Turning the computer inside out
on
A New Kind of OS
·
· Score: 4, Interesting
Almost since the inception of computers and then later modern OS design we've been trapped in a paradigm that although mirroring some aspects of the real world (the desktop, tools, etc), is quite backwards from other aspects. I think it is time we ditched some of these decades old concepts. For one the concept of an "application" has to go. It's an outdated and locks us down and restricts what we can do. See it's not about the applications; it's about the data. The data is the most important thing. Data should not be imprisoned in an application or even a series of compatible applications. Rather than the application being the focus of our OS and UIs, we should make the data, or the "document" be the focus. Instead of applications we have smaller, simpler, tools that can be applied to the documents (data objects or whatever). Common tools can work equally well on like data objects no matter where they reside. A spell checker would spell check anything that is text. A pen could draw on anything that is a drawable (a surface of some kind). If you needed a better pen, you'd buy a better pen that would work on the same surfaces as the old one (but in a better way perhaps). Everything would be document-centric with the concept of, perhaps, tool palettes or something. But it would be very modular and loosely coupled. The irony of loose coupling is that it could lead to the integration of widely differing sets of tools. For years Microsoft has tought us that to have good integration between the various tasks (word processing, spreadsheets, etc) we need a tightly intergrated application. This is false. We really need just open document objects that can support a variety of types of data and the tools to work on them. The OS becomes the app and *everything* is then integrated, but in a more open and extensible way. Of course this dramatic shift would lead to the demise of many major software houses until they can learn to adapt to the new way of doing things. But in the end the OS gets out of the way and lets us *work*.
If some of these concepts sound familiar, it is because they are not new. Apple and IBM once talked about this in their Taliget (http://en.wikipedia.org/wiki/Taligent) project which died. Unfortunately while we talk about technologies like OOP, they really haven't moved very much beyond languages. OSs are modular and even object-oriented to a degree, but they haven't quite arrived at the things I describe yet. Having the KDE libraries being object-oriented and manipulatable over RPC and DCOP is a step towards a possible document-centric future.
Because the free download actually costs more. Your time may be free, but mine certainly is not.
As for being dodgy, if you're only worried getting ripped off by some shady Russian person, this is not going to happen at allofmp3.com. Of the other hand, who knows who ends up with the money. But you will get your "product."
Have you tried allofmp3.com? They may have a lot of the stuff (you already own) in flac or mp3 or ogg. I find it more convenient to pay them for the ogg files than to rip my own CDs. Anyone remember the old mp3.com service where you'd insert your music CD and then they'd let you download the mp3 version for free? Didn't last long.
I'd start by learning about the specialized distros for things like linksys. For example, http://www.openwrt.org/ . Google also has lots of things when last I checked. I'm currently running a WRT54GS (older non-broken version) with openwrt. I have a custom bash script that generates the iptables on boot, I run an openvpn client to bridge my home network with another network, and I have iptables rules that allow transparent http filtering and proxying through an external machine (the linksys isn't beefy enough to run squid, privoxy, dansguardian).
I just tried this using a restricted user and it said, simply that "Access is Denied" when I tried to run it. So I don't think this is really as big a hole as you suggest, although power users may be able to get slightly elevated priviledges this way. But certainly not the end of the world.
While such things do influence people for the worse, this is not the root cause. The root cause of Islamic facism and terrorism is due to a couple of things. First we need to acknowledge that the largest Islamic country in the world is Indonesia which does not seem to produce large numbers of terrorists and suicide bombers. Therefore the root cause cannot be Islam itself. Rather the root cause has more to do with the fact that the Arab world (not necessarily the Islamic world) is, in terms of thinking, almost a century behind the rest of the world. Repressive regimes (which we prop up) create essentially a gang-neighborhood environment. In fact terrorism is caused by the same things that the inner-city gang problems are caused by. Instead of drugs and rap music we have extreme religious points of view that people turn to to get a sense of belonging and value, even though these things are ultimately destructive and cause a chain of violence. All of these factors combine to cause Arabs in particular to feel that they are victims. They are victims of Israel, victims of the US, victims of the west. Even if the US was not in Iraq, and if Israel did not exist, many would still feel this way. It's really a way of avoiding responsibility for one's own actions (you caused me to blow myself up in your city because you're oppressing me).
So the combination of having one's society be so far behind the rest of the world and the feeling of victimhood are the real root causes. Sure our foreign policy doesn't help things, but on the other hand I don't see that changing anytime soon. There is no ethical way the US can abandon Israel. There is, however, lots we can do to build good will in the Arab world (and no, abandoning Iraq will *not* foster good will on the part of the Arab world).
Nope. The Saturn processor is a 9-register cpu (each register is 64-bits) coupled with an outrageous 4-bit data bus. RPL and the stack are created in software.
However, if the compiler that they use to compile the binary versions of the Kernel is licenced under the GPL v3, then wouldn't the Kernel also need to be licenced under the GPL v3?
Umm, no. The output of GCC has never been under any license; only the code of the compiler suite itself. If GPLv3 changes this, then no wonder Linus is so opposed to it. I doubt that, even under the GPLv3, GCC's output will be under any license.
I guess you've never had any experience with the linksys family of wireless routers. Until recently the entire WRT series ran Linux. And yes, cups can fit in the ROM of several models (8 MB FLASH only). The WRT54GSL, though, has a usb port and enough flash to run quite a bit of stuff. Samba, cups, even a small proxy server.
There are many laser printers under $300. Samsung makes a few fairly nice ones. But networkable, that's another story. Probably your best bet is a USB printer supported by linux and then stick CUPs on a Linksys 54GSL or whatever it's called (the wireless router with USB support).
Although toner is very expensive, you can get a decent full-color laser for about $300-600. The HP Laserjet 2600n. See http://www.nextag.com/hp-2600n/search-html . The n designation means it comes with jetdirect too. The only downside is the printer language is not PCL or PS, but there is a CUPS driver for linux available. The extra money for color may be worth the extra couple of hundred.
As for myself, I have an old Okidata 10ex LED printer that is parallel only. I use a USB-Parallel adapter and plug it into my linux box (cups server) with the USB. The linux box shares it to windows and linux clients. I recently upgraded the RAM to 32 MB, so it should be able ot handle anything I throw at it for years to come, even if I have to wait a while for the pages to spit out. Toner is separate from the drum, so it's dirt cheap to fill. I recently bought a new drum for it for $60. This printer has been one of my best computer investments.
Well except for the fact that extreme cold appears to increase the rate of decay (according to these scientists), which if it works for carbon-14 would show the sample to be significantly older than it really is, people already claim carbon-14 dating is inaccurate due to temperature fluxuation.
The thing is that carbon dating is really useless for the things that many people want to prove. Since the accuracy is plus or minus thousands of years, I don't see it proving anything about, say, a wood sample purported to be from Noah's ark. On the other hand it *does* show us that the earth is millions of years old generally.
That's really 10000 writes to a single cell on the flash chip. All flash memory has built-in algorithms to statistically spread out those writes over all the cells of the chip, so it's not like if you wrote a file 10000 times you'd have an instant failure. If you only write to on average, say 10% of the disk, then you wouldn't see a failure for probably 100000 writes. Given a 4 GB flash chip, the average write is probably just a few megabytes, so works out to even longer time between failures. Still not a huge lifetime, but not as short as you say.
The oven in my kitchen is a slightly evolved version of ovens 50 years ago. Heck even my microwave is just a slightly improved version of the first microwaves back in the late 70s. Some technologies have advanced, but others have just gotten slightly improved over the years, because as poor as they are, they are still the best. Consider the internal combustion engine. While computer controls have increased reliability and decreased emissions dramatically (and increased power and efficiency), the engine is basically unchanged from what it was in the 1920s. Overhead cams, mulitple valves per cylinder, super and turbo-chargers where around from the 20s. Even fuel injection has been here for years. Yet engines steadily improve in capability and power, even if we're using 100-year-old base technology. Similarly, until solid-state memory gets cheap enough and reliable enough, spinning magnetic disks will likely remain the main storage solution for years to come. And I think that's okay. I don't think it means we're somehow behind. No matter what technology comes along, I will always have a love/hate relationship with it.
"With a brand that said 'business machine' and an open architecture that invited third-party innovation, the IBM PC transformed the IT industry."
It seems we forget that when the PC was first introduced it was closed and proprietary. It wasn't until Compaq clean-room reverse-engineered the BIOS that the PC revolution really got started. If IBM had had their way the PC would have been locked down and controlled by IBM forever. Remember they used to call clones "IBM compatible." After Compaq started the cloning revolution, and Microsoft moved to make IBM-specific aspects of DOS irrelevant, not long after that IBM started to become less and less relevant. They no longer directed where the platform was going. By the i386, one could no longer talk about IBM-compatible. IBM tried to start over with a proprietary system (careful not to let cloning happen this time) withe Microchannel Architecure. Fortunately the market said, we'll stick with ISA, VESA-Local and PCI (even if MCA was superior at the time). Had IBM been successful in keeping the PC proprietary, I don't know what computers we would be using today. Maybe DEC alphas or Sparcstations. Or maybe we'd be paying $10000 a pop to IBM.
Apple never supported the open source version of darwin in any way beyond lip services, some server space, and releasing source packages in mostly unbuildable form. They took from many open source projects but returned precious little to the community. At the end of the day Apple does what immediately benefits Apple. It's sad, but it's likely the threat of hacking OS X to run on white box computers likely is the greatest reason for Apple to not release vital parts of the latest OS X source code. Yet this will still happen. In the meantime, Linux continues to grow and become better all the time. There just was no need for OpenDarwin without Aqua. If all you want is a unix-like OS to run servers, Linux suits the bill just fine.
I think the main problem for you is that SQL and relational databases were designed to store relational database data, not be a persistant store for objects. Indeed if you use a database as it was designed to be used, SQL is highly appropriate as a query language.
I think Microsoft is correct, as stated in another post, that ORM is inherently flawed. It may be the best thing we have, but shoe-horning database data into classes and vice versa is problematic. The time it takes to mess with Java's XML mapping files, for example, often exceeds the time it would take to write some decent stored procedures, use proper database views, and then do simple SQL queries directly. And I believe it is no less manageable or scaleable.
What you are looking for is really just a better persistant-object storage scheme. It doesn't matter what the backend is like, so long as your objects are accessible in an easy and logical fashion.
Sure but until the Mactel towers are release with pci-express slots in them, there's no real support in OSX86 for nvidia cards at all. In fact there aren't even any drivers available yet, except from a third-party site. Currently OSX86 only ships with intel and ATI drivers. This says nothing about the future, though.
The article is wrong about the flexibility of answer files and streamlined installations only applying to windows itself and not applications. We use http://unattended.sf.net/ and it works very well. 20 minutes later windows is install, updated, and all our standard packages installed including MS Office. In my opinion this beats the heck out of imaging XP using Ghost in terms of flexibility. We just boot into our installer image using PXE; no disks required at all. Ironically unattended uses a linux boot image with dosemu to bootstrap the Windows installer. Very slick. Unfortunately it looks like this whole process may have to be scrapped with Vista, sadly. We'll likely avoid vista for at least a full year after it is released, to give us time to integrate it properly into our operation (our servers are and will remain Linux based--can't wait to try Samba 4). The article is right about one thing, though. Setting up unattended installations is tedius.
As people have pointed out already, by the time you spec a machine similarly from Dell, the cost is very similar. For me the weight, battery life, and the linux compatibility are worth a $300 premium. People also like to complain about the expensive Apple laptops too until I show them that Apple is often cheaper than the equivalent Thinkpad or sometimes Dell.
Since the masses do think like you do, you are probably right, though. Most folks probably will take the lower-quality Dell over this laptop. This company will probably be a niche seller, but I hope they do well enough. Now that the 12" Apple laptop is no more (no the MacBook doesn't count), I may just abandon the sleek aluminum look and go for a maximum value laptop, such as this one. I could always run OS X on it...;)
Slashdot Mirror
Almost since the inception of computers and then later modern OS design we've been trapped in a paradigm that although mirroring some aspects of the real world (the desktop, tools, etc), is quite backwards from other aspects. I think it is time we ditched some of these decades old concepts. For one the concept of an "application" has to go. It's an outdated and locks us down and restricts what we can do. See it's not about the applications; it's about the data. The data is the most important thing. Data should not be imprisoned in an application or even a series of compatible applications. Rather than the application being the focus of our OS and UIs, we should make the data, or the "document" be the focus. Instead of applications we have smaller, simpler, tools that can be applied to the documents (data objects or whatever). Common tools can work equally well on like data objects no matter where they reside. A spell checker would spell check anything that is text. A pen could draw on anything that is a drawable (a surface of some kind). If you needed a better pen, you'd buy a better pen that would work on the same surfaces as the old one (but in a better way perhaps). Everything would be document-centric with the concept of, perhaps, tool palettes or something. But it would be very modular and loosely coupled. The irony of loose coupling is that it could lead to the integration of widely differing sets of tools. For years Microsoft has tought us that to have good integration between the various tasks (word processing, spreadsheets, etc) we need a tightly intergrated application. This is false. We really need just open document objects that can support a variety of types of data and the tools to work on them. The OS becomes the app and *everything* is then integrated, but in a more open and extensible way. Of course this dramatic shift would lead to the demise of many major software houses until they can learn to adapt to the new way of doing things. But in the end the OS gets out of the way and lets us *work*.
If some of these concepts sound familiar, it is because they are not new. Apple and IBM once talked about this in their Taliget (http://en.wikipedia.org/wiki/Taligent) project which died. Unfortunately while we talk about technologies like OOP, they really haven't moved very much beyond languages. OSs are modular and even object-oriented to a degree, but they haven't quite arrived at the things I describe yet. Having the KDE libraries being object-oriented and manipulatable over RPC and DCOP is a step towards a possible document-centric future.
Because the free download actually costs more. Your time may be free, but mine certainly is not.
As for being dodgy, if you're only worried getting ripped off by some shady Russian person, this is not going to happen at allofmp3.com. Of the other hand, who knows who ends up with the money. But you will get your "product."
Have you tried allofmp3.com? They may have a lot of the stuff (you already own) in flac or mp3 or ogg. I find it more convenient to pay them for the ogg files than to rip my own CDs. Anyone remember the old mp3.com service where you'd insert your music CD and then they'd let you download the mp3 version for free? Didn't last long.
I'd start by learning about the specialized distros for things like linksys. For example, http://www.openwrt.org/ . Google also has lots of things when last I checked. I'm currently running a WRT54GS (older non-broken version) with openwrt. I have a custom bash script that generates the iptables on boot, I run an openvpn client to bridge my home network with another network, and I have iptables rules that allow transparent http filtering and proxying through an external machine (the linksys isn't beefy enough to run squid, privoxy, dansguardian).
I just tried this using a restricted user and it said, simply that "Access is Denied" when I tried to run it. So I don't think this is really as big a hole as you suggest, although power users may be able to get slightly elevated priviledges this way. But certainly not the end of the world.
While such things do influence people for the worse, this is not the root cause. The root cause of Islamic facism and terrorism is due to a couple of things. First we need to acknowledge that the largest Islamic country in the world is Indonesia which does not seem to produce large numbers of terrorists and suicide bombers. Therefore the root cause cannot be Islam itself. Rather the root cause has more to do with the fact that the Arab world (not necessarily the Islamic world) is, in terms of thinking, almost a century behind the rest of the world. Repressive regimes (which we prop up) create essentially a gang-neighborhood environment. In fact terrorism is caused by the same things that the inner-city gang problems are caused by. Instead of drugs and rap music we have extreme religious points of view that people turn to to get a sense of belonging and value, even though these things are ultimately destructive and cause a chain of violence. All of these factors combine to cause Arabs in particular to feel that they are victims. They are victims of Israel, victims of the US, victims of the west. Even if the US was not in Iraq, and if Israel did not exist, many would still feel this way. It's really a way of avoiding responsibility for one's own actions (you caused me to blow myself up in your city because you're oppressing me).
So the combination of having one's society be so far behind the rest of the world and the feeling of victimhood are the real root causes. Sure our foreign policy doesn't help things, but on the other hand I don't see that changing anytime soon. There is no ethical way the US can abandon Israel. There is, however, lots we can do to build good will in the Arab world (and no, abandoning Iraq will *not* foster good will on the part of the Arab world).
No bounds checking. Instead, always use strncpy.
Nope. The Saturn processor is a 9-register cpu (each register is 64-bits) coupled with an outrageous 4-bit data bus. RPL and the stack are created in software.
o r)
http://en.wikipedia.org/wiki/Saturn_(microprocess
http://softpear.sf.net/
For those of us thinking about buying Parallels, what are the bugs and issues you found?
However, if the compiler that they use to compile the binary versions of the Kernel is licenced under the GPL v3, then wouldn't the Kernel also need to be licenced under the GPL v3?
Umm, no. The output of GCC has never been under any license; only the code of the compiler suite itself. If GPLv3 changes this, then no wonder Linus is so opposed to it. I doubt that, even under the GPLv3, GCC's output will be under any license.
I guess you've never had any experience with the linksys family of wireless routers. Until recently the entire WRT series ran Linux. And yes, cups can fit in the ROM of several models (8 MB FLASH only). The WRT54GSL, though, has a usb port and enough flash to run quite a bit of stuff. Samba, cups, even a small proxy server.
There are many laser printers under $300. Samsung makes a few fairly nice ones. But networkable, that's another story. Probably your best bet is a USB printer supported by linux and then stick CUPs on a Linksys 54GSL or whatever it's called (the wireless router with USB support).
Although toner is very expensive, you can get a decent full-color laser for about $300-600. The HP Laserjet 2600n. See http://www.nextag.com/hp-2600n/search-html . The n designation means it comes with jetdirect too. The only downside is the printer language is not PCL or PS, but there is a CUPS driver for linux available. The extra money for color may be worth the extra couple of hundred.
As for myself, I have an old Okidata 10ex LED printer that is parallel only. I use a USB-Parallel adapter and plug it into my linux box (cups server) with the USB. The linux box shares it to windows and linux clients. I recently upgraded the RAM to 32 MB, so it should be able ot handle anything I throw at it for years to come, even if I have to wait a while for the pages to spit out. Toner is separate from the drum, so it's dirt cheap to fill. I recently bought a new drum for it for $60. This printer has been one of my best computer investments.
Well except for the fact that extreme cold appears to increase the rate of decay (according to these scientists), which if it works for carbon-14 would show the sample to be significantly older than it really is, people already claim carbon-14 dating is inaccurate due to temperature fluxuation.
The thing is that carbon dating is really useless for the things that many people want to prove. Since the accuracy is plus or minus thousands of years, I don't see it proving anything about, say, a wood sample purported to be from Noah's ark. On the other hand it *does* show us that the earth is millions of years old generally.
That's really 10000 writes to a single cell on the flash chip. All flash memory has built-in algorithms to statistically spread out those writes over all the cells of the chip, so it's not like if you wrote a file 10000 times you'd have an instant failure. If you only write to on average, say 10% of the disk, then you wouldn't see a failure for probably 100000 writes. Given a 4 GB flash chip, the average write is probably just a few megabytes, so works out to even longer time between failures. Still not a huge lifetime, but not as short as you say.
Yes, but without the BIOS, the hardware was useless.
The oven in my kitchen is a slightly evolved version of ovens 50 years ago. Heck even my microwave is just a slightly improved version of the first microwaves back in the late 70s. Some technologies have advanced, but others have just gotten slightly improved over the years, because as poor as they are, they are still the best. Consider the internal combustion engine. While computer controls have increased reliability and decreased emissions dramatically (and increased power and efficiency), the engine is basically unchanged from what it was in the 1920s. Overhead cams, mulitple valves per cylinder, super and turbo-chargers where around from the 20s. Even fuel injection has been here for years. Yet engines steadily improve in capability and power, even if we're using 100-year-old base technology. Similarly, until solid-state memory gets cheap enough and reliable enough, spinning magnetic disks will likely remain the main storage solution for years to come. And I think that's okay. I don't think it means we're somehow behind. No matter what technology comes along, I will always have a love/hate relationship with it.
You mean drivers like, say the operating system?
From the article:
"With a brand that said 'business machine' and an open architecture that invited third-party innovation, the IBM PC transformed the IT industry."
It seems we forget that when the PC was first introduced it was closed and proprietary. It wasn't until Compaq clean-room reverse-engineered the BIOS that the PC revolution really got started. If IBM had had their way the PC would have been locked down and controlled by IBM forever. Remember they used to call clones "IBM compatible." After Compaq started the cloning revolution, and Microsoft moved to make IBM-specific aspects of DOS irrelevant, not long after that IBM started to become less and less relevant. They no longer directed where the platform was going. By the i386, one could no longer talk about IBM-compatible. IBM tried to start over with a proprietary system (careful not to let cloning happen this time) withe Microchannel Architecure. Fortunately the market said, we'll stick with ISA, VESA-Local and PCI (even if MCA was superior at the time). Had IBM been successful in keeping the PC proprietary, I don't know what computers we would be using today. Maybe DEC alphas or Sparcstations. Or maybe we'd be paying $10000 a pop to IBM.
Apple never supported the open source version of darwin in any way beyond lip services, some server space, and releasing source packages in mostly unbuildable form. They took from many open source projects but returned precious little to the community. At the end of the day Apple does what immediately benefits Apple. It's sad, but it's likely the threat of hacking OS X to run on white box computers likely is the greatest reason for Apple to not release vital parts of the latest OS X source code. Yet this will still happen. In the meantime, Linux continues to grow and become better all the time. There just was no need for OpenDarwin without Aqua. If all you want is a unix-like OS to run servers, Linux suits the bill just fine.
I think the main problem for you is that SQL and relational databases were designed to store relational database data, not be a persistant store for objects. Indeed if you use a database as it was designed to be used, SQL is highly appropriate as a query language.
I think Microsoft is correct, as stated in another post, that ORM is inherently flawed. It may be the best thing we have, but shoe-horning database data into classes and vice versa is problematic. The time it takes to mess with Java's XML mapping files, for example, often exceeds the time it would take to write some decent stored procedures, use proper database views, and then do simple SQL queries directly. And I believe it is no less manageable or scaleable.
What you are looking for is really just a better persistant-object storage scheme. It doesn't matter what the backend is like, so long as your objects are accessible in an easy and logical fashion.
Sure but until the Mactel towers are release with pci-express slots in them, there's no real support in OSX86 for nvidia cards at all. In fact there aren't even any drivers available yet, except from a third-party site. Currently OSX86 only ships with intel and ATI drivers. This says nothing about the future, though.
The article is wrong about the flexibility of answer files and streamlined installations only applying to windows itself and not applications. We use http://unattended.sf.net/ and it works very well. 20 minutes later windows is install, updated, and all our standard packages installed including MS Office. In my opinion this beats the heck out of imaging XP using Ghost in terms of flexibility. We just boot into our installer image using PXE; no disks required at all. Ironically unattended uses a linux boot image with dosemu to bootstrap the Windows installer. Very slick. Unfortunately it looks like this whole process may have to be scrapped with Vista, sadly. We'll likely avoid vista for at least a full year after it is released, to give us time to integrate it properly into our operation (our servers are and will remain Linux based--can't wait to try Samba 4). The article is right about one thing, though. Setting up unattended installations is tedius.
As people have pointed out already, by the time you spec a machine similarly from Dell, the cost is very similar. For me the weight, battery life, and the linux compatibility are worth a $300 premium. People also like to complain about the expensive Apple laptops too until I show them that Apple is often cheaper than the equivalent Thinkpad or sometimes Dell.
;)
Since the masses do think like you do, you are probably right, though. Most folks probably will take the lower-quality Dell over this laptop. This company will probably be a niche seller, but I hope they do well enough. Now that the 12" Apple laptop is no more (no the MacBook doesn't count), I may just abandon the sleek aluminum look and go for a maximum value laptop, such as this one. I could always run OS X on it...
Many high-end engines require the highest octane gas, so you'd likely have to pay for the more expensive gas anway.