Sure about that?
Helium can be forced to form a few compounds, including He2. In real life, I doubt you would find any in a hard disk.
1452 posts • joined 19 Oct 2007
It is a secret. The Royal Mint's explanatory web pages and video say it is excellent, but does not give a word of detail. So if you get an ISIS coin, you can tell it is genuine because ... err ... erm ... well you cannot tell it is genuine because all the new security features are secret. I assume the security features are like the emperor's new clothes - if you cannot see them you are not fit for your job.
Lightsquared and GPS use different frequencies - but they are close to each other. GPS receivers should filter out frequencies not used for GPS, and be immune to Lightsquared's transmissions. In real life, GPS manufacturers used cheap filters that let enough of Lightsquared's signals through to cause confusion.
Lightsquared think this is not their fault, so GPS manufacturers should use better filters and everyone should buy new receivers. The FCC say that Lightsquared's license is dependent on their signals not effecting GPS - even though GPS receivers should use better filters. Lightsquared became a litigation company specialising in suing the FCC.
The real purpose of chapter 11 is to keep the creditors at bay while lawyers transfer the company's remaining assets to each other. This can go horribly wrong if the largest creditors agree to form a committee to run the company in chapter 11.
1) Handed some data on a USB flash key. Cheap laptop: plug it in. New Mac: need to carry an adapter.
2) Handed some data on an SDHC card. Cheap laptop: plug it in. New Mac: need to carry an adapter.
3) Need a fast network connection. Cheap laptop: plug it in. New Mac: need to carry an adapter.
4) Device gets broken or nicked. Cheap laptop: can replace screen or whole device for a pittance. New Mac: ouch.
There are some cool things that could be done with USB C. The charger could be a USB hub with an HDMI socket, ethernet port and SDHC slots. Apple have taken care to let their customers pay extra for the benefits of USB C, but that seems to be what their customers want: "Look! My computer is more expensive, shiny and fragile than yours!"
Here is the dr̷aft law.
Overview of legislation in draft: 226 pages
Draft clauses and explanatory notes for Finance Bill 2015: 552 pages
Some UK MP's can read a page or two. I am sure nearly two of the 650 can read five pages. If you want to find someone who can read and understand all 552 pages, try asking a Micrappoogle accountant. One of them probably wrote it, and made it that long to hide a dozen loopholes.
The good news is that as the MPs are wasting so much time and effort on something that will achieve nothing, they should be too distracted to do anything more damaging.
Which lies have you heard about your 'free' smart meter?
'It will only cost an extra £20 on your bill.'
'Your bill will only go up by £20 per year.'
'It will only increase you bill by £20 for the fist year, £40 for the second and £60 for the third.'
The staged increases in electricity bills have already been approved to cover the cost and installation of smart meters. There is no requirement that bills should go back down afterwards. I am looking forward to MPs getting an e-mail like this:
I pwn your smart meter. I will let you have power between 12:00 and 13:00. If you want power for a whole month, send me a bitcoin.
Jason Bourne records a phone conversation with Noah Bosun, and plays back the first two words (Noah Bosun) to Noah's safe. The safe opens and is full of incriminating evidence.
Years ago, early attempts at speech recognition (understanding what was said) succeeded at voice recognition (identifying who is speaking). I could say 'Help! Help! He has a gun!' and voice recognition would happily allow access to my account. Someone can do an excellent impression of me saying 'Flock of crows', and get access to his own account.
Finger prints are just as good as voice: they give you a list of account names of people with similar fingers/voices. If you have few enough customers, that list might have only one entry, and you have a useful identification device. Identification (the account name) is not the same as authentication (confirming the user is the owner of the account).
Understanding the difference between voice and speech recognition is beyond the ability of most PHBs. Clearly no-one has yet been able to explain the difference between identification and authentication to a bank manager.
Has she been replaced by a digital alternative?
Why are there lots of craters, but no rivers or clouds?
1) Look at all these patents. It would be a shame if one of your products faced an injunction. You need protection. Would you like to pre-install Office 365?
2) All our distributors get a contribution towards marketing funds that are a big fraction of the cost of Windows licenses. It would be a shame if you received a much smaller contribution than your competitors. Would you like to pre-install Office 365?
3) Some boiling frogs are so locked into MS Office that they actually like it pre-installed.
Have you got over $100 in your bank account? Is your credit limit over $100? Can you borrow $101 from Wonga?
The satellite will be over the equator, but will have several directional antennas concentrating signals north or south.
SpaceX concentrated on minimising R&D cost and time. Last year's prices were $61.5million for 13150Kg to low Earth orbit. After paying off R&D, the launch cost is expected to fall to $1100/Kg assuming recovery of stage 1.
Skylon is a much more challenging design. The budget figures are $12billion R&D, 15000Kg to low Earth orbit for £650/Kg (including R&D). I could not get dates for the prices (probably 2004), so I have not tried to adjust them for inflation. Skylon has not yet received 1% of its R&D budget. If the money appeared tomorrow, the first test flight could be in 2021.
Skylon would have to stay on budget for years to compete against a mature Falcon in 2022. On the other hand, SpaceX could keep their prices near current levels and buy Skylon. Plenty could happen in the next seven years. The Chinese are eating their own dog food, even though it costs more than SpaceX. The EU are looking for ways to cut costs. The US government are looking for ways to increase launch costs and I have no idea what the Russians will do.
Real quotes from PHBs: "I know how to tighten screw", "Using the torque screwdriver makes me look incompetent" and "Look! I got it to hold together using only two screws!".
At 16:9 aspect ratio, a 152" screen + bezel just about fits through a standard EU door without tipping it diagonally. If you have a desk that can take the weight, you must either trim the legs, remove the ceiling or tilt the screen to make it fit.
I have always wanted the screen size to go the other way: a few mm across, with a lens so screens can be a few mm from each eye. So far, I have only seen such screens with low resolutions and high prices.
Hyperloop is a solar powered transport system that will compete with traffic jams and light aircraft. Anyone would think you wrote you comment without checking your facts.
Gigabit ethernet + USB3 is a rare combination on ARMs, and pushes you over half way to the cost of an Intel box. I started reading more carefully when I saw USB3. A quick check of what is available puts Mad Catz Mojo on the short list for when I need a new high spec cheap silent computer.
Recovering a big spinning disk from backup over USB2 should take about 50 hours. Half that if you have 2 USB2 interfaces (most of the time, the hardware is a few USB1 interfaces, a USB2 interface, and a port multiplexer that will assign USB1 interfaces to ports with a USB1 devices attached until it runs out of USB1 interfaces. All the remaining ports share the same USB2 interface). USB3 would be limited by the sustained transfer rate of a spinning disk (~7 hours for a 4TB disk). eSata + an Sata hub should be quick too, if the chips are compatible. (You are lucky to get 1 Sata port on an ARM. I have never seen 2).
USB3 has value, but it is not something I need every day.
The fission explosion creates a burst of neutrons that smash lithium into tritium and helium. This gets you your fission fuel when you need it without the hassle of trying to store a radioactive cryogenic liquid.
The purpose of the source article is to demonstrate the importance of keeping up to date with the patches with whatever software you are using. No-one gets to sit back and say "I don't need no steeking patches", no matter what OS they are using. The statistics do point at two important security tips not mentioned in the article: "If you do not need it, do not install it", and "If at all possible, turn it off".
For a proper comparison, you need to know what is being defended, and who it is being defended against. Publicised exploit statistics are not a good source for comparison. I would suggest setting up multiple high value targets with the same budget, regularly pulling the hard disks, comparing the contents to a clean install and seeing which OS survives the longest.
If you search hard enough, you can find computers with no OS installed. They usually cost more than the same hardware with Windows. Years ago, that was because you were still paying the Microsoft tax even though the software was not installed. These days, crapware can more than pay for the minimum Windows license.
I used to be annoyed by the lack of crapware available for Linux. Now all the crapware in the world cannot bring the price of a new Intel box down to the price of an ARM sufficient to replace a dead desktop.
Superfish's biggest achievement is to educate some noobies about the value of a clean install.
Standards can last. I expect that when 7nm becomes mainstream, there will be an ATX motherboard for it.
The big problem with ATX is that almost all customers can keep their old monitor, keyboard and mouse when they 'refresh' (does anyone say 'upgrade' anymore?). Customers able to use a screwdriver can replace the motherboard, CPU, memory, graphics card, power supply, optical disk and hard disk individually as required. The laptop was a great leap forward, requiring a regular purchase of a full set of new components. Modern designs include cases that crack if you try to upgrade the hard disk or memory, glued-in batteries and self destruct when the warranty expires.
Customers with a clue have wanted modular laptops with standard parts for over a decade. The big manufacturers have worked hard never to repeat the mistakes they made with ATX. Sometimes a small player proposes a modular laptop (or phone). All goes well until people see the high price caused by lack of economies of scale and a poor deal on the crapware. It would be great if more customers could appreciate the long term savings available when upgrading only components that matter.
OLPC have been around since 2005. By past performance, I would expect a modular computer in 2017, and an upgrade module in 2020. Some way will be found to avoid providing a machine suitable to large numbers of people in wealthy countries.
Makes the machine cheaper. Even if I was convinced a new computer came with a clean install of the OS of my choice, wiping and installing is required to proove I can restore from backups.
Everyone with one of these routers can find the private key. If the key is not on the internet already it will be soon. Everyone who knows how to set up an ssh server will be able to pretend that their box is one of the 250,000 routers. After they have stolen all the underwear, how to they profit?
A public key is the product of two large primes. The corresponding private key is the two primes not multiplied together, but in either order, so there are only two possible private keys. It is almost certain that if two public keys are the same, then so are the private keys. In this case, any competent cracker with physical access to the device can read the unencrypted private key. (In the other case, any cracker able to get the unencrypted private key from the telco could just as easily get every unencrypted private key if they were all different). There is nothing to decrypt here.
If all the keys were different, and I had physical access to Alice's router, I could install my device that can pretend to the telco that it is Alice's router. As all the keys are the same, I cannot do that because the telco knows beyond all possible doubt that the secret key is not secret.
<voice style="John Cleese/Romanes eunt domus">
If understanding this is beyond the ability of the average commentard, imagine the near impossibility of explaining to a PHB why the telco needs to spend money maintaining a database of which customer has which key. If by some fiendishly cunning stratagem you sneak the database into the telco's budget how on earth are you going to explain what is going on to the customer when his router does not have the secret key in the database?
You are not making any sense. You can try to crack a public key - create the secret key by factorising the public key. You can try to crack an encrypted secret key by guessing the password. Even when I did not know which key this was about, the private key/keys were not encrypted. The challenge is to steal the unencrypted key/keys from where they are stored.
Dropbear's banner tells you the public key used to authenticate the device, so the secret key is on the device. With physical access and a little hacking, you can copy the secret key to some other device. You can put some other device where the router is, and when the telco tries to log in, you can fool them into thinking your device is the router they supplied.
As the same key is used by many routers, if you can get between the telco and someone else's router, you can convince the telco that they are updating the customer's router when they are really talking to yours. This leaves the customer's router with out-of-date software.
Now imagine what changes when every router has its own key. The telco could keep a database of which key belongs to the router of each customer. When they do updates, they can check that they are talking to a device with the right secret key. If they assume someone with physical access to the router did not copy the key to another device, then they can have confidence that they are updating the router in the customer's home and not some man-in-the middle device.
I can easily imagine the majority of ISP's not bothering to maintain the database. Pretend one does, and finds the wrong key where they expect a customer's device. There are plenty of legitimate reasons: mixing up which device went where so the database is wrong. The customer using his own router - and giving the unused one to a friend whose router broke. The customer generated a new key pair, or the NSA are preventing router updates. So the telco knows something is going on. What are they going to do?
customer man in the middle with an explanation of the issue? Phone up the customer and explain what a secret key is?
Is there a genuine threat that could realistically be countered by telling each router to generate its own keys? If all the keys were different, would you assume that your new device is the only place where the secret key is stored?
If someone finds Linus's secret key, all those computers could be fooled into thinking Linux signed some source code that he didn't! Even worse, type:
gpg --recv-keys 79BE3E4300411886
and you get a copy of Linus's public key, and something similar will get you anyone else's (if they have one). It is almost as if public keys were available to anyone!
The routers have two obvious uses for ssh keys. One use is for authenticating the router - in which case the same secret key is on every router. I could copy that key to another device, and the telco could be fooled into thinking they are talking to any one of their routers when they are really talking to my laptop.
The other use is for remote administration. Each router could have its own key. When the telco does an update, the computer doing the update needs to know the secret key for every router. If a cracker can get one, she can get any or every secret key, so having only one key does not remove any security.
What is the issue here?
The big secret is to be Western Digital, Seagate or Toshiba. Those are the only drive manufacturers left. The margins on drives are so thin that enormous economies of scale are required to make any profit. The spinning disks market is in its final stage of consolidation. A new player would need to commit running their business at a loss until they can get above 10% of the market and refine their manufacturing process to the same efficiency levels as one of the big two. In real life, a new manufacturer will implode before they have a business worth being crushed and bought out by WD or Seagate.
WD and Seagate could release their firmware under GPL without harming their businesses. The thing is, I am not sure it is their firmware. It certainly used to be a component they bought in - like the controller cards. If this problem gets fixed, it will be by people creating 'The Open Rotating Disk Initiative Obnubilating NSA' for themselves.
If your example is correct, the phone makers should have bought a chip they could control.
Personally, I blame the customers. They should have checked for a cyanogen installer before purchase.
"... presents identical and similar product offers that may have lower prices"
In this context is 'may' equivalent to 'almost never'?
Did I fall asleep?
The original Russian disclosure says SSH, not SSL, so step one is to install and enable SSH, on a port where people can find it. This is a terrible idea on any machine visible on the internet because the machine will be found and hit with a continuous stream login requests attempting to find an account by brute force. Although this stands no chance of success with even basic precautions, it does waste a little CPU time and lots of network bandwidth. (The most popular way to avoid the network traffic is to set up port knocking.)
Next, you will have to make some changes to /etc/ssh/sshd_config. The one that is definitely required is 'PasswordAuthentication yes' otherwise all attempts to log in with a password will fail (sshd should be set up to require one type of public key authentication, and have all other methods disabled). You can save the crackers some time with 'PermitRootLogin yes'. Without that, the cracker will need to use some sort of privilege escalation - which a competent cracker probably knows. Next you need an account with a password that was not created with a random character generator. If you permitted people to log is as root, make sure you set root's password to a word or two out of the dictionary, swapping i to 1 and o to 0. You can save some network bandwidth by using the most popular pasword: 123456. (Logging in as root should require logging in as an ordinary user, then upgrading to root access).
Next up, this malware requires a bash script in /etc/init.d/ to install itself. The vast majority of them are sh scripts, but I did find a couple of bash scripts. The malware is looking for '#!/bin/bash', which is the way to specify the bash interpreter in Linux. The BSDs require '#! /bin/bash', and Linux accepts that too for compatibility. You can trip up this version 1 installer by adding a space to bash scripts in /etc/init.d/ - if you have any.
The translation of the incident page said something about using a virus scanner to detect infection. I stopped reading at that point because the advice is clearly bollocks. If you installed and configured sshd to use the ssh port and password authentication with a brute forceable root password then you computer will be infected with something that can hide from any virus scanner running on the computer. You might be able to find the malware by pulling out the hard disk, putting it in a USB enclosure, attaching it to a different computer and comparing it to your backup.
I think the biggest barrier to catching this malware is that something more nasty will get in first and close up the configuration errors before everyone and his dog pwns the machine.
Speech recognition is for working out which words were spoken, but this article is about voice recognition: identifying the speaker. Make sure to create a recording of your voice so you can watch films you purchased even if you catch a cold.
The dielectric constant changes with frequency (it is called dispersion). This was even a problem with 28800 bit per second modems. The signal for a single bit was spread all over the audio spectrum, and spread out in time between the customer's modem and the exchange. One proposed solution was to send many slow bit streams at the same time - each using a small amount of bandwidth centred on a different frequency. High speed optical links send many bit streams at different frequencies to counter dispersion.
If you sent an analogue signal over hundreds of kilometres of ethernet cable, dispersion might make a measurable difference. Back in the stone age, that is pretty much how the phone system worked. All the other sources of noise hid the effects of dispersion.
Some materials have lower dispersion than others, and it is possible to select pairs of materials that cancel out dispersion over a limited frequency range. For ethernet cables, lots of effort goes into reducing cross talk, but I have not seen any mention of dispersion.
The obvious thing these cables are missing is some carved amber end-caps so the electrons don't fall out while the cable is in the post. Only $1000 each - you know they're worth it.
In a good conductor electrons travel millimeters per second. A semiconductor has far fewer electrons (or holes) that can move, so given the same current density, the electrons travel much faster. If they want fast electrons, silver is a really bad choice.
When one electron moves, it leaves behind an excess of positive charge that attracts electrons. The place it arrives at gets an excess negative charge that repels electrons. Although the electrons themselves barely move, regions with extra or missing electrons move fast - like a ripple on a pond moves far more than individual water molecules.
An excess of charge in one place is a voltage. A change in voltage moves along a pair of wires at the speed of light in the insulator between them. Light travels through popular insulators at between one half and one third the speed of light in vacuum. High frequency traders have already switched to air to reduce latency.
I can just imagine audiophools listening to their music with vacuum spaced ethernet cables while a pump chugs away to maintain the vacuum.
When someone wants to sell a book that proves their cult religion is sciency, they usually follow the sequence: quantum -> observer -> consciousness -> bullshit
Inside the box: Either [the atom decayed and the cat is dead] or [the atom did not decay and the cat is alive]. The Geiger counter amplified the energy difference between a decayed/non-decayed atom until there was a clear macro-scale difference: a cat the is either angry about being shut in a box or dead. Inside the box, the Geiger counter is an observer.
Outside the box: In the thought experiment, the box is so magical that no clue about the state of the cat can escape. Not the faintest vibration from her breath or heartbeat. No hint of RF from nerve impulses. No difference in which warm atoms vibrate differently because either the cat's immune system is digesting bacteria or the bacteria are digesting a dead cat. Because the box is magical, the wave function that describes the contents of the box is a superposition of states of a live cat and more states of a dead cat.
As soon as the box is opened, the wave function collapses. That term needs some explanation: either the probabilities for different states of a dead cat collapse to zero, and the sum of probabilities for states with a live cat zoom up to one, or the other way around. The rate at which the probabilities change depend on the sum of the energy differences between the possible states. A nerve impulse from a live cat might cause a photon of RF energy to escape from the box just as the lid opens. That photon could cause a molecule outside the box to vibrate. That vibration could change the way other molecules vibrate. Because the difference between a dead and a live cat is macro-scale, billions and billions of differences between the states leap out of the box the moment the lid starts to open. Those differences create other differences outside the box that grow exponentially. That exponential growth or amplification is what collapses the wave function.
The key feature of an observer is amplification. If one atom can change electronic state, and the only possible result of that change is another atom changes electronic state then the result is a wave function in a superposition of states. If that second atom consequently emits a photon into a photomultiplier, the photomultiplier amplifies the difference between possible states and collapses the wave function.
Tiny parts often have a warning in the data sheet. The packaging is so thin that light can get through turning diodes and transistors become photodiodes and phototransistors. You do not normally see this in action because of EMC shielding or the box that the device comes in.
The size of a resource is usually estimated as some multiple of the reserve, and the size of the reserve is set by the market. If a new use for helium is discovered, more will be trapped from oil wells, the reserve will increase, and we will have two or three decades of the resource again.
I found more information here. As I became a penguin last millennium, I could easily have misunderstood some the things I read about Windows for Pi.
If I understand what I read there and elsewhere correctly, a Windows Pi is not a developer's computer. A developer works on some other computer, creates a Windows Pi executable and transfers it to a Pi to run. When the target is a 300MHz MIPS, OK. When the target is a single core 700MHz ARMv6, and the application is big, then OK, but for 4x900MHz ARMv7: Why?
I think I found a why. If I understood correctly, Windows has some alternative to NFS with some weird authentication protocol that only Windows understands (excuse: I am a penguin, so really do not know what the Microsofties were talking about). Now you can connect to this stuff with a $35 pi instead of the $299 box Microsoft needed to run their software.
The only really consistent thing I have seen about Windows for Pi is that Microsoft will make an announcement real soon now. I get the impression Marketing do not know anything yet and Microsoft's techies found out about the project from The Register.
Last time I saw Microsoft's numbers, Microsoft scored 52/year (patch Tuesday) and the Linux numbers included every package multiplied by the number of distributions. I admit that was a long time ago and things have changed - these days Microsoft do not update every Tuesday.
Abiword on Pi.1 worked, but you could see the screen updates. It was tollerable for trivial work, but I used something bigger unless someone else was using it. Abiword and LibreOffice have been running fine on my 4 core 1GHz ARMv7 box since early 2012 (tripple the cost of a Pi). For the vast majority of office work, this is fine. The NIC is '1GHz', but ⅓GHz would be more honest when you look at the memory bandwidth. It is attached to a 100MHz switch. It was using an SSD connected by USB until the eSATA cable arrived. Lack of 100MHz network and lack of SATA are not noticable issues for office work or for a ripped DVD client / server. On a good day, two users can watch 1080p over a 100MHz net.
Linux users have been able to set up an 'austerity computer' for years. Sometimes they are even for sale directly to computer somewhat-literates. In the past, such computers vanished half way through an exhibition and an underpowered windows box appeared after a couple of months later for twice the price.
I have no idea what hardware Windows 10 + MS Office really requires but I would expect a Pi.2 to be a perfectly capable Libreoffice box. I have had problems using a Pi.1 as a print server, and would make sure I had a plan B before trying a Pi.2 as a print server.
Back in the day, I could compile for i386, i386+i387, i486(dx), i486sx, i487, Pentium, Pentium MMX and the AMD/VIA variations. I could also compile for any subset of them at the cost of reducing performance. Debian has ports for two variations to keep the size of the repositories sane. (Gentoo supports your exact hardware by downloading the source code and compiling it). ARM is just about leave a period of diversity that used to afflict x86. Debian has two ARM ports: armel (much older hardware than Pi1) or armhf (a little too modern for Pi1). Rasbian (a Debian port specifically for Pi1) is needed to get reasonable performance out of an old Pi. A Pi2 should be able to use the standard armhf port without a significant performance loss.
ARM SoCs have something resembling a BIOS: an on-chip ROM than can just about read a boot loader from SDHC or SATA or whatever device exists. At this point, things get unpleasant. Each ROM works differently, and the documentation is usually secret, missing, non-existent, badly translated and full of errors. Where there is a standard, it is often outright hostile to prevent you from installing Linux on a landfill RT tablet. The fix is called Das U-Boot. If there is a branch for your SoC, Das U-Boot can be compiled and installed on flash where and how your particular boot ROM expects it.
The next disaster is that every SoC has a different mix of on chip components, and usually far more than can access the outside world because there are not enough pins on the chip. The actual hardware available depends on what verison of the PCB the SoC is soldered onto. On modern x86 systems, most devices can be found from their PCI id, or by hoping the BIOS will tell you (gray beards can regale with tales of ISA and plug 'n pray). On ARM, you can create flattened device trees (lists of available hardware) when you compile the kernel.
Getting an ARM to boot requires partitioning some flash device the way the boot ROM expects, installing the correct branch of U-Boot where the ROM expects it, and pointing U-Boot at the right FDT and you distribution's partition. The kernel itself can be the standard one from your distribution.
The current system may be vile, but it could easily be worse. When BIOS was the 'standard', some manufacturers implemented it so badly that the Linux BIOS project was created to replace it (that project suffered from all the horrors we currently see with ARM). There are standard boot sequences for ARM, usually designed to lock you into Chrome, RT, Android, Winphone or whatever you want to replace the day you get the device.
The real solution is the same as it has always been: research the install process and state of hardware support before you make a purchase decision.
If you do not decrypt an encrypted file when the police tell you to, you get five years in prison. A paedo or terrorists would get far worse if they did. The fun comes when you email a file full of random numbers to Theresa May called 'plans_for_wmd.txt.gpg'. How is she going to decrypt it?
Same for the house of commons.
I tried your credibility link to a blog written by Anonymous ExNokian. AXN has convinced me that TA is being honest. The link complains about three graphs, and has a link promising to tell me what is wrong with the first graph and probably has similar links to the next two graphs, but I never got that far.
The first graph illustrates Elop's promise to convert Nokia's smart phone customers into Winphone customers. TA says his graph is a tidied up version of one from slashgear. AXN points out that TA missed out 'not a prediction' from slashgear's graph, and changed the vertical axis from net sales mix to revenue mix. Both are careless/naughty mistakes, but both graphs clearly represent Elop's promise to convert all of Nokia's smart phone customers into Winphone customers. TA and AXN both have graphs for what really happened.
AXN's graph shows the proportion of smart and dumb phones sold by Nokia. It looks a lot like the original 'not a prediction' graph, and makes it look like the 'not a prediction' graph was an accurate prediction.
TA's graph includes a white region at the top that grows with time. The area represents Nokia's smart customers buying Android/iOS. Elop retained 3 out of 20 smart phone customers.
AXN's graph is at best completely useless for checking Elop's promise to convert all of Nokia's smart phone customers into Winphone customers. What you need is something like TA's graph. Of course AXN disagrees with TA's numbers, and promises to explain in a link to another of his posts. I got part of the way through that, and found more rants, and promises to substantiate them in other articles.
Where AXN does have numbers, they are numbers shipped, not numbers sold and they are often for only one region and not the whole world. Personally, I disagree with TA's assumption that the difference between 'units shipped' and 'units activated' represents unsold Lumia's in boxes with the retailers. I always thought Lumias were shipped out, shipped back and shipped to the region where Microsoft wanted to quote a large number of units shipped (opinion - no evidence).
I would agree that TA rabidly despises Elop, and by association, does not like Microsoft at all. Where did this hatred come from? TA makes it quite clear that he is unhappy about Nokia's loss of profits, loss of unit sales and loss of market share, all of which happened while Elop was in charge. TA places almost all of the blame on Elop (Ballmer gets a some blame too). Given TA's feelings on the matter, I can understand why you might question his figures, and conclusions, but if you want to discredit TA, AXN is the wrong choice.