My brother in law was an avid Microsoft fan until XP. As a busy doctor with 7 kids, he liked the "no decisions" aspect of Windows. (All that choice can be time consuming - which boot loader? which window manager? which desktop? which 2.4 kernel VM?...)
However, the Windows XP EULA horrified him, and he has switched to Linux. Having a distribution make most of the decisions for him helps a lot. He uses Mandrake.
The setup has to handle sudden interruption of the WAN link at any stage of the transfer (e.g. primary site gets bombed). rsync is OK for email directories and such, but not for the entire system.
There needs to be some sort of journalling - which you get for free when mirroring updates to an ext3 filesystem at the block level (and copy blocks in order of update). There is a semi-free (open source + proprietary extensions to make it useful) linux block driver from a company in Austria which sounds more promising.
A number of other replies echo similar and worse problems with EMC.
I tried an
Ask Slashdot article on ideas for open source implementations of disk mirroring over WAN. I was roundly flamed for failing to find EMC on google. I had no idea that EMC was as bad as it is, but I did know that it was way way way too expensive for our customers. I am looking for disaster recovery that is better than offsite tapes, and affordable for small companies with < 100 employees (All of which enter large volumes of transaction data daily.)
We did get an email from MiraLink which is still way too expensive, but they promised to review their pricing.
You shouldn't assume the person you are talking to doesn't know any chemistry.
I would certainly support efforts to ban transport of free chlorine (the thought of a tanker of the stuff in an accident gives me the creeps), and perhaps even to restrict manufacturing processes using it. But the Greenpeacer wanted to ban the _element_ !
FYI, there are much nastier poisons than dioxins containing oxygen, but no chlorine - many of them natural! Poison ivy comes to mind . . .
I had a subscription to a wonderful "Rational Environmentalism" magazine called "Garbage". It was edited by Patritia Poore. The magazine dissappeared suddenly in 1996. I still have all my issues, but have been unable to find out what happened to it - or Patricia Poore. There were many very valuable articles. For instance, a toxicologist gave an introduction to toxins, and the difference between acute and cumulative poisoning.
I could sure use the Garbage approach to sort out the competing claims concerning global warming and other alleged crises. Does anyone else remember "Garbage"? Can you recommend a replacement in the same spirit? I support the Nature Conservancy, which I find very rational (unlike GreenPeace), but they deal only with species conservation.
I wrote off Greenpeace after their campaign to ban the element Chlorine (1995) because so many deadly poisons contain chlorine. I had a Greenpeacer come to my house asking for funds. I mentioned the many deadly poisons containing Oxygen, and suggested they might want to ban that also.
Our customers insist on using Outhouse. Some of them use Netscape instead, but it doesn't help. While Netscape properly pops up a window saying, "You are about to execute an untrusted program from the internet which may contain a virus... blah blah blah" - the user *always* clicks OK to install the virus.
What *has* helped is installing a
sendmail milter to remove executable Windoze attachments.
I can easily list all the files a binary RPM will install with their modes, ownership, and size (rpm -qlvp), plus any scripts it will execute during/after installation. RPM will automatically checks for conflicts with existing RPM controlled files. Configure/make/install do not do this. One of the major pains when creating a new RPM.spec is trying to figure out exactly what files "make install" is going to install:-(
I can get the source RPM, unpack it, look things over, tweak a few things, then build my own custom binary RPM. This gives me all the benefits of configure/make/make install, plus the package tracking benefits of RPM.
Communist economics is based on top-down centralized control of production. Government committees determine how many aspirin to manufacture/sell/stockpile in each district. I have personally seen this work well on a small scale: up to several hundred people in a commune. Beyond that, it fails, with or without corruption. A central committee simply has no idea now many aspirin are or will be needed in a remote province. Communism succedes when each decision maker has at least some personal contact with everyone affected by his decisions.
Free markets are about bottom-up decentralized control of production. The distinction has little to do with making money.
Microsoft is all about top down control of software research and production.
Free Software is all about localized, bottom-up control software research and production.
When a commune goes bad, you can leave - unless it is one of those really nasty cults. Microsoft is so big, that you can't get away from them. They is why they can get away with ever more oppressive licensing terms.
There are advantages to centralized production. It is nice when Microsoft drives standardization of PC hardware. But participation needs to be voluntary. Microsoft is becoming like a cult - it becomes harder and harder for its members to leave.
If its less than $100, a lot of people will be tempted to give it a try - even if they already have MSOffice, and especially if they don't have an office suite yet.
VLIW is a good argument for dynamic compilation such as used by Smalltalk and Java (HotSpot) and AS400. I see a future where most programming is done on a virtual machine. This insulates the code from changes to the CPU. One problem with VLIW is that tiny variations in functional units, etc, tend to require recompilation.
I mentioned "genetic programming" - which is what you mean by genetic algorithms. All of these that I have ever seen do not create new code, but find optimal combinations and/or configurations of pre-written code modules.
I'll let the experts argue about fossils. However, I do know something about writing code.
Bacteria adapt to antibiotics by exchanging existing genetic code (via plasmids, for example) with other bacteria, like script kiddies exchanging root kits. Organisms like us adapt by exchanging code via sexual reproduction, and to a lesser extent via imprinting. Natural selection works to find the optimal combination of code modules and the optimal configuration (via suppressing and promoting genes). We have even taken genetic code from one species, and injected it into the genome of another - e.g. spider genes in goats (they secret spider silk in the milk).
What is in question is the ability to write new code. What has never been observed, or a detailed specific mechanism proposed for, is the creation of new genetic code. When a programmer uses "genetic programming", they supply pre-written code modules, and let natural selection find the optimum combination and configuration. Genetic programming doesn't write the code for you.
One proposed mechanism for evolution is to start off with a "super stem cell" containing all the genetic code of all orgamisms today. It would reproduce - identically at first - but soon differentiating and specializing, like a developing foetus. Natural selection finds the optimum specialization for each habitat.
One problem with this scenario is that we don't know how organisms could shed unused genes. A bacterium has far less code than a human. However, this is easier to imagine than writing new code from thin air.
The idea that random mutations can turn a functioning gene into another functioning gene (with no fatal in between states) makes exactly as much sense as the idea the random bit mutations can turn a functioning method into a new working method with a different function (without core dumping in the process).
The whole point is to *not* wait for completion on the remote network. Since the journalling protects against corruption from the system being interrupted at any point, then interrupting the same journalling stream to the remote location at any point will not corrupt it either. You lose some data, but a whole lot less than nightly FedEx backups.
Some years ago, I was introduced to a man who had invented a "Zero Bandwidth Transmitter". He claimed to be able to transmit FM quality voice while using no bandwidth at all, and had a working prototype of a transmitter and receiver.
You know, and I know, that a "zero bandwidth" transmitter makes as much sense as 1 + 1 == 3 (or, for that matter, compression of "random" data). For this reason, despite a working prototype, the poor man had been unable to obtain a patent for his invention, despite 10 years of trying. (The Patent Office seems to be a lot looser when it comes to software.) He was very bitter and convinced that everyone else in the world was an idiot.
However, when the invention was described to me, it turns out that by "zero bandwidth", he meant "undetectable by FCC compliance measuring equipment", and that what he had really invented was a "Spread Spectrum" transmitter! What a sad story. Someone else got the patent because they could communicate it better.
So, even though compression of "random" data is mathematical nonsense, it is likely that "random" is not being used in the standard matematical sense, but in the Marketroid sense - and the new compression algorithm might actually be useful.
My father-in-law is learning to use a computer for the first time in his life (at age 70). He practices on Windows at the senior center, and on a Gnome desktop at my home. Coming without any prejudice, he finds both systems equally friendly. Overall, he would prefer Linux because he can run multiple apps at once without worrying that it might trigger a bug and crash the system . . .
How do they determine how many Linux desktops there are? We have up to seventy (70) desktops looking like a single IP to the web sites. Of course, Windows can use proxies also, but is it a good assumption to assume proxy use is the same between Linux and Windows users?
Truly good accounting packages are customized to the business. Not only regulations are constantly changing, but you want to capture business rules to code as well.
Ideally, there should be a open-source framework for the accounting - like what IBM is doing with "San Francisco", but lighter and open source. Then, it is the job of a (paid) consultant to create very specific modules for core business transactions - or a computer savvy user might undertake it themselves.
The nice thing about being that paid consultant is that the rules are constantly changing - which means the customer needs more modifications. Our company handles US import/export accounting and paperwork, and every day we get e-notices from the Feds about rule changes (with effective dates - you still have to handle the old rules with older shipments)! Needless to say, we like dynamic languages like Python and Java - and C++ doesn't cut it, except for rarely changing performance critical components.
No matter how illegal unfettered crypto becomes, there is no reason not to use it if you are planning highly illegal activities - what is there to lose?
Suppose using crypto with no government backdoor gets the death penalty - this is going to stop a suicide bomber???
Furthermore, there is no cost to use it since there is no physical product. The government could concievably control every arms manufacturing plant in the world - and this would mean criminals would have to pay some stiff bribes to obtain weapons. But the investment would really pay off since law abiding citizens would be defenselss without a police presence.
On the other hand, anyone who read the Scientific American article on RSA could produce a rough program to implement it in a few days. (No PKI, but good enough for terrorists.) For terrorists with no math background, all they need is a T-shirt with the perl code.
While I don't like their attitude, Adobe does have a long established trademark. It is not worth fighting over a name. I suggest changing the KDE product to KIlluminator. I love medieval illuminated manuscripts. In fact, "illuminator" sounds much classier than "illustrator". Adobe will be jealous.
The same people who tell us computer graphics can never be fine art, tell us that flipping a brush to spray random paint droplets is "fine art". I agree that computer graphics emphasizing the random number generator are not art. But when the computer is a medium for the artist, it is a different story. I especially enjoy literally 3D paintings that you can walk around in - e.g. Myst and Riven.
That said, the critics may be talking about the distinction between fine art, folk art, and pop art. Fine art is founded on a long tradition, and requires extensive study to fully create or enjoy it. This is because there are rich unwritten symbolisms in every stroke of fine art. In the same way, classic literature is full of references to Greek and Roman Myths, and other traditions, which is why you study them in school.
Folk art is similar, but the traditions are more regional. Pop art requires no such depth. You simply paint by the numbers:-)
Computer art is laboring to establish those essential connections to the past. They are tenous because it is new. And also because the computer art derives a great deal of its heritage from the computer culture. This is a culture that has attained at least the level of "Folk Art". But art critics are ignorant of computer culture - just as I am ignorant of most of French Impressionism.
Slashdot Mirror
However, the Windows XP EULA horrified him, and he has switched to Linux. Having a distribution make most of the decisions for him helps a lot. He uses Mandrake.
There needs to be some sort of journalling - which you get for free when mirroring updates to an ext3 filesystem at the block level (and copy blocks in order of update). There is a semi-free (open source + proprietary extensions to make it useful) linux block driver from a company in Austria which sounds more promising.
I tried an Ask Slashdot article on ideas for open source implementations of disk mirroring over WAN. I was roundly flamed for failing to find EMC on google. I had no idea that EMC was as bad as it is, but I did know that it was way way way too expensive for our customers. I am looking for disaster recovery that is better than offsite tapes, and affordable for small companies with < 100 employees (All of which enter large volumes of transaction data daily.)
We did get an email from MiraLink which is still way too expensive, but they promised to review their pricing.
I would certainly support efforts to ban transport of free chlorine (the thought of a tanker of the stuff in an accident gives me the creeps), and perhaps even to restrict manufacturing processes using it. But the Greenpeacer wanted to ban the _element_ !
FYI, there are much nastier poisons than dioxins containing oxygen, but no chlorine - many of them natural! Poison ivy comes to mind . . .
I could sure use the Garbage approach to sort out the competing claims concerning global warming and other alleged crises. Does anyone else remember "Garbage"? Can you recommend a replacement in the same spirit? I support the Nature Conservancy, which I find very rational (unlike GreenPeace), but they deal only with species conservation.
I wrote off Greenpeace after their campaign to ban the element Chlorine (1995) because so many deadly poisons contain chlorine. I had a Greenpeacer come to my house asking for funds. I mentioned the many deadly poisons containing Oxygen, and suggested they might want to ban that also.
What *has* helped is installing a sendmail milter to remove executable Windoze attachments.
Free markets are about bottom-up decentralized control of production. The distinction has little to do with making money.
Microsoft is all about top down control of software research and production.
Free Software is all about localized, bottom-up control software research and production.
When a commune goes bad, you can leave - unless it is one of those really nasty cults. Microsoft is so big, that you can't get away from them. They is why they can get away with ever more oppressive licensing terms.
There are advantages to centralized production. It is nice when Microsoft drives standardization of PC hardware. But participation needs to be voluntary. Microsoft is becoming like a cult - it becomes harder and harder for its members to leave.
If its less than $100, a lot of people will be tempted to give it a try - even if they already have MSOffice, and especially if they don't have an office suite yet.
But if this works, then all the CPU functional units will get higher duty cycles, and now we'll need a bigger fan!
VLIW is a good argument for dynamic compilation such as used by Smalltalk and Java (HotSpot) and AS400. I see a future where most programming is done on a virtual machine. This insulates the code from changes to the CPU. One problem with VLIW is that tiny variations in functional units, etc, tend to require recompilation.
How would *you* like to be deprived of a flesh and blood mother, and then "terminated to comply with regulations"?
I mentioned "genetic programming" - which is what you mean by genetic algorithms. All of these that I have ever seen do not create new code, but find optimal combinations and/or configurations of pre-written code modules.
Bacteria adapt to antibiotics by exchanging existing genetic code (via plasmids, for example) with other bacteria, like script kiddies exchanging root kits. Organisms like us adapt by exchanging code via sexual reproduction, and to a lesser extent via imprinting. Natural selection works to find the optimal combination of code modules and the optimal configuration (via suppressing and promoting genes). We have even taken genetic code from one species, and injected it into the genome of another - e.g. spider genes in goats (they secret spider silk in the milk).
What is in question is the ability to write new code. What has never been observed, or a detailed specific mechanism proposed for, is the creation of new genetic code. When a programmer uses "genetic programming", they supply pre-written code modules, and let natural selection find the optimum combination and configuration. Genetic programming doesn't write the code for you.
One proposed mechanism for evolution is to start off with a "super stem cell" containing all the genetic code of all orgamisms today. It would reproduce - identically at first - but soon differentiating and specializing, like a developing foetus. Natural selection finds the optimum specialization for each habitat.
One problem with this scenario is that we don't know how organisms could shed unused genes. A bacterium has far less code than a human. However, this is easier to imagine than writing new code from thin air.
The idea that random mutations can turn a functioning gene into another functioning gene (with no fatal in between states) makes exactly as much sense as the idea the random bit mutations can turn a functioning method into a new working method with a different function (without core dumping in the process).
The problem with DRBD is that *all* disk writes are sent over the network. This is twice the bandwidth over sending just the journal data.
The whole point is to *not* wait for completion on the remote network. Since the journalling protects against corruption from the system being interrupted at any point, then interrupting the same journalling stream to the remote location at any point will not corrupt it either. You lose some data, but a whole lot less than nightly FedEx backups.
You know, and I know, that a "zero bandwidth" transmitter makes as much sense as 1 + 1 == 3 (or, for that matter, compression of "random" data). For this reason, despite a working prototype, the poor man had been unable to obtain a patent for his invention, despite 10 years of trying. (The Patent Office seems to be a lot looser when it comes to software.) He was very bitter and convinced that everyone else in the world was an idiot.
However, when the invention was described to me, it turns out that by "zero bandwidth", he meant "undetectable by FCC compliance measuring equipment", and that what he had really invented was a "Spread Spectrum" transmitter! What a sad story. Someone else got the patent because they could communicate it better.
So, even though compression of "random" data is mathematical nonsense, it is likely that "random" is not being used in the standard matematical sense, but in the Marketroid sense - and the new compression algorithm might actually be useful.
My father-in-law is learning to use a computer for the first time in his life (at age 70). He practices on Windows at the senior center, and on a Gnome desktop at my home. Coming without any prejudice, he finds both systems equally friendly. Overall, he would prefer Linux because he can run multiple apps at once without worrying that it might trigger a bug and crash the system . . .
How do they determine how many Linux desktops there are? We have up to seventy (70) desktops looking like a single IP to the web sites. Of course, Windows can use proxies also, but is it a good assumption to assume proxy use is the same between Linux and Windows users?
Ideally, there should be a open-source framework for the accounting - like what IBM is doing with "San Francisco", but lighter and open source. Then, it is the job of a (paid) consultant to create very specific modules for core business transactions - or a computer savvy user might undertake it themselves.
The nice thing about being that paid consultant is that the rules are constantly changing - which means the customer needs more modifications. Our company handles US import/export accounting and paperwork, and every day we get e-notices from the Feds about rule changes (with effective dates - you still have to handle the old rules with older shipments)! Needless to say, we like dynamic languages like Python and Java - and C++ doesn't cut it, except for rarely changing performance critical components.
Suppose using crypto with no government backdoor gets the death penalty - this is going to stop a suicide bomber???
Furthermore, there is no cost to use it since there is no physical product. The government could concievably control every arms manufacturing plant in the world - and this would mean criminals would have to pay some stiff bribes to obtain weapons. But the investment would really pay off since law abiding citizens would be defenselss without a police presence.
On the other hand, anyone who read the Scientific American article on RSA could produce a rough program to implement it in a few days. (No PKI, but good enough for terrorists.) For terrorists with no math background, all they need is a T-shirt with the perl code.
I recommend KIlluminator as the new name. After the illuminated manuscripts done by monks.
While I don't like their attitude, Adobe does have a long established trademark. It is not worth fighting over a name. I suggest changing the KDE product to KIlluminator. I love medieval illuminated manuscripts. In fact, "illuminator" sounds much classier than "illustrator". Adobe will be jealous.
That said, the critics may be talking about the distinction between fine art, folk art, and pop art. Fine art is founded on a long tradition, and requires extensive study to fully create or enjoy it. This is because there are rich unwritten symbolisms in every stroke of fine art. In the same way, classic literature is full of references to Greek and Roman Myths, and other traditions, which is why you study them in school.
Folk art is similar, but the traditions are more regional. Pop art requires no such depth. You simply paint by the numbers :-)
Computer art is laboring to establish those essential connections to the past. They are tenous because it is new. And also because the computer art derives a great deal of its heritage from the computer culture. This is a culture that has attained at least the level of "Folk Art". But art critics are ignorant of computer culture - just as I am ignorant of most of French Impressionism.