Makes no sense. "Corporate upheavals nothwithstanding" -- how do corporates have anything to do with Linux?
Several commercial distributions (RedHat, Novell-SuSE) use their own branch of the Linux kernel.
GPL -- I thought the very spirit of the GPL was to allow everyone to make their own forks
On the other hand the GPL allows all the changes occuring in separate kernel branches to be incorporated in the main branch.
Centralised Research Lab: I thought the very spirit of Linux was to have voluntary developers from around the world, not people working in some centralised lab.
This one makes sense in connection with:
Linus Torvalds' unassailable position --- This is the only one that makes sense.
Because of the centralised research lab OSDN channeling all work and giving Linus Torvalds (and the maintainers for the older Kernel versions 2.2 and 2.0) the organisational tools at hand to maintain the central kernel branch this branch even exists.
As it seems Asia and Africa are much longer inhabited by humans than both of the Americas (at least 1,7 mio years for Africa and at least 40.000 years for Asia, but newer founds point at the same time range for Asia than for Africa). So in Africa and Asia all large mammals that have survived until 10.000 years ago were surely adapted to the presence of humans in their neighbourhood, quite different for the Americas.
But the extinction of large mammals could also have been connected to several different factors. For European cave bears there is valid evidence that they couldn't really cope with the climate during the Ice Age. Most finds so far point at diseases spread within the population. Basicly the cold and humid air in the European caves caused the whole population to have a sore throat and a bad cold;). With the appearance of the modern human following the retracting ice the bears were to weak to deal with the new challenge. They didn't have the time to recover from the diseases because they were already hunted by cave dwellers driving them out of their homes.
It doesn't even count all those desktop boxes acting as servers. Once, when the Xbox was still a good deal for the hardware you got in it we even considered buying Xboxes and installing linux on them to have cheap servers which don't run too hot. No, it just counts physical entities sold with the label "server" and tries to determine what the people paid for them and what software was delivered with them. But that's nothing new.
"free" in this case means "free of a direct fee or charge".
And I agree: Not everyone who pays with his taxes will actually need this service but is forced to pay for it.
But: Not everyone needs bicycle lanes in the town, but nevertheless he has to pay for it. Not everyone needs a new coat of paint at the town hall, but he has to pay for it. Not everyone needs flowers at the central plaza, but he has to pay for it. Not everyone needs the new painting in the mayor's office, but he has to pay for it.
Just because a single person doesn't need something that is paid for by tax money, doesn't mean it is bad if the town pays for it anyway. There are things that get paid for even though they aren't needed by many people.
That's what a town council is for: To determine what to do with tax money. If the town council throws the money at stuff no one really needs, the next town council will be different. And there are always control mechanisms in place to point out where tax money gets wasted for meaningless, dangerous or otherwise undesirable projects. You don't need a special law for that. Imagine a town where everyone agrees in a voting, that wireless LAN services paid for by tax money makes sense. But a law voted for by completely other people forbids it for no other reason than some companies crying loudly.
A community is an economic entity like others are too: single persons, charities, corporations. A community is even (like a business) out to make money: tax money, fees, money from business it owns. Why should we exclude a community from doing business? There is only one valid reason: monopoly considerations. If a community is able to control a whole branch of business because of a monopoly in another branch, it may be bad, because the service in the first branch may be detoriated but there is no business model to offer a better service because of the monopoly in another branch.
But free wireless LAN in general doesn't look to me affected by those dangers. Just because it's free it doesn't mean it's bad. Most telcos offer free local calls also, so where is the harm? Having free wireless LAN is just a matter of local infrastructure, and this surely is within a communal regulation. And if the inhabitants of a community vote to have wireless LAN offered by a communal utility, who should forbid this by a state law? It's clearly against the democratically expressed will of the inhabitants. And they are paying for it with their own taxes. Only if a large enough number of people don't want wireless LAN provided by the town and don't agree to pay for it by other means (yes, someone has to pay) there is a case against free wireless LAN.
Originally everyone tried the RMS 100% free approach but that lead to no acceptance outside of the geeks.
Which RMS 100% free approach? RMS's approach was always to subsequently replace every program necessary to run a computer by a free equivalent. So 100% free was the goal far away on the horizon, and until then it was meant to replace proprietary software part by part.
It's exactly the same approach you are proposing: Make the environment to the user consisting of more and more free software, until the underlying operating system can be replaced without looking back. Why do you think the long time goal was to create a complete free tool chain first (editor, compiler, linker, shell...) and then with those tools create more software and finally an operating system? Because the working environment should be free first, so the code creating the futural software was not tainted by proprietary parts.
Of course this approach was "for geeks" first, because the GNU tool chain is more often used by sysadmins and developers (typical geek ecology) than by normal users. Now the tool chain is completed, and at least one operating system is completely GNU licensed, and the programs for other people are coming in.
The question was not, if.NET (version 1.0, version 1.1) are vaporware. They are present and usable.
The question was, if the.NET OS is vaporware, and if the rumours are somewhat related to the truth, it may be. The question was, if the rush to rebuild everything on.NET to be able to serve Microsoft's next generation OS, was founded on vaporware. It might be, if the rumours contain a grain of correctness.
I think for everyone who makes PC add on hardware, it's interesting to go into chipset design, it's just not affordable for everyone. Especially if the hardware requires high bandwiths and/or low latency, knowing the chipset inside out helps to quench out that last drop of latency the specification allows.
So if you are already in a position to know chipsets in every detail, why not go and redesign them to better support your own hardware designs?
Don't you have a second hand trader around, who sells you an old ATi Mach or RagePro for $5? That's what I am using for servers. I always have some of them lying around.
You are missing the other point: Reboot if you do something that affects the boot process. Normally you don't have to do anything to the boot process or the software involved in the boot process. Only if you patch kernel and loadable modules, you have to reboot. If you patch software that is started at boot time it should be enough to just restart the services. That's why you have the/etc/init.d/ directory (or/etc/rc.d/init.d or/etc/system/init.d or whatever...) where all the scripts are to be used to start and stop services. The one who screwed up his firewall was actually recompiling the kernel without testing if it runs. That's bad. The one with 550days uptime didn't have any kernel patches to test. So no reboot. If you look at the patch descriptions they actually tell you if a reboot is necessary after applying. And no. You don't fiddle around in the boot scripts of a working server. You have a second server where you can happily reboot as you like to test necessary changes to boot scripts. And even then above said still applies: Restarting the service using the changed boot script should be a sufficient test. UNIX is designed to have as much as possible independent of each other. And a vendor changing a boot script in a way that it requires to change the sequence the boot scripts should be called will always put BIG WARNING SIGNs at this patch. As does the free and open source community.
The Eiger is a mountain top in Switzerland, and notoriously famous for people dying while trying to climb the northern side. The first successful attempt took place in 1938, but two years early four people died tragically in front of the eyes of witnesses, who were watching the climbing from nearby views, the last one only a few yards away from a rescue team.
I thus don't think it's a good idea to name a project after a mountain top famous for failed and deadly attempts to finally climb it.
Because at all companies I worked so far all root access had to be documented and stored separately from the workplaces. This is called a business continuity procedure. Imagine yourself running over by a truck.
It's easy. Natural diamonds grow very slowly, and every disturbing of the growth process show in the diamonds. If you find a perfect crystal (*) of a certain size, you can be pretty sure it's artificial.
(*) Perfect as in "every layer of atoms the same", not necessarily "every layer consisting solely of carbon atoms, and all carbon atoms in a tetraedic order"
This was, what made MUDs, MUSHs, MOOs and all those other games so cool. Of course since the advent of EverQuest, no one knows them anymore.:) But with MUDs we got even psychologists starting to program who were shying away from computers like the devil from holy water.
I did a similar test, driving a distance of about 400 mls with the old Audi 100 2.3CC of my father. Going one direction I drove as fast as the traffic was allowing, up to 90mph (this was in Germany after all;) ), and the car consumed nearly the whole gas tank (20 gallons). Travelling back I did never accelerate to more than 60mph, and the car consumed about 11 gallons. Interestingly though both times I managed to drive the whole distance in about 6:10h.
On the other hand: About 20 years ago a local university tested different strategies to get through town traffic: Accelerating sharply until the town's speed limit is reached (32mph in Germany) and then switching in the 4th gear, or accelerating slowly, and switching gear early, or trying to accelerate in the highest gear possible. The first alternative yielded the best mileage: Accelerate sharply until you reach the speed you are desiring and then switch to a high gear. The conclusion was that accelerating puts the car into a non economic state of work anyway, so you should try to go through this state as fast as possible und try to reach a very economic state: Driving with low revs at a constant speed.
Having Not-MS running at school lowers all the cost associated to children constantly reconfiguring the software, installing 'cool' stuff and otherwise render the PCs unusable. Of course you can try to lock down the PCs as much as possible, having them reinstalled for each course and all the other ways to keep the PCs in a non-surprising and workable state. But all those are associated with additional cost (either having someone knowledgable setting up the labs, probably to be hired from outside, or sending the responsible teachers to training courses or whatever). For Linux there are educational distributions (in Germany for instance ask Schulen ans Netz e.V.), which take care of the special problems of educational computer labs. You can create workable computer images with ease, and without violating the license agreements that came with the software. You have a very good set of computer work related tools already within every distribution, so there is no cost for additional software. And: for a school it could be very important: You keep a lot of computer players out of the lab and thus are freeing seats for people who might actually do their homework or class projects after regular hours.
There is another aspect: Chip sets. To have a chipset being accepted by the motherboard manufacturers, it has to come with easy implementable reference designs and with a small board real estate, that is: a minimum of used area for the chips itself and for the connections to other functions. So chip sets tend to integrate as many functions as possible, because thus board manufactures need less additional circuits and wiring to put those functions at the boards. A large portion of the boards in turn are sold to assemblers anyway, which are keen on boards with many functions already builtin, because then they don't need to put additional cards and ports into the boxes they are assembling. So for a board manufacturer to get into the assembler business to sell boards he has to offer fully integrated boards, and he will choose fully integrated chipsets to deliver. And if he has once designed fully integrated board series, what's the point in designing stripped down versions again? He will use the same chipsets anyway, because he has all the testing equipment in place for those, he reuses the design of the wiring (once the masks for wiring are done, the manufactunging costs are the same, independent of the number of wires), so all he saves are the few cents for the actual sockets and the soldering of the sockets to the boards. Basicly a "bare" board then is nothing else than a board without the sockets, but electrically the functions are there. But to manufacture those boards intentionally you need another production process, another QA process, another packaging. It might be cheaper to just sort the boards according to their final testing results and then specify which functions are 'there' (those that work in the tests).
If there is no evidence or proof for a higher power yet, it just means that there was nothing that could be solely explained by the existance of such a higher power. Thus for any reasoning about anything we have evidence for, we don't need the higher power.
There are the so called algebraic numbers. That are numbers which are the solutions to rational polynoms (polynoms whose elements are all rational numbers). You can't create a polynom with rational numbers, for which Pi is a solution. So for the algebraic numbers, there is no "evidence" for Pi. But Pi exists in non rational polynomic worlds.
On the other hand: If you are looking for a solution to an algebraic problem consisting of rational polynoms, wondering about the existance of Pi doesn't help you. It is not part of any solution you are seeking.
It's the same with God and his creation. There is no scientific question yet, whose answer is 'God', which is just another term for 'there is no scientific evidence for the existance of God'. So wondering about God's existance doesn't solve any scientific problems and doesn't answer any scientific questions for now.
And here comes my purely private speculation: If we ever discover anything resembling what we call 'God' today, we are just making another turn and find out that also this God might have restrictions put at his doing, which are nothing else than our physical laws on the next level. Imagine the small lifeforms on your skin discovering that you have a conciousness and free will and are able to determine your future. Does this invalidate anything they found out so far about the chemical processes on your skin? About the physical laws determining the outcomes of their experiments? About the cosmos surrounding them with day and night, cold and heat, dry and wet air?
No, just someone who tried to make sense from the article, and what Lyon was described to be doing.
Lyon basicly built a big application proxy which simulated the application of the gambling site in question and was able to filter requests at every stage of the frontend data processing. Lyon thus was able to scan the logfiles for apparent attack patterns (for instance a sudden spike in requests for a certain sub service of the whole application) and answer with filtering patterns. Only requests got through that were considered 'normal' traffic for technical and heuristic reasons. The article talked also about the constant stress in adapting the application proxy to the attack.
So if you look at the defense described in the article, you can try to find out what type of attack it was designed to withstand. There is the big bandwith usage. The article talks about 1Gb/sec up to 3 Gb/sec. There is the application proxy, which points to an direct application attack, not just easily filterable noise traffic like ping of death or SYN-ACK-attacks. And there is the constant stress of the people working frantically at the proxy which points to a variable attack with changing patterns and changing targets within the application. This leads to the four descriptions I gave for the DDoS attack.
Slashdot Mirror
Several commercial distributions (RedHat, Novell-SuSE) use their own branch of the Linux kernel.
On the other hand the GPL allows all the changes occuring in separate kernel branches to be incorporated in the main branch.
This one makes sense in connection with:
Because of the centralised research lab OSDN channeling all work and giving Linus Torvalds (and the maintainers for the older Kernel versions 2.2 and 2.0) the organisational tools at hand to maintain the central kernel branch this branch even exists.
As it seems Asia and Africa are much longer inhabited by humans than both of the Americas (at least 1,7 mio years for Africa and at least 40.000 years for Asia, but newer founds point at the same time range for Asia than for Africa). So in Africa and Asia all large mammals that have survived until 10.000 years ago were surely adapted to the presence of humans in their neighbourhood, quite different for the Americas.
;). With the appearance of the modern human following the retracting ice the bears were to weak to deal with the new challenge. They didn't have the time to recover from the diseases because they were already hunted by cave dwellers driving them out of their homes.
But the extinction of large mammals could also have been connected to several different factors. For European cave bears there is valid evidence that they couldn't really cope with the climate during the Ice Age. Most finds so far point at diseases spread within the population. Basicly the cold and humid air in the European caves caused the whole population to have a sore throat and a bad cold
It doesn't even count all those desktop boxes acting as servers. Once, when the Xbox was still a good deal for the hardware you got in it we even considered buying Xboxes and installing linux on them to have cheap servers which don't run too hot.
No, it just counts physical entities sold with the label "server" and tries to determine what the people paid for them and what software was delivered with them. But that's nothing new.
What a nicely fitting signature...
"free" in this case means "free of a direct fee or charge".
And I agree: Not everyone who pays with his taxes will actually need this service but is forced to pay for it.
But:
Not everyone needs bicycle lanes in the town, but nevertheless he has to pay for it.
Not everyone needs a new coat of paint at the town hall, but he has to pay for it.
Not everyone needs flowers at the central plaza, but he has to pay for it.
Not everyone needs the new painting in the mayor's office, but he has to pay for it.
Just because a single person doesn't need something that is paid for by tax money, doesn't mean it is bad if the town pays for it anyway. There are things that get paid for even though they aren't needed by many people.
That's what a town council is for: To determine what to do with tax money. If the town council throws the money at stuff no one really needs, the next town council will be different. And there are always control mechanisms in place to point out where tax money gets wasted for meaningless, dangerous or otherwise undesirable projects. You don't need a special law for that. Imagine a town where everyone agrees in a voting, that wireless LAN services paid for by tax money makes sense. But a law voted for by completely other people forbids it for no other reason than some companies crying loudly.
A community is an economic entity like others are too: single persons, charities, corporations. A community is even (like a business) out to make money: tax money, fees, money from business it owns. Why should we exclude a community from doing business? There is only one valid reason: monopoly considerations. If a community is able to control a whole branch of business because of a monopoly in another branch, it may be bad, because the service in the first branch may be detoriated but there is no business model to offer a better service because of the monopoly in another branch.
But free wireless LAN in general doesn't look to me affected by those dangers. Just because it's free it doesn't mean it's bad. Most telcos offer free local calls also, so where is the harm? Having free wireless LAN is just a matter of local infrastructure, and this surely is within a communal regulation. And if the inhabitants of a community vote to have wireless LAN offered by a communal utility, who should forbid this by a state law? It's clearly against the democratically expressed will of the inhabitants. And they are paying for it with their own taxes. Only if a large enough number of people don't want wireless LAN provided by the town and don't agree to pay for it by other means (yes, someone has to pay) there is a case against free wireless LAN.
Which RMS 100% free approach? RMS's approach was always to subsequently replace every program necessary to run a computer by a free equivalent. So 100% free was the goal far away on the horizon, and until then it was meant to replace proprietary software part by part.
It's exactly the same approach you are proposing: Make the environment to the user consisting of more and more free software, until the underlying operating system can be replaced without looking back. Why do you think the long time goal was to create a complete free tool chain first (editor, compiler, linker, shell...) and then with those tools create more software and finally an operating system? Because the working environment should be free first, so the code creating the futural software was not tainted by proprietary parts.
Of course this approach was "for geeks" first, because the GNU tool chain is more often used by sysadmins and developers (typical geek ecology) than by normal users. Now the tool chain is completed, and at least one operating system is completely GNU licensed, and the programs for other people are coming in.
Freetrain ;)
The question was not, if .NET (version 1.0, version 1.1) are vaporware. They are present and usable.
.NET OS is vaporware, and if the rumours are somewhat related to the truth, it may be. The question was, if the rush to rebuild everything on .NET to be able to serve Microsoft's next generation OS, was founded on vaporware. It might be, if the rumours contain a grain of correctness.
The question was, if the
I think for everyone who makes PC add on hardware, it's interesting to go into chipset design, it's just not affordable for everyone. Especially if the hardware requires high bandwiths and/or low latency, knowing the chipset inside out helps to quench out that last drop of latency the specification allows.
So if you are already in a position to know chipsets in every detail, why not go and redesign them to better support your own hardware designs?
To answer this you should read "Der kleine Riese" (The small giant) by Gerd Prokop.
Don't you have a second hand trader around, who sells you an old ATi Mach or RagePro for $5? That's what I am using for servers. I always have some of them lying around.
You are missing the other point: Reboot if you do something that affects the boot process. Normally you don't have to do anything to the boot process or the software involved in the boot process. Only if you patch kernel and loadable modules, you have to reboot. /etc/init.d/ directory (or /etc/rc.d/init.d or /etc/system/init.d or whatever...) where all the scripts are to be used to start and stop services.
If you patch software that is started at boot time it should be enough to just restart the services. That's why you have the
The one who screwed up his firewall was actually recompiling the kernel without testing if it runs. That's bad. The one with 550days uptime didn't have any kernel patches to test. So no reboot.
If you look at the patch descriptions they actually tell you if a reboot is necessary after applying. And no. You don't fiddle around in the boot scripts of a working server. You have a second server where you can happily reboot as you like to test necessary changes to boot scripts. And even then above said still applies: Restarting the service using the changed boot script should be a sufficient test. UNIX is designed to have as much as possible independent of each other. And a vendor changing a boot script in a way that it requires to change the sequence the boot scripts should be called will always put BIG WARNING SIGNs at this patch. As does the free and open source community.
As a german I hereby confirm. The word is not in wide use though, because ogres in fairy tales are mostly called "Riesen" (giants).
Lukily I am no native speaker and thus not subject to the Grammar Police. You might still try the LGRT (Luser's Grammar Readjustment Tool).
The Eiger is a mountain top in Switzerland, and notoriously famous for people dying while trying to climb the northern side. The first successful attempt took place in 1938, but two years early four people died tragically in front of the eyes of witnesses, who were watching the climbing from nearby views, the last one only a few yards away from a rescue team.
I thus don't think it's a good idea to name a project after a mountain top famous for failed and deadly attempts to finally climb it.
Because at all companies I worked so far all root access had to be documented and stored separately from the workplaces. This is called a business continuity procedure. Imagine yourself running over by a truck.
It's easy. Natural diamonds grow very slowly, and every disturbing of the growth process show in the diamonds. If you find a perfect crystal (*) of a certain size, you can be pretty sure it's artificial.
(*) Perfect as in "every layer of atoms the same", not necessarily "every layer consisting solely of carbon atoms, and all carbon atoms in a tetraedic order"
This was, what made MUDs, MUSHs, MOOs and all those other games so cool. Of course since the advent of EverQuest, no one knows them anymore. :)
But with MUDs we got even psychologists starting to program who were shying away from computers like the devil from holy water.
I did a similar test, driving a distance of about 400 mls with the old Audi 100 2.3CC of my father. Going one direction I drove as fast as the traffic was allowing, up to 90mph (this was in Germany after all ;) ), and the car consumed nearly the whole gas tank (20 gallons). Travelling back I did never accelerate to more than 60mph, and the car consumed about 11 gallons. Interestingly though both times I managed to drive the whole distance in about 6:10h.
On the other hand: About 20 years ago a local university tested different strategies to get through town traffic: Accelerating sharply until the town's speed limit is reached (32mph in Germany) and then switching in the 4th gear, or accelerating slowly, and switching gear early, or trying to accelerate in the highest gear possible. The first alternative yielded the best mileage: Accelerate sharply until you reach the speed you are desiring and then switch to a high gear. The conclusion was that accelerating puts the car into a non economic state of work anyway, so you should try to go through this state as fast as possible und try to reach a very economic state: Driving with low revs at a constant speed.
Having Not-MS running at school lowers all the cost associated to children constantly reconfiguring the software, installing 'cool' stuff and otherwise render the PCs unusable. Of course you can try to lock down the PCs as much as possible, having them reinstalled for each course and all the other ways to keep the PCs in a non-surprising and workable state. But all those are associated with additional cost (either having someone knowledgable setting up the labs, probably to be hired from outside, or sending the responsible teachers to training courses or whatever).
For Linux there are educational distributions (in Germany for instance ask Schulen ans Netz e.V.), which take care of the special problems of educational computer labs. You can create workable computer images with ease, and without violating the license agreements that came with the software. You have a very good set of computer work related tools already within every distribution, so there is no cost for additional software.
And: for a school it could be very important: You keep a lot of computer players out of the lab and thus are freeing seats for people who might actually do their homework or class projects after regular hours.
There is another aspect: Chip sets.
To have a chipset being accepted by the motherboard manufacturers, it has to come with easy implementable reference designs and with a small board real estate, that is: a minimum of used area for the chips itself and for the connections to other functions. So chip sets tend to integrate as many functions as possible, because thus board manufactures need less additional circuits and wiring to put those functions at the boards. A large portion of the boards in turn are sold to assemblers anyway, which are keen on boards with many functions already builtin, because then they don't need to put additional cards and ports into the boxes they are assembling.
So for a board manufacturer to get into the assembler business to sell boards he has to offer fully integrated boards, and he will choose fully integrated chipsets to deliver.
And if he has once designed fully integrated board series, what's the point in designing stripped down versions again? He will use the same chipsets anyway, because he has all the testing equipment in place for those, he reuses the design of the wiring (once the masks for wiring are done, the manufactunging costs are the same, independent of the number of wires), so all he saves are the few cents for the actual sockets and the soldering of the sockets to the boards.
Basicly a "bare" board then is nothing else than a board without the sockets, but electrically the functions are there. But to manufacture those boards intentionally you need another production process, another QA process, another packaging. It might be cheaper to just sort the boards according to their final testing results and then specify which functions are 'there' (those that work in the tests).
If there is no evidence or proof for a higher power yet, it just means that there was nothing that could be solely explained by the existance of such a higher power. Thus for any reasoning about anything we have evidence for, we don't need the higher power.
There are the so called algebraic numbers. That are numbers which are the solutions to rational polynoms (polynoms whose elements are all rational numbers). You can't create a polynom with rational numbers, for which Pi is a solution. So for the algebraic numbers, there is no "evidence" for Pi. But Pi exists in non rational polynomic worlds.
On the other hand: If you are looking for a solution to an algebraic problem consisting of rational polynoms, wondering about the existance of Pi doesn't help you. It is not part of any solution you are seeking.
It's the same with God and his creation. There is no scientific question yet, whose answer is 'God', which is just another term for 'there is no scientific evidence for the existance of God'. So wondering about God's existance doesn't solve any scientific problems and doesn't answer any scientific questions for now.
And here comes my purely private speculation: If we ever discover anything resembling what we call 'God' today, we are just making another turn and find out that also this God might have restrictions put at his doing, which are nothing else than our physical laws on the next level. Imagine the small lifeforms on your skin discovering that you have a conciousness and free will and are able to determine your future. Does this invalidate anything they found out so far about the chemical processes on your skin? About the physical laws determining the outcomes of their experiments? About the cosmos surrounding them with day and night, cold and heat, dry and wet air?
Check Google for "KDE 3.4" and "zeroconf" to answer that.
No, just someone who tried to make sense from the article, and what Lyon was described to be doing.
Lyon basicly built a big application proxy which simulated the application of the gambling site in question and was able to filter requests at every stage of the frontend data processing. Lyon thus was able to scan the logfiles for apparent attack patterns (for instance a sudden spike in requests for a certain sub service of the whole application) and answer with filtering patterns. Only requests got through that were considered 'normal' traffic for technical and heuristic reasons. The article talked also about the constant stress in adapting the application proxy to the attack.
So if you look at the defense described in the article, you can try to find out what type of attack it was designed to withstand. There is the big bandwith usage. The article talks about 1Gb/sec up to 3 Gb/sec. There is the application proxy, which points to an direct application attack, not just easily filterable noise traffic like ping of death or SYN-ACK-attacks. And there is the constant stress of the people working frantically at the proxy which points to a variable attack with changing patterns and changing targets within the application. This leads to the four descriptions I gave for the DDoS attack.