Of course, putting a dollar cost on wasted bandwidth creates a financial incentive to remove the waste. e.g. Symantec could set the price of Windows Update Cache Pro(TM) just below the cost of bandwidth that it saves...
Metered bandwidth would be good for several reasons.... it would provide a financial disincentive for people to use file sharing software for illegal reasons...
Maybe, maybe not. In that scenario, if I download an episode of "24" from iTunes, I have to pay Apple and my ISP, but if I download it through unauthorized channels I only pay my ISP. But metering would certainly shift the balance from sharing towards leeching.
there is no way Microsofts OS can differentiate between user-installed spyware and legitimate apps.
Even assuming that is true, it is still possible to design an OS where legitimate software works and spyware doesn't. It's called the principle of least authority.
is there something fundamental about the Java language that lends itself to or requires the whole JIT-bytecode thing?
No, not really.
I mean, if it's a Turing-complete language, can't it be compiled into any machine code you want, with the right compiler? Is there some reason why one has never been built?
It's called GCJ or JET (and in the old days there was BulletTrain and another one I don't remember).
But why would Java be inherently any more slow than any other language, if it were directly compiled on an optimized x86 compiler?
Because the language semantics are different, which affects the allowable optimizations. (This effect is not unique to Java; if you write "the same" program in C and Fortran, in some cases the Fortran version will be faster because Fortran has stricter aliasing rules.)
How about a package manager that downloads the code, lets you inspect, customize, or debug it, then compiles it and adds it to your modules list once you approve it?
Any code that is compiled locally is by definition untested, and many people don't want to run untested drivers.
But if you want Gentoo, you know where to find it.
If the hardware manufacturers publish thier source code it wouldn't be a problem.
Not necessarily. For example, if a vendor publishes source code for a driver, it takes months for that driver to appear in the next stable kernel, and likewise months until the next SLES or RHEL quarterly update. Novell's process allows the vendors to ship drivers whenever they are ready, without waiting for anyone else. But as I said, the cost is that the vendors have to do a lot of work.
This "breakthrough" requires device vendors to recompile (and possibly port) their driver for every distro, every time that distro updates their kernel ABI. The only thing that has really changed seems to be that Novell will keep track of when the kernel ABI changes and notify driver developers.
IP address space has not run out anywhere, because all ISPs (eventually) get address blocks from the same pool. And when that pool does run out, everyone will run out at the same time.
That is true, but I don't think you can buy a PC with a good old Matrox G400 or Radeon 9200 any more; you can't even buy one of those cards on eBay and put it in a modern PC, because AGP is obsolete. For PCs that were manufactured in 2006, it's pretty much Intel or binary drivers.
Intel publishes open source drivers for their latest integrated video chipsets.
While people may report that some Matrox or ATI cards work fine with open source drivers, those are all old, discontinued cards. If you want to buy new hardware, Intel is basically your only choice for open source drivers.
More like his wallet couldn't handle the $1,000 extreme processor. Considering the poor value of Intel's extreme editions, maybe the people who buy them are the suckers.
In general, there is no secondary market for digital goods. Either it has DRM which disallows resale completely, or it has no DRM in which case people just copy it for $0. In theory there could be "friendly DRM' that allows resale, but if publishers feel threatened by it they can simply not use it.
Slashdot Mirror
I sense a little bias here; the fastest Intel and AMD processors are actually $1,000.
Of course, putting a dollar cost on wasted bandwidth creates a financial incentive to remove the waste. e.g. Symantec could set the price of Windows Update Cache Pro(TM) just below the cost of bandwidth that it saves...
Metered bandwidth would be good for several reasons. ... it would provide a financial disincentive for people to use file sharing software for illegal reasons...
Maybe, maybe not. In that scenario, if I download an episode of "24" from iTunes, I have to pay Apple and my ISP, but if I download it through unauthorized channels I only pay my ISP. But metering would certainly shift the balance from sharing towards leeching.
So why not just borrow it outright and not be stuck paying per hour?
AFAIK, the whole point of pay-as-you go is for people with no credit.
So why should Apple re-host Darwin on some new microkernel? What benefit do they get?
And the benefit of encrypting BitTorrent connections is what? All the *AA have to do is ask the tracker who's downloading what, and it will tell them.
there is no way Microsofts OS can differentiate between user-installed spyware and legitimate apps.
Even assuming that is true, it is still possible to design an OS where legitimate software works and spyware doesn't. It's called the principle of least authority.
is there something fundamental about the Java language that lends itself to or requires the whole JIT-bytecode thing?
No, not really.
I mean, if it's a Turing-complete language, can't it be compiled into any machine code you want, with the right compiler? Is there some reason why one has never been built?
It's called GCJ or JET (and in the old days there was BulletTrain and another one I don't remember).
But why would Java be inherently any more slow than any other language, if it were directly compiled on an optimized x86 compiler?
Because the language semantics are different, which affects the allowable optimizations. (This effect is not unique to Java; if you write "the same" program in C and Fortran, in some cases the Fortran version will be faster because Fortran has stricter aliasing rules.)
Nah, they can just charge millions for testing and logos.
How about a package manager that downloads the code, lets you inspect, customize, or debug it, then compiles it and adds it to your modules list once you approve it?
Any code that is compiled locally is by definition untested, and many people don't want to run untested drivers.
But if you want Gentoo, you know where to find it.
If the hardware manufacturers publish thier source code it wouldn't be a problem.
Not necessarily. For example, if a vendor publishes source code for a driver, it takes months for that driver to appear in the next stable kernel, and likewise months until the next SLES or RHEL quarterly update. Novell's process allows the vendors to ship drivers whenever they are ready, without waiting for anyone else. But as I said, the cost is that the vendors have to do a lot of work.
This "breakthrough" requires device vendors to recompile (and possibly port) their driver for every distro, every time that distro updates their kernel ABI. The only thing that has really changed seems to be that Novell will keep track of when the kernel ABI changes and notify driver developers.
Sun already tried shipping the Linux-based "Java Desktop System", and then canceled it in favor of Solaris on the desktop.
The userland has nothing to do with it; people have been running GPL userland on proprietary kernels for years (decades?).
IP address space has not run out anywhere, because all ISPs (eventually) get address blocks from the same pool. And when that pool does run out, everyone will run out at the same time.
That is true, but I don't think you can buy a PC with a good old Matrox G400 or Radeon 9200 any more; you can't even buy one of those cards on eBay and put it in a modern PC, because AGP is obsolete. For PCs that were manufactured in 2006, it's pretty much Intel or binary drivers.
Intel publishes open source drivers for their latest integrated video chipsets.
While people may report that some Matrox or ATI cards work fine with open source drivers, those are all old, discontinued cards. If you want to buy new hardware, Intel is basically your only choice for open source drivers.
Mondrian memory protection (also see Mondrix) has been suggested as a hardware solution to this problem, but hardware tends to change very slowly.
I'm curious to the kernel hackers out there whether a 100-core CPU will see an advantage in either the micro-kernel vs. monolithic approach?
The K42 team certainly thinks that large multiprocessor systems benefit from a different OS stucture and one of their ingredients is a microkernel.
More like his wallet couldn't handle the $1,000 extreme processor. Considering the poor value of Intel's extreme editions, maybe the people who buy them are the suckers.
The solution is GStreamer; it's modular and Fluendo is working on legal proprietary plug-ins for it. Or you can download the underground plug-ins.
I don't know about that; very few motherboards have HTX slots, but lots of motherboards have multiple processor sockets.
Music, for instance, is trivially easy to copy, and yet purchases continue.
I was talking about the secondary market for digital goods. No one pays for "used MP3s" or "used iTunes songs".
In general, there is no secondary market for digital goods. Either it has DRM which disallows resale completely, or it has no DRM in which case people just copy it for $0. In theory there could be "friendly DRM' that allows resale, but if publishers feel threatened by it they can simply not use it.
64GB two-socket Opteron systems (e.g. IWill DK88) are rare but available for people who need them.