Performance?? I can't imagine a real-time rendering engine (i.e., games) in.NET or java.
You might need a bit of code in C for that for performance, but most of a game is better done in Java or C#. Games like World of Warcraft actually include their own scripting language and garbage collector. These could be compiled to bytecodes on load and run at full speed instead of being interpreted and using a much slower GC. It would also enable them to do a lot more cool stuff; instead of C++ where you have to explicitly write code to let a script do something you could, in Java, say what it couldn't do if you wanted to.
There are plenty of examples of game engines being written other languages you might not think are suitable, like LISP for instance. The game industry mostly uses C++ these days due to inertia (all those existing engines / code written in C++) and ignorance rather than performance. Incidentally, that's also why the OS isn't written in a safe language -- there's no technical reason favoring C/C++ over Java/C# for an OS.
The only reason I haven't purchased a Mac is because I do not have the option of dual-booting to play games on Windows. Sad, but true. I would but a x86 Mac today if I could, and use it as the main OS. And why the hell would I write open-source software for an OS I don't use every day? Let some Linux fan port the code if they want to use it.
There are lots of developers out there like me, who write code for the OS they use, and who would get a Mac86. Linux will be hurt by this move, but it will also keep growning well. Windows will be the real loser. People talk about how Mac will lose performance benchmarks, which is probably true, but forget that it will crush Linux and Windows on boot time, resume-from-suspend time, GUI l&f, and other benchmarks end-users actually care about.
Hopefully Apple will also be able to influence the direction of BIOS and other non-os-specific stuff to be less crappy...
Actually VMware is normally within 2-3% of native performance for anything besides context switches*, which are very slow (although they are slow on any OS), and IO. So if you are doing lots of maths, ie photoshop, then it will be essentially indistinguishable from native speed.
* - which incidentally is why Fedora Core runs like a turd on VMware... they lowered the timeslice a LOT to appear responsive to desktop users, when they should have raised the priority of the X server and window manager.
Also don't forget that Intel's processors actually run microcode, not x86 instructions. The microcode can be patched, but only if it is encrypted/signed by the right key (that only intel has), so not too much is actually known about its capabilities, but it might be possible for some Intel chips to natively run PPC instruction set using microcode.
The main sucky feature of Intel chips is the x86 instruction set, which is basically pure crap -- it couldn't really get any worse. So even if there is a 20% overhead from running a non-optimal instruction set, if you have a mobile chip that is 4x faster than a G4 (G5 runs too hot for a laptop) it could be a significant performance improvement, provide Intel with a migration away from x86 instruction set, and not have any affect on the Mac software.
I don't care who pays for it as long as I don't have to do anything. As long as I can call for special pickup and dump it on the curb like anything else then it could cost me $50 more for my state tax for all I care.
But if I have to do anything different from what I would with normal odd-sized trash then I'd just throw it in the dump when they aren't looking and then everybody loses.
ST has $1.1 billion cash, so Sun is really spending more like $3 billion. At $191 million profit that's over 6% return on their investment. Plus you have to expect sales to increase due to companies storing more data due to the recent demise of that accounting firm due to aggressively destroying documents.
Then you factor in the forthcoming zfs, which should make Solaris far better than any other operating system for handling mass data storage and they could do very well by this deal.
There is really no way to prevent technically savy people from making copies of content which is distributed on media that does not have user specific encryption without owning the complete system that is responsible for playback.
Completely wrong... They only have to control one component of the system, not all components. For example, they don't have to ban winamp if they have a device driver that does audio fingerprinting, because it can then detect any software playing the song. You might get 5 seconds into a song before the dialog comes up demanding you pay for the song to continue.
Sure, you could record the analog and play it back on your 1970s 8-track. Or you could alter the audio data to avoid a fingerprint and then use special hardware before the speakers to decode it... but then you are already on the defensive because any encoding that becomes popular can be specifically recognized and disabled. In any case, a fingerprinting driver and secure audio path will cut out 99% of piracy and there's nothing you can do about it.
You guys are talking about cost effectiveness of chips used by, to be generous, maybe 5% of the computers. The fact that a chip used by that few consumers can ever compete with ones used by the other 95% on price/performance should really be telling you something important: yes, Intel chips really are the suck... even with margins of scale they still are often left in the dust.
Or maybe you just didn't understand the guy? He was probably just saying you couldn't play a video and do any other math at the same time at the processor level which seem true enough to me. Probably the fact he was fat, greasy, and lacked social skills predisposed you to misinterpreting him. Yeah, ok that's still a bad image for Apple...
...but Intel processors are the suck. If processors weren't a prime example of market lock-in there would be no reason to use any x86 (even AMD) over Power, or over 68k back in the day. Check sandpile.org and see if you can made any sense in the instruction set.
Microkernels still rely on separate memory spaces, so will always be too slow. System calls in Linux take the quivalent of about ~1000 instructions. Microkernels improve on this a bit but do it far more often. That's a lot of wasted time.
If you look at a typical executable, about 75% of the instructions executed were run within the last 1000. IOW, the address in the instruction pointer register was the same within the last 1000 instructions 3 times in 4. This is due to loops, common functions, etc, but what it really means is that if you ask anything from the system in any part of your program requiring performance it'll run at least half as fast as it could.
That's why the virtual machine should be the os. Then performance critical code that wants to use OS services can safely run directly in the kernel.
First off, COM wasn't a "sufficiently expressive language" since it was a bunch of macros and conventions on top of C/C++, which puts it into the same camp as interfaces in Simula. Second, if you refuse to believe the Java developers themselves and take into account their extensive background in Smalltalk and Obj C and instead believe they ripped it from COM of all things because, well, that's just what you believe then that's pretty lame.
asAnInterface: ^self
-- (c) 1991 Microsoft Corporation. Patent pending
0x: In any case, the idea that Microsoft Research invented interfaces and that's where Java got them is simply wrong on any number of counts.
Bn: Who said that? Me?... Bn: Finally, you want to claim Microsoft Research are credit-stealing whores over the Java community? Where do you think the idea of objects supporting multiple interfaces for the purposes of inclusion polymorphism (seperated from the nightmare of multiple inheritance that plagues/criples OO-languages) comes from? Java?
Java owes a lot (beyond even that) to MS Research.
So, you both implied that Microsoft originated interfaces and asserted it. And you are simply wrong about it.
COM is a piece of poo rip off with interfaces, delegation, and thikly caked layers of crud calling itself OOP; like Windows, Microsoft popularized it. Harldy something to be proud of IMO.
And here the British were supposed be good with sarcasm, but the point that interfaces in OOP have been done since time immemorial obviously escaped you. Regardless, it certainly is possible to have interfaces in Simula, by grouping the interface methods into a Class, having the interface class as an attribute on each Class "implementing" that interface, and delegating the instance methods to the enclosing class methods. Then you can test for implementation of the interface, access methods on the interface without knowing the concrete type or supertype(s) of the object, and it is inherited to subclasses.
I should hope your experts could come up with better ways to implement interfaces in Simula. In any case, the idea that Microsoft Research invented interfaces and that's where Java got them is simply wrong on any number of counts.
First, Microsoft have a longer history of developing virtual machines for programming languages (in Visual Basic).
Riiight, because Microsoft has so many ex-Smalltalk pople working for it unlike Sun*... Sun knows more about VMs just from Self then Mircosoft ever will.
Java owes a lot (beyond even that) to MS Research. Don't judge a company by its Operating Systems division..
Java owes nothing to Microsoft Research. Nothing. Your lack of seemingly any context regarding VMs is simple astounding.
Finally, you want to claim Microsoft Research are credit-stealing whores over the Java community? Where do you think the idea of objects supporting multiple interfaces for the purposes of inclusion polymorphism (seperated from the nightmare of multiple inheritance that plagues/criples OO-languages) comes from?
Um, Simula way back in 1967.
As for being whores, look at some of the other 'innovations' they have discovered:
Singularity: "safe" C#-based operating system done by a total of over 50 people, with 4 paid full-time for 2 years. DOS interface.
vs.
jxos: "safe" Java-based OS done by a handful of people in basically their spare time. AWT interface with lists, buttons, textboxes, etc. Plays minesweeper.
One of them is a direct rip, apparently from the interview including even the underlying architecture.
Users don't give a flip about how anything you mentioned is implemented, which is all you are talking about when you say "relational filesystem this and that": how it is implemented.
Apple's OS X has plenty of directories that appear as files to users, and that is fine since users don't care about the details. If all docuements and files were in a "relational filesystem" (whatever that is), you would still need something like Spotlight to convert all the different schemas and data into a usable, searchable model.
So what problem is a relational filesystem going to solve? Unless users will start dragging tables around instead of icons then the only thing it could do is make some operations more convenient and faster -- but the price will be lots of interdependent schemas and tables. For example, everything will break when you change a file size from being a simple value to being a key into the file size table, which stores the type-specific size (ie width x height for images, duration for songs, etc).
Sure that's a simple example that maybe nobody would do, but the fact is that a relational filesystem will just jumble lots of currently separate information together making for even more dependencies and likelyhood for breaking. Would you let any random joe add new tables, columns, etc to your companies databases? So why would you want them to mess with your personal data either?
Is there some place on the.net where a person can get an overview on what company owns what other companies? I know that First USA was taken over by Bank One and then by Chase in like less than a year. They could all be owned by one of the affected banks for all I know!
How can anybody reasonable know these things? Is there a www.CorporateTree.com or something?
I'm not trying to get more karma, but there are other advantages that I thought of:
Firefox could have an 'identity manager' that stores your public/private keys along with the name, address, phone, etc for that key. Then for a "fox-id" enabled site FF can automatically insert that data into the appropriate fields; the user would still have the ability to edit/delete individual data before sending it. So from a user perspective it could be a simple one-click "Use this id" drop down selecting the id to send. Most users would just use one id because they don't care about their info being private, but savy people could still create a different id per site or just different levels.
Companies don't like relying anything external that could go down and impact their business, so they would take to a system where for most users there is no 3rd party -- all interactions would be between their site and the user's browser. Companies also like to have real information, so sites that really care could cross-check your public key against some authority (ie ask VISA if your public key matches your credit card info). Most sites would still allow anonymous access, but ones that cared like online shops could validate your info for cheap and so would reduce fraud -- you could no longer buy just by stealing somebody's credit card number, you would also need their private key and personal info.
Also sites could just include the user's login in the URL as the login, so going to slashdot.org/~0xABADC0DA would automatically log me in but nobody else because only I have my private key.
Of course there are downsides, like companies that buy your public key and info so they can put your name on the page even the first time you visit, but this could be managed by needing to add trusted sites in your browser config (like for cookies). It would also let companies tie together all you purchases even across keys based on matching the names, addresses, etc. But this already happens, so what's the difference?
What I want is a system where I go to a site requiring a login and it asks my browser to sign some data with my private key. During the account creation I send the server my public key and that's that -- no need for a password and the login could be done automatically using cookies or something. Then there is no need for a single sign-on provider and nobody can globally revoke my account at all sites.
You could still have an 'id provider' that could sign the data on your behalf if you are on a internet cafe for instance, but it would not be required by design. So in 'kiosk mode' the browser could just forward signiture requests to the authority after you logged into it (which could even be your home computer).
This should be pretty easy to do as a firefox plug-in.
Does anybody know where I can read the original article that O'Gara wrote? I can't find it anywhere; all I can find are stories talking about it and making vague references. I am curious what was actually written in the article.
I always thought once something got onto the web it was there forever, but even with the controversy and massive flap it's like the article just vanished completely!
"Until Longhorn is fully operational we are vulnerable. Linux is too well equipped. It's more dangerous than you realize."
"Dangerous to your OS division, not to my Office line."
"Linux will continue to gain support along with OpenOffice as long as Red Hat continues to..."
"Red Hat will no longer be of any conern to us. I've just received word that Emperor Gates has disbanded the Red Hat. The last remants of Unix have been swept away."
"But without Red Hat, how will the local servers be kept on-line?"
"Fear. Fear of lawsuits will keep the local system on-line. This monopoly is now the ultimate power in the internet."
"Do not be so proud of this technological terror you have constructed; it is insignificant next to the power of the Source"
"Don't try to frighten us with your Sourcers ways, Lord Bahlmer..."
No, but if you have vmware or the right hardware you might want to check out JXOS an all-Java operating system. It was done by I think 2 people in their spare time, and it's like 50% of Linux speed at an actual benchmark (nfs filesystem). That's pretty good for two guys writing everything from scratch.
IMO, Freenet is trying to solve the wrong problem by trying to make it impossible to detect even through statistics, impossible to shut down, etc. What they should be doing is building the freenet features as a trojan part of a larger software. Make a 'plausable deniability' P2P system that works and is fast and you can send any other data you want as a small amount of overhead. For example, if as you are using bittorrent you are also sending.5k / sec of encrypted "free" data out of China then your problem is solved.
Basically, include lots of Joe Innocents and porn downloads in the network as white noise.
Slashdot Mirror
Performance?? I can't imagine a real-time rendering engine (i.e., games) in .NET or java.
You might need a bit of code in C for that for performance, but most of a game is better done in Java or C#. Games like World of Warcraft actually include their own scripting language and garbage collector. These could be compiled to bytecodes on load and run at full speed instead of being interpreted and using a much slower GC. It would also enable them to do a lot more cool stuff; instead of C++ where you have to explicitly write code to let a script do something you could, in Java, say what it couldn't do if you wanted to.
There are plenty of examples of game engines being written other languages you might not think are suitable, like LISP for instance. The game industry mostly uses C++ these days due to inertia (all those existing engines / code written in C++) and ignorance rather than performance. Incidentally, that's also why the OS isn't written in a safe language -- there's no technical reason favoring C/C++ over Java/C# for an OS.
The only reason I haven't purchased a Mac is because I do not have the option of dual-booting to play games on Windows. Sad, but true. I would but a x86 Mac today if I could, and use it as the main OS. And why the hell would I write open-source software for an OS I don't use every day? Let some Linux fan port the code if they want to use it.
There are lots of developers out there like me, who write code for the OS they use, and who would get a Mac86. Linux will be hurt by this move, but it will also keep growning well. Windows will be the real loser. People talk about how Mac will lose performance benchmarks, which is probably true, but forget that it will crush Linux and Windows on boot time, resume-from-suspend time, GUI l&f, and other benchmarks end-users actually care about.
Hopefully Apple will also be able to influence the direction of BIOS and other non-os-specific stuff to be less crappy...
Actually VMware is normally within 2-3% of native performance for anything besides context switches*, which are very slow (although they are slow on any OS), and IO. So if you are doing lots of maths, ie photoshop, then it will be essentially indistinguishable from native speed.
* - which incidentally is why Fedora Core runs like a turd on VMware... they lowered the timeslice a LOT to appear responsive to desktop users, when they should have raised the priority of the X server and window manager.
Also don't forget that Intel's processors actually run microcode, not x86 instructions. The microcode can be patched, but only if it is encrypted/signed by the right key (that only intel has), so not too much is actually known about its capabilities, but it might be possible for some Intel chips to natively run PPC instruction set using microcode.
The main sucky feature of Intel chips is the x86 instruction set, which is basically pure crap -- it couldn't really get any worse. So even if there is a 20% overhead from running a non-optimal instruction set, if you have a mobile chip that is 4x faster than a G4 (G5 runs too hot for a laptop) it could be a significant performance improvement, provide Intel with a migration away from x86 instruction set, and not have any affect on the Mac software.
I don't care who pays for it as long as I don't have to do anything. As long as I can call for special pickup and dump it on the curb like anything else then it could cost me $50 more for my state tax for all I care.
But if I have to do anything different from what I would with normal odd-sized trash then I'd just throw it in the dump when they aren't looking and then everybody loses.
ST has $1.1 billion cash, so Sun is really spending more like $3 billion. At $191 million profit that's over 6% return on their investment. Plus you have to expect sales to increase due to companies storing more data due to the recent demise of that accounting firm due to aggressively destroying documents.
Then you factor in the forthcoming zfs, which should make Solaris far better than any other operating system for handling mass data storage and they could do very well by this deal.
There is really no way to prevent technically savy people from making copies of content which is distributed on media that does not have user specific encryption without owning the complete system that is responsible for playback.
Completely wrong... They only have to control one component of the system, not all components. For example, they don't have to ban winamp if they have a device driver that does audio fingerprinting, because it can then detect any software playing the song. You might get 5 seconds into a song before the dialog comes up demanding you pay for the song to continue.
Sure, you could record the analog and play it back on your 1970s 8-track. Or you could alter the audio data to avoid a fingerprint and then use special hardware before the speakers to decode it... but then you are already on the defensive because any encoding that becomes popular can be specifically recognized and disabled. In any case, a fingerprinting driver and secure audio path will cut out 99% of piracy and there's nothing you can do about it.
You guys are talking about cost effectiveness of chips used by, to be generous, maybe 5% of the computers. The fact that a chip used by that few consumers can ever compete with ones used by the other 95% on price/performance should really be telling you something important: yes, Intel chips really are the suck... even with margins of scale they still are often left in the dust.
Or maybe you just didn't understand the guy? He was probably just saying you couldn't play a video and do any other math at the same time at the processor level which seem true enough to me. Probably the fact he was fat, greasy, and lacked social skills predisposed you to misinterpreting him. Yeah, ok that's still a bad image for Apple...
...but Intel processors are the suck. If processors weren't a prime example of market lock-in there would be no reason to use any x86 (even AMD) over Power, or over 68k back in the day. Check sandpile.org and see if you can made any sense in the instruction set.
Microkernels still rely on separate memory spaces, so will always be too slow. System calls in Linux take the quivalent of about ~1000 instructions. Microkernels improve on this a bit but do it far more often. That's a lot of wasted time.
If you look at a typical executable, about 75% of the instructions executed were run within the last 1000. IOW, the address in the instruction pointer register was the same within the last 1000 instructions 3 times in 4. This is due to loops, common functions, etc, but what it really means is that if you ask anything from the system in any part of your program requiring performance it'll run at least half as fast as it could.
That's why the virtual machine should be the os. Then performance critical code that wants to use OS services can safely run directly in the kernel.
First off, COM wasn't a "sufficiently expressive language" since it was a bunch of macros and conventions on top of C/C++, which puts it into the same camp as interfaces in Simula. Second, if you refuse to believe the Java developers themselves and take into account their extensive background in Smalltalk and Obj C and instead believe they ripped it from COM of all things because, well, that's just what you believe then that's pretty lame.
asAnInterface: ^self
-- (c) 1991 Microsoft Corporation. Patent pending
0x: In any case, the idea that Microsoft Research invented interfaces and that's where Java got them is simply wrong on any number of counts.
...
Bn: Who said that? Me?
Bn: Finally, you want to claim Microsoft Research are credit-stealing whores over the Java community? Where do you think the idea of objects supporting multiple interfaces for the purposes of inclusion polymorphism (seperated from the nightmare of multiple inheritance that plagues/criples OO-languages) comes from? Java?
Java owes a lot (beyond even that) to MS Research.
So, you both implied that Microsoft originated interfaces and asserted it. And you are simply wrong about it.
COM is a piece of poo rip off with interfaces, delegation, and thikly caked layers of crud calling itself OOP; like Windows, Microsoft popularized it. Harldy something to be proud of IMO.
And here the British were supposed be good with sarcasm, but the point that interfaces in OOP have been done since time immemorial obviously escaped you. Regardless, it certainly is possible to have interfaces in Simula, by grouping the interface methods into a Class, having the interface class as an attribute on each Class "implementing" that interface, and delegating the instance methods to the enclosing class methods. Then you can test for implementation of the interface, access methods on the interface without knowing the concrete type or supertype(s) of the object, and it is inherited to subclasses.
I should hope your experts could come up with better ways to implement interfaces in Simula. In any case, the idea that Microsoft Research invented interfaces and that's where Java got them is simply wrong on any number of counts.
Riiight, because Microsoft has so many ex-Smalltalk pople working for it unlike Sun*... Sun knows more about VMs just from Self then Mircosoft ever will.
Java owes nothing to Microsoft Research. Nothing. Your lack of seemingly any context regarding VMs is simple astounding.
Um, Simula way back in 1967.
As for being whores, look at some of the other 'innovations' they have discovered:
Singularity: "safe" C#-based operating system done by a total of over 50 people, with 4 paid full-time for 2 years. DOS interface.
vs.
jxos: "safe" Java-based OS done by a handful of people in basically their spare time. AWT interface with lists, buttons, textboxes, etc. Plays minesweeper.
One of them is a direct rip, apparently from the interview including even the underlying architecture.
* -- that's sarcasm btw.
Note to self: when you type in the new script-prevention code and hit return, it posts immediately intead of previews
Note to self: check spelling before posting...
Note to self: when you type in the new script-prevention code and hit return, it posts immediately intead of previews
Users don't give a flip about how anything you mentioned is implemented, which is all you are talking about when you say "relational filesystem this and that": how it is implemented.
Apple's OS X has plenty of directories that appear as files to users, and that is fine since users don't care about the details. If all docuements and files were in a "relational filesystem" (whatever that is), you would still need something like Spotlight to convert all the different schemas and data into a usable, searchable model.
So what problem is a relational filesystem going to solve? Unless users will start dragging tables around instead of icons then the only thing it could do is make some operations more convenient and faster -- but the price will be lots of interdependent schemas and tables. For example, everything will break when you change a file size from being a simple value to being a key into the file size table, which stores the type-specific size (ie width x height for images, duration for songs, etc).
Sure that's a simple example that maybe nobody would do, but the fact is that a relational filesystem will just jumble lots of currently separate information together making for even more dependencies and likelyhood for breaking. Would you let any random joe add new tables, columns, etc to your companies databases? So why would you want them to mess with your personal data either?
Is there some place on the .net where a person can get an overview on what company owns what other companies? I know that First USA was taken over by Bank One and then by Chase in like less than a year. They could all be owned by one of the affected banks for all I know!
How can anybody reasonable know these things? Is there a www.CorporateTree.com or something?
I'm not trying to get more karma, but there are other advantages that I thought of:
Firefox could have an 'identity manager' that stores your public/private keys along with the name, address, phone, etc for that key. Then for a "fox-id" enabled site FF can automatically insert that data into the appropriate fields; the user would still have the ability to edit/delete individual data before sending it. So from a user perspective it could be a simple one-click "Use this id" drop down selecting the id to send. Most users would just use one id because they don't care about their info being private, but savy people could still create a different id per site or just different levels.
Companies don't like relying anything external that could go down and impact their business, so they would take to a system where for most users there is no 3rd party -- all interactions would be between their site and the user's browser. Companies also like to have real information, so sites that really care could cross-check your public key against some authority (ie ask VISA if your public key matches your credit card info). Most sites would still allow anonymous access, but ones that cared like online shops could validate your info for cheap and so would reduce fraud -- you could no longer buy just by stealing somebody's credit card number, you would also need their private key and personal info.
Also sites could just include the user's login in the URL as the login, so going to slashdot.org/~0xABADC0DA would automatically log me in but nobody else because only I have my private key.
Of course there are downsides, like companies that buy your public key and info so they can put your name on the page even the first time you visit, but this could be managed by needing to add trusted sites in your browser config (like for cookies). It would also let companies tie together all you purchases even across keys based on matching the names, addresses, etc. But this already happens, so what's the difference?
What I want is a system where I go to a site requiring a login and it asks my browser to sign some data with my private key. During the account creation I send the server my public key and that's that -- no need for a password and the login could be done automatically using cookies or something. Then there is no need for a single sign-on provider and nobody can globally revoke my account at all sites.
You could still have an 'id provider' that could sign the data on your behalf if you are on a internet cafe for instance, but it would not be required by design. So in 'kiosk mode' the browser could just forward signiture requests to the authority after you logged into it (which could even be your home computer).
This should be pretty easy to do as a firefox plug-in.
Followed by:
"Delisting: when short-sellers attack".
Does anybody know where I can read the original article that O'Gara wrote? I can't find it anywhere; all I can find are stories talking about it and making vague references. I am curious what was actually written in the article.
I always thought once something got onto the web it was there forever, but even with the controversy and massive flap it's like the article just vanished completely!
almost...
"Until Longhorn is fully operational we are vulnerable. Linux is too well equipped. It's more dangerous than you realize."
"Dangerous to your OS division, not to my Office line."
"Linux will continue to gain support along with OpenOffice as long as Red Hat continues to..."
"Red Hat will no longer be of any conern to us. I've just received word that Emperor Gates has disbanded the Red Hat. The last remants of Unix have been swept away."
"But without Red Hat, how will the local servers be kept on-line?"
"Fear. Fear of lawsuits will keep the local system on-line. This monopoly is now the ultimate power in the internet."
"Do not be so proud of this technological terror you have constructed; it is insignificant next to the power of the Source"
"Don't try to frighten us with your Sourcers ways, Lord Bahlmer..."
No, but if you have vmware or the right hardware you might want to check out JXOS an all-Java operating system. It was done by I think 2 people in their spare time, and it's like 50% of Linux speed at an actual benchmark (nfs filesystem). That's pretty good for two guys writing everything from scratch.
IMO, Freenet is trying to solve the wrong problem by trying to make it impossible to detect even through statistics, impossible to shut down, etc. What they should be doing is building the freenet features as a trojan part of a larger software. Make a 'plausable deniability' P2P system that works and is fast and you can send any other data you want as a small amount of overhead. For example, if as you are using bittorrent you are also sending .5k / sec of encrypted "free" data out of China then your problem is solved.
Basically, include lots of Joe Innocents and porn downloads in the network as white noise.