It's not free, it's taxpayer funded. The lecturers don't work for free and the universities don't run for free. They are paid, by the government, in proportion to the number of students that they have plus their research grants. Depending on the university, the research grants pay somewhere between 20 and 100% of any given lecturer's salary. For universities without such a strong research reputation, the money from tuition can be a significant amount of a department's total budget. If your choice is either lower standards of make a lecturer redundant (which weakens your ability to get research grants and lowers your teaching quality, which makes it harder to attract students), what do you do?
The writers forgot about that on several occasions. For example, in both Redemption (the episode immediately following Orac) and Ultraworld, ORAC manages to control alien computers that would not use these chips.
While I agree that protesting at any funeral is a bit rude, there is a big difference between protesting at a private ceremony intended for family and protesting at a public procession intended to portray the person as a positive social force.
And threads alone aren't that interesting a metric anyway. A single UltraSPARC T3 has 16 cores, 128 threads. They work in up to 4-socket configurations, so that's 512 per board. Four boards will give you 2048 threads. The 4-socket boards come in 5U boxes, so that's 9 per 47U rack, giving 4608 threads. A high-end GPU has 512 hardware threads and you can stick 4 of them in a single desktop and get your 2048 threads. How does a box full of GPUs or a rack full of T3s compare to this system? The thread counts give you absolutely no information to determine that.
Do you remember the launch of USB? It was released in January 1996, and it wasn't until Windows 95 OSR 2.1 in August 1997 that it was even supported by Windows 95, and then only if you bought a new computer (OSR 2.1 was only available to OEMs, it wasn't an update to existing computers that shipped with USB). It wasn't until around 1998 that there started to be a number of USB mice and keyboards available, and most of those were in translucent plastic to go with the iMacs that Apple had released without legacy I/O - most PC users didn't buy them because they were more expensive than PS/2 or serial ones (lots of Apple users did because the original iMac shipped with the worst-designed mouse in human history). In 2000, I got my first USB-only device (I also owned a joypad that could connect to either USB or a game port, and a keyboard and mouse that were USB or PS/2), and it was a gaming mouse. It wasn't until around 2001 that USB stuff was really ubiquitous, and even in 2001/2002 I remember gamers complaining that you should always use PS/2 mice because of USB latency. It took 5 years for USB to become a ubiquitous connector, and it was starting at the low end, with cheap devices with short(ish) lifespans as the primary use cases. The early devices came with a hefty early-adopter premium.
With Intel pushing it, I expect Thunderbolt will have the same adoption pattern: it will appear in chipsets and become a standard connector on all Intel motherboards before it really becomes widely used, and prices won't drop until economies of scale reach the point where they're selling at least tens and probably hundreds of thousands of Thunderbolt peripherals.
That depends on the desktop. Lots of popular cases are small and don't have the space for any more expansion cards. It also depends on the user: how many typical users would prefer to open up the case and install something when they could just plug something into the back?
Two: it requires a special graphics card and some hacky loop-back connector.
Really? The port just exposes some PCIe lanes and the graphics card plugs into the PCIe bus. I thought the standard way of doing it was to stick a PCIe-PCIe bridge chip on the graphics card reserve four of the 16 lanes that the slot provides for the Thunderbolt interface.
Back in 2003, I had a PowerBook that came with FireWire 400 and FireWire 800. I had two LaCie BigDisks, which had USB, FireWire 400 and two FireWire 800 ports. I could connect one to the other and then daisy chain them to the FW800 port. USB 2 was the bottleneck when using a single drive. With FW400, the bottleneck became the drive speed. With FW800, I could have both drives running at their full speed. This was amazing for video editing on the laptop. I had the source DV files on one disk and the scratch disk on the other, and compositing passes became CPU-bound, rather than I/O bound (as they were to a painful degree on the internal disk drive). Back then, half decent video cameras were also FW400-only, and with an adaptor cable I was able to plug the camera into the end of the chain (I only did this once to see if it worked, because I had another FW400 port for it in normal use) and it worked perfectly, streaming the data to a disk on the same chain. On that machine, I used the FireWire ports a lot more than the USB ports - I initially used them for a keyboard and mouse, but later replaced those with Bluetooth.
It wasn't a separate technology required for hard drives, it was a separate technology required for things that needed either high bandwidth or isochronous transfer. Unfortunately, that basically meant hard disks and video cameras. It failed because USB 2 was only slightly worse than FW400, and you needed USB (and so got USB2 effectively for free) for keyboards and other cheap devices. A FireWire mouse or keyboard would have needed the $1 host controller chip, whereas a USB one could use much simpler hardware, so would always be cheaper. You might have got less lag with a FireWire input device, but I doubt anyone human would have noticed the difference and been willing to pay the price (although you might have been able to sell them to gamers). USB was in the Intel chipsets, FireWire was an extra IC on the motherboard, so it added cost. It's hard to make people pay extra to go from good enough to better. The proliferation of USB2 ports and fast CPUs made this even worse: the issues with USB2 largely go away when you have one port per device and enough cores that you can effectively designate one to running the USB devices and still not notice.
It depends a lot on what you're doing. What do people do on Macs Pro that are that processor intensive these days? I do a fair bit of compute-bound work (big compile jobs, FPGA place-and-route, and so on), but I do it sitting in front of my MacBook Pro and ssh'd into some big servers each with 32 cores and 256GB of RAM. The current Mac Pro maxes out at 12 cores and 64GB of RAM, and for that price costs more than we paid for the servers (which can be easily shared between people, as we all tend to have bursty workloads). Interactive video / image processing tasks tend to be offloaded to the GPU these days and there isn't such a huge difference between the desktop and mobile parts there. Bigger jobs are best run on a load of cheap commodity hardware in a rack than on a workstation.
Android can't easily do this, because it uses discretionary access control stuff to protect applications, which means that if your SD card if FAT formatted then it can't impose any access restrictions because the filesystem doesn't provide per-user permissions. In contrast, iOS uses the MAC framework from SEBSD and so grants access to files per system call based on a policy that is changed at run time. An even cleaner mechanism would be to use Capsicum and grant access per file descriptor (they're effectively trying to emulate Capsicum on SEBSD at the moment, because they haven't imported Capsicum from FreeBSD yet).
Like who is going to just put a fish in their ear?
Anyone who read the books? On the other hand, getting the fish required first of all that you remember to pick up the junkmail at the very start of the game (and don't forget to feed the dog soon after, or it will eat your space fleet two hours further into the game) and then arrange it in a complex way to make the vending machine work without the robot catching the fish. The hint book entry for this was two pages long.
The compile cycle is annoying, but viewing the document that you're writing in semantic markup is very valuable if you want to write quickly. I write using a set of semantic annotations that are implemented in two ways. One is a set of LaTeX macros. The other is a set of classes in a program that generates XHTML. When I am writing (which I do in vim), I am only looking at the semantic markup and am only thinking about the presentation. I'll often write an entire chapter before doing the first compile pass. I'll then do some tweaking, but this happens much later on. The important thing is the separation of the writing, typesetting, and editing tasks.
TeX does stupid things with floats. For example, a (simple) style rule most publishers use is that a figure should appear after the first reference to it. Now, try encoding that in [La]TeX.
Does visio not have some kind of scripting interfaces? I used OmniGraffle a lot on OS X with LaTeX and had a little graffle2pdf AppleScript. The Makefile for my PhD thesis used this so if the.graffle is ever older than the accompanying.pdf then it automatically fired up OmniGraffle and did the export.
I've switched to Beamer for my presentations, but for OS X there was a nice system service that sent the selected text through pdflatex and returned the resulting PDF. It also did some tricks to embed the TeX source in the PDF, so there was an inverse operation. That made formulae in Keynote presentations easy: just type the TeX version, hit a keyboard shortcut, and you get it replaced by the PDF. Hit another shortcut and you can edit it again.
something that would be unlikely to happen with alcohol
It's also unlikely to happen with marijuana. It's even unlikely to happen with LSD, although probably more likely. Unfortunately, unlike tobacco and alcohol, there is no requirement to put warning labels on marijuana when it's sold. It is also difficult to do detailed studies on the effects of the drug, and it is not possible to go to a doctor and be tested for the latent conditions that can be triggered by certain chemicals, if those chemicals happen to be illegal.
Yup, the power boxes model works quite well in general. It integrates nicely with the UI for things that are files. It's a bit more complicated to get right, for example, for the address book. For example, I want a QR code reader to be able to create new vCards and pass them to the address book, but it shouldn't be able to see my contacts. The responsibility for merging the new vCard into the address book should belong to the Address Book application. Similarly, I may want other applications to be able to add calendar entries, but not see all of my appointments.
I gave up on Ars when Jon Stokes left, as I realised that 95% of the articles they published that I actually read and then didn't come away thinking that the writer was an ignorant hack were written by him. They stayed in my RSS feeds for about a month without producing a single article worth reading, and then I removed them. Don't be too hard on their writers though. I responded to their last call for freelancers, but didn't end up doing anything for them because they pay somewhere between a quarter and a half of the market rate, so it's not surprising that they have difficulty getting competent people.
Note that he said the same use case, not specifications. I have a nice Android tablet (Asus TransformerPad Infinity), and a few of my friends have Android tablets and iPads and pretty much everything we do on them can also be done on my HP TouchPad. There are very few things that currently tax the CPU in tablets, especially now that Flash doesn't work on modern Android systems. Some games do, but most web browsing, email, and so on tasks don't come close to taxing them.
In 1993, if you were using a Microsoft product it would either have been DOS or Windows 3.x. In DOS, PrtScn wrote the current terminal contents to lpt1. In Windows 3.x, it copied the current screen contents to the clipboard.
It probably helps that Unix was developed from the beginning as a multi-user system, where you had to think about not letting one user trample all over another, whereas Windows started out as a single-user system where users could only f*ck up their own stuff if they did something stupid. The whole multi-user security thing was bolted on afterwards.
This is not true of Windows NT, which started with fine-grained access control to every kernel object (files, but also network interfaces and IPC primitives), when UNIX only had coarse-grained user-group-everyone permissions, and then only on things that showed up in the filesystem namespace (which, contrary to popular belief, is not everything).
Windows users say: "Of course there aren't any Linux viruses. It's too small a target." Linux users say: "Of course there aren't any Linux viruses, it has far better security." The truth probably lies somewhere in the middle.
You need three things to have viruses in the wild. One is to have some available exploits. Linux and Windows both have an abundance of these. The next is to have an installed base large enough to be worth targeting. Windows has this, Linux also does on the server, which is why you periodically see things like PHP worms that hijack web servers and use them to infect other servers and clients. Finally, you need sufficient density of installation that the infection can spread. A Linux machine on a Windows network is typically safe, because the only things that will try to infect it will be trying Windows exploits. The converse is also true. When you factor in the web and email, you have some additional attack vectors, but you still need the person running the evil web server to bother writing the Linux version of the exploit, and it often isn't worth it for an extra 1% infection rate. It's a lot more common to see, for example, a piece of malware that contains code for attacking a web browser running on Windows and a web server running on Linux (or any *NIX, as often the server-side exploits are in the web server and don't rely on kernel vulnerabilities). The client attacks are more likely to be OS-specific than the server attacks, as web browsers implement sandboxing using different OS primitives.
The main issue with sandboxing in the style of Android is that it doesn't actually work. Any application that needs to be able to modify documents has the 'access SD card' permission, and that gives it access to all other documents. There is an inherent problem with this approach, in that you either give such coarse-grained permissions that they leak like a sieve, or you have such fine-grained ones that 99% of users don't understand them.
Currently, the most common way of deploying Android malware is to find a fairly popular app in the Market (sorry, Play Store), download it, add the trojan, reupload it with a different name and a price of zero, download it a few dozen times and rate it 5 stars, and then wait for users to install it. You can usually even keep the same permissions as the original, as it is pretty much the norm for applications to ask for far more permissions than they actually need.
If you did this on Windows, you'd have screensavers saying that they needed full access to C: to run, and users would just click 'yes, show me the kittens'.
No, economics predicts that two factors will affect the transactions. One is that, as a commodity with a long-term trend that goes upwards, the rational strategy is to hold. The other is that, as a highly volatile commodity, it is possible to make a lot of money by buying at the peaks and selling at the troughs. Currently, Bitcoin is sufficiently volatile that the effects of the latter outweigh the effects of the former: it is possible to make a lot more money from the noise than from the growth. The same is true of any highly volatile stock with a long-term growth trend, for example Apple stocks over much of the last decade.
Slashdot Mirror
It's not free, it's taxpayer funded. The lecturers don't work for free and the universities don't run for free. They are paid, by the government, in proportion to the number of students that they have plus their research grants. Depending on the university, the research grants pay somewhere between 20 and 100% of any given lecturer's salary. For universities without such a strong research reputation, the money from tuition can be a significant amount of a department's total budget. If your choice is either lower standards of make a lecturer redundant (which weakens your ability to get research grants and lowers your teaching quality, which makes it harder to attract students), what do you do?
And she could introduce a Poll Tax, sorry, Community Charge, sorry, Flat Tax, and solve all of the country's problems!
I've been using nVidia Linux drivers since, 1992 (*gasp*)
Gasp indeed. I'd be very impressed by this, given that nVidia was only founded in 1993 and released its first graphics card in 1995.
The writers forgot about that on several occasions. For example, in both Redemption (the episode immediately following Orac) and Ultraworld, ORAC manages to control alien computers that would not use these chips.
While I agree that protesting at any funeral is a bit rude, there is a big difference between protesting at a private ceremony intended for family and protesting at a public procession intended to portray the person as a positive social force.
And threads alone aren't that interesting a metric anyway. A single UltraSPARC T3 has 16 cores, 128 threads. They work in up to 4-socket configurations, so that's 512 per board. Four boards will give you 2048 threads. The 4-socket boards come in 5U boxes, so that's 9 per 47U rack, giving 4608 threads. A high-end GPU has 512 hardware threads and you can stick 4 of them in a single desktop and get your 2048 threads. How does a box full of GPUs or a rack full of T3s compare to this system? The thread counts give you absolutely no information to determine that.
Do you remember the launch of USB? It was released in January 1996, and it wasn't until Windows 95 OSR 2.1 in August 1997 that it was even supported by Windows 95, and then only if you bought a new computer (OSR 2.1 was only available to OEMs, it wasn't an update to existing computers that shipped with USB). It wasn't until around 1998 that there started to be a number of USB mice and keyboards available, and most of those were in translucent plastic to go with the iMacs that Apple had released without legacy I/O - most PC users didn't buy them because they were more expensive than PS/2 or serial ones (lots of Apple users did because the original iMac shipped with the worst-designed mouse in human history). In 2000, I got my first USB-only device (I also owned a joypad that could connect to either USB or a game port, and a keyboard and mouse that were USB or PS/2), and it was a gaming mouse. It wasn't until around 2001 that USB stuff was really ubiquitous, and even in 2001/2002 I remember gamers complaining that you should always use PS/2 mice because of USB latency. It took 5 years for USB to become a ubiquitous connector, and it was starting at the low end, with cheap devices with short(ish) lifespans as the primary use cases. The early devices came with a hefty early-adopter premium.
With Intel pushing it, I expect Thunderbolt will have the same adoption pattern: it will appear in chipsets and become a standard connector on all Intel motherboards before it really becomes widely used, and prices won't drop until economies of scale reach the point where they're selling at least tens and probably hundreds of thousands of Thunderbolt peripherals.
One: they have slots and don't really need it.
That depends on the desktop. Lots of popular cases are small and don't have the space for any more expansion cards. It also depends on the user: how many typical users would prefer to open up the case and install something when they could just plug something into the back?
Two: it requires a special graphics card and some hacky loop-back connector.
Really? The port just exposes some PCIe lanes and the graphics card plugs into the PCIe bus. I thought the standard way of doing it was to stick a PCIe-PCIe bridge chip on the graphics card reserve four of the 16 lanes that the slot provides for the Thunderbolt interface.
Back in 2003, I had a PowerBook that came with FireWire 400 and FireWire 800. I had two LaCie BigDisks, which had USB, FireWire 400 and two FireWire 800 ports. I could connect one to the other and then daisy chain them to the FW800 port. USB 2 was the bottleneck when using a single drive. With FW400, the bottleneck became the drive speed. With FW800, I could have both drives running at their full speed. This was amazing for video editing on the laptop. I had the source DV files on one disk and the scratch disk on the other, and compositing passes became CPU-bound, rather than I/O bound (as they were to a painful degree on the internal disk drive). Back then, half decent video cameras were also FW400-only, and with an adaptor cable I was able to plug the camera into the end of the chain (I only did this once to see if it worked, because I had another FW400 port for it in normal use) and it worked perfectly, streaming the data to a disk on the same chain. On that machine, I used the FireWire ports a lot more than the USB ports - I initially used them for a keyboard and mouse, but later replaced those with Bluetooth.
It wasn't a separate technology required for hard drives, it was a separate technology required for things that needed either high bandwidth or isochronous transfer. Unfortunately, that basically meant hard disks and video cameras. It failed because USB 2 was only slightly worse than FW400, and you needed USB (and so got USB2 effectively for free) for keyboards and other cheap devices. A FireWire mouse or keyboard would have needed the $1 host controller chip, whereas a USB one could use much simpler hardware, so would always be cheaper. You might have got less lag with a FireWire input device, but I doubt anyone human would have noticed the difference and been willing to pay the price (although you might have been able to sell them to gamers). USB was in the Intel chipsets, FireWire was an extra IC on the motherboard, so it added cost. It's hard to make people pay extra to go from good enough to better. The proliferation of USB2 ports and fast CPUs made this even worse: the issues with USB2 largely go away when you have one port per device and enough cores that you can effectively designate one to running the USB devices and still not notice.
It depends a lot on what you're doing. What do people do on Macs Pro that are that processor intensive these days? I do a fair bit of compute-bound work (big compile jobs, FPGA place-and-route, and so on), but I do it sitting in front of my MacBook Pro and ssh'd into some big servers each with 32 cores and 256GB of RAM. The current Mac Pro maxes out at 12 cores and 64GB of RAM, and for that price costs more than we paid for the servers (which can be easily shared between people, as we all tend to have bursty workloads). Interactive video / image processing tasks tend to be offloaded to the GPU these days and there isn't such a huge difference between the desktop and mobile parts there. Bigger jobs are best run on a load of cheap commodity hardware in a rack than on a workstation.
Android can't easily do this, because it uses discretionary access control stuff to protect applications, which means that if your SD card if FAT formatted then it can't impose any access restrictions because the filesystem doesn't provide per-user permissions. In contrast, iOS uses the MAC framework from SEBSD and so grants access to files per system call based on a policy that is changed at run time. An even cleaner mechanism would be to use Capsicum and grant access per file descriptor (they're effectively trying to emulate Capsicum on SEBSD at the moment, because they haven't imported Capsicum from FreeBSD yet).
Like who is going to just put a fish in their ear?
Anyone who read the books? On the other hand, getting the fish required first of all that you remember to pick up the junkmail at the very start of the game (and don't forget to feed the dog soon after, or it will eat your space fleet two hours further into the game) and then arrange it in a complex way to make the vending machine work without the robot catching the fish. The hint book entry for this was two pages long.
LaTeXiT is a newer tool. The one I was using is no longer developed and was PowerPC-only.
The compile cycle is annoying, but viewing the document that you're writing in semantic markup is very valuable if you want to write quickly. I write using a set of semantic annotations that are implemented in two ways. One is a set of LaTeX macros. The other is a set of classes in a program that generates XHTML. When I am writing (which I do in vim), I am only looking at the semantic markup and am only thinking about the presentation. I'll often write an entire chapter before doing the first compile pass. I'll then do some tweaking, but this happens much later on. The important thing is the separation of the writing, typesetting, and editing tasks.
TeX does stupid things with floats. For example, a (simple) style rule most publishers use is that a figure should appear after the first reference to it. Now, try encoding that in [La]TeX.
Does visio not have some kind of scripting interfaces? I used OmniGraffle a lot on OS X with LaTeX and had a little graffle2pdf AppleScript. The Makefile for my PhD thesis used this so if the .graffle is ever older than the accompanying .pdf then it automatically fired up OmniGraffle and did the export.
Using PP for math will make you go blind
I've switched to Beamer for my presentations, but for OS X there was a nice system service that sent the selected text through pdflatex and returned the resulting PDF. It also did some tricks to embed the TeX source in the PDF, so there was an inverse operation. That made formulae in Keynote presentations easy: just type the TeX version, hit a keyboard shortcut, and you get it replaced by the PDF. Hit another shortcut and you can edit it again.
something that would be unlikely to happen with alcohol
It's also unlikely to happen with marijuana. It's even unlikely to happen with LSD, although probably more likely. Unfortunately, unlike tobacco and alcohol, there is no requirement to put warning labels on marijuana when it's sold. It is also difficult to do detailed studies on the effects of the drug, and it is not possible to go to a doctor and be tested for the latent conditions that can be triggered by certain chemicals, if those chemicals happen to be illegal.
Yup, the power boxes model works quite well in general. It integrates nicely with the UI for things that are files. It's a bit more complicated to get right, for example, for the address book. For example, I want a QR code reader to be able to create new vCards and pass them to the address book, but it shouldn't be able to see my contacts. The responsibility for merging the new vCard into the address book should belong to the Address Book application. Similarly, I may want other applications to be able to add calendar entries, but not see all of my appointments.
I gave up on Ars when Jon Stokes left, as I realised that 95% of the articles they published that I actually read and then didn't come away thinking that the writer was an ignorant hack were written by him. They stayed in my RSS feeds for about a month without producing a single article worth reading, and then I removed them. Don't be too hard on their writers though. I responded to their last call for freelancers, but didn't end up doing anything for them because they pay somewhere between a quarter and a half of the market rate, so it's not surprising that they have difficulty getting competent people.
Note that he said the same use case, not specifications. I have a nice Android tablet (Asus TransformerPad Infinity), and a few of my friends have Android tablets and iPads and pretty much everything we do on them can also be done on my HP TouchPad. There are very few things that currently tax the CPU in tablets, especially now that Flash doesn't work on modern Android systems. Some games do, but most web browsing, email, and so on tasks don't come close to taxing them.
In 1993, if you were using a Microsoft product it would either have been DOS or Windows 3.x. In DOS, PrtScn wrote the current terminal contents to lpt1. In Windows 3.x, it copied the current screen contents to the clipboard.
It probably helps that Unix was developed from the beginning as a multi-user system, where you had to think about not letting one user trample all over another, whereas Windows started out as a single-user system where users could only f*ck up their own stuff if they did something stupid. The whole multi-user security thing was bolted on afterwards.
This is not true of Windows NT, which started with fine-grained access control to every kernel object (files, but also network interfaces and IPC primitives), when UNIX only had coarse-grained user-group-everyone permissions, and then only on things that showed up in the filesystem namespace (which, contrary to popular belief, is not everything).
Windows users say: "Of course there aren't any Linux viruses. It's too small a target." Linux users say: "Of course there aren't any Linux viruses, it has far better security." The truth probably lies somewhere in the middle.
You need three things to have viruses in the wild. One is to have some available exploits. Linux and Windows both have an abundance of these. The next is to have an installed base large enough to be worth targeting. Windows has this, Linux also does on the server, which is why you periodically see things like PHP worms that hijack web servers and use them to infect other servers and clients. Finally, you need sufficient density of installation that the infection can spread. A Linux machine on a Windows network is typically safe, because the only things that will try to infect it will be trying Windows exploits. The converse is also true. When you factor in the web and email, you have some additional attack vectors, but you still need the person running the evil web server to bother writing the Linux version of the exploit, and it often isn't worth it for an extra 1% infection rate. It's a lot more common to see, for example, a piece of malware that contains code for attacking a web browser running on Windows and a web server running on Linux (or any *NIX, as often the server-side exploits are in the web server and don't rely on kernel vulnerabilities). The client attacks are more likely to be OS-specific than the server attacks, as web browsers implement sandboxing using different OS primitives.
The main issue with sandboxing in the style of Android is that it doesn't actually work. Any application that needs to be able to modify documents has the 'access SD card' permission, and that gives it access to all other documents. There is an inherent problem with this approach, in that you either give such coarse-grained permissions that they leak like a sieve, or you have such fine-grained ones that 99% of users don't understand them.
Currently, the most common way of deploying Android malware is to find a fairly popular app in the Market (sorry, Play Store), download it, add the trojan, reupload it with a different name and a price of zero, download it a few dozen times and rate it 5 stars, and then wait for users to install it. You can usually even keep the same permissions as the original, as it is pretty much the norm for applications to ask for far more permissions than they actually need.
If you did this on Windows, you'd have screensavers saying that they needed full access to C: to run, and users would just click 'yes, show me the kittens'.
No, economics predicts that two factors will affect the transactions. One is that, as a commodity with a long-term trend that goes upwards, the rational strategy is to hold. The other is that, as a highly volatile commodity, it is possible to make a lot of money by buying at the peaks and selling at the troughs. Currently, Bitcoin is sufficiently volatile that the effects of the latter outweigh the effects of the former: it is possible to make a lot more money from the noise than from the growth. The same is true of any highly volatile stock with a long-term growth trend, for example Apple stocks over much of the last decade.