at least a year.
The FTC started looking at Microsoft's anti-competitive activities on 1991, and closed the investigation in 1993 without taking action. In 1993 the DOJ started investigating, and reached a settlement with Microsoft in 1994, which Microsoft promptly broke when they required all customers to pay for IE bundled with new computers whether they wanted it or not. The trial began in 1998. Microsoft were found guilty of using the monopoly in operating systems to crush competitors, and of tying. The court sentenced Microsoft to be broken into two (an OS company and an applications company) in 2000. In 2001 Microsoft wrote their own sentence which was accepted by the DOJ in 2002, and allowed Microsoft to continue tying.
If Microsoft do crush non-Microsoft Windows App stores, then I fully expect the FTC and DOJ to huff and puff for another decade and achieve nothing.
The new law is over 500 pages long. I have no idea how much tax law the UK has, but I doubt that there is a shortage. The new law is vague. The only practical solution is to negotiate an agreement with a tax man. In a banana republic, this would result in a poor government and rich tax men. I would expect that Google's accountants are not dim, and that they bought a piece of paper that says they have paid all the required back taxes in full.
The fun comes when the government tries to explain why that piece of paper is worthless, and why Google should buy another piece of paper.
It didn't look particularly like Z80 or 68000. It screamed 6502 to me.
I regret that I do not have to source code to the most horrible bug I have ever found. The project involved using DSPs to send position data regularly by radio (pre-GSM), but with each radio transmitting at a different time. I had time from GPS, so sync was easy. Apart from a little assembly language for the signal processing, the bulk of the code was in C. The kit worked fine in the office. Hours of testing with no problems at all. On the next day, I took it to the customer for a demonstration, and the kit sat there doing nothing.
After some frantic bug hunting I discovered the fault was in the C library for the DSP. The 32-bit add function was made out of a pair of _signed_ 16 bit shifted add instructions. The DSP was massively optimised for 16 bit variables (and fixed point arithmetic). CHAR_MAX was 32767 (sizeof(char)==1 by definition in C). As a result, almost all of my C code used 16-bit (int) variables and long int was only ever used for the time. During testing, bit 15 was zero, but the demonstration was at a time when bit 15 was 1.
The easy bit was fixing the code during the customer's lunch break. The difficult bit was explaining what I had fixed.
Busemann's biplane design created shock waves on the leading edges and anti-shock waves where the wing thickness reduced. The idea was that the shock waves from each wing would hit the anti shock wave form the other wing - when the plane was going at precisely the right speed. The symmetry of the design ensured no sonic boom. Unfortunately that symmetry also meant the lift was zero. Not a little lift or low lift, but precisely zero lift. Unless the biplane has more buoyancy than mass it isn't going to fly.
The article My-Handle pointed at did not mention how they solved this problem. Does anyone know?
The reason for the upgrade from Pi1 to Pi2: The usual increases in speed and decreases in power, size and cost with time in electronics allowed a faster CPU and more RAM in the same size without increasing the cost. As a 'bonus', the instruction set for the new CPU matched the one Microsoft supports, so Microsoft could, with a little effort, trim down version of their OS to fit the memory and speed constraints of a Pi2.
Precisely why Microsoft bothered is not clear to me. I could buy a Windows PC, buy some Microsoft development tools, learn how to use them, create something to run on a Windows Pi, then read through the huge license agreements to see if I can distribute the result, and hope that Microsoft's tools will still be maintained at an reasonable fee in five years time. On the other hand, I can use all the tools I am familiar with for free, have confidence that they will be maintained and free for at least a decade, I can distribute my work without fear that a part of it will only be available from a monopoly at an ever increasing price, and I get a full OS tested on millions of servers, not some cut down thing tested almost once by a pair of Microsoft employees.
The reason for the Pi2 to Pi3 upgrade is precisely the same as last time - newer, better kit was available without increasing the cost. Pi3 supports the same instruction set as Pi2 and a 64bit instruction set. Microsoft's IoT for Pi2 should work with little or no modification on a Pi3.
Microsoft are perfectly welcome to distribute an OS for Pi3, or any other computer they want. The fun thing about a Pi, and many other small cheap computers is that they do not come with bundled storage or operating systems. The buyers get to make their own choices. This is a new experience for Microsoft, and I for one welcome them to the twenty first century and wish them luck catching up.
A Pi2 can handle TheRegister, internet shopping, and the enormous attack surface required for internet banking. I am sure there are sites that will bring it to its knees, but a do not visit sites that try to mine bitcoins with javascript. Openoffice did the minimal tasks I require of it (convert docx and xls to odt and csv). Cropping and scaling a 6Mpixel image to 1920x1080 with the gimp required patience.
The solution?
Independent news took a severe beating when people stopped paying for it. Hands up all those who paid to read news on The Register. The Trump 'solution' is to make vague threats against Amazon and Bezos. [I think a more productive solution would be to convince people to pay for their news. It wouldn't work because "Who's boinking who" would get a ton of cash for fictional articles, but I have yet to see a better idea.]
If Trump gets nominated and elected (or wins as an independent), he becomes commander in chief of the US military. In theory he cannot declare a war on Bezos - that power remains with congress, but this has not stopped other presidents. He can order troops to surround Bezos for up to 60 days. He can promise a presidential pardon to the person who murders Bezos. He cannot create a new law directly: He can ask/pay a congressman to introduce his law. Then he has to get/buy enough votes to have to law passed, and finally the law will probably get struck down if it is not constitutional. In theory, he cannot buy congressmen with the military budget because that requires congressional approval (there could be a design flaw in the US system of checks and balances here). Finally Trump can appoint federal officials. Bezos could be investigated by the IRS and the Antitrust Division while getting raided daily by the ATF. President Trump could cause 'such trouble' for Jeff Bezos, but I am not convinced this will cause fair and honest newspapers, TV and internet news any better than asking people to pay for proper investigative reporting.
In real life, Trump talks to his faction, and gets hardly any votes from the other Republican factions. Trump has got this far because the other factions all had multiple candidates splitting the vote. Trump is now down to four opponents and two of those will be gone soon. His loyal supporters are not enough to get him nominated in a three way split. His next step to being nominated has to be a completely new message that appeals to at least two factions - and to find a really good excuse to avoid Megyn Kelly. If he does not get nominated, Trump is sufficiently big headed to run as an independent.
Although Trump is Hillary's dream opponent (doubly so as an independent), I doubt they are working together because Trump would never make a deal where he gets second place.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
