Yeah, there's no room in the Internet for new protocols to be used along side the old protocols. No one adopted AIM, Gnutella, Real, or BitTorrent because they're not backwards compatible.
You are right that the people who designed TCP are smart, and error correction was understood at the time TCP was designed. But at that time computers weren't powerful enough to use FEC in TCP. TCP may have been the optimal protocol back in the 80s, but thanks to Moore's Law it may not be optimal today.
So you can't tolerate less than the original amount data being sent before retransmit?
Correct. That is dictated by information theory.
Why not just stick with TCP?
TCP only retransmits after it discovers that a packet has been dropped, which takes many milliseconds. Since forward error correction schemes are constantly transmitting check blocks, the lost data can be recovered much sooner.
Companies like Minerva and Pace have TV over IP stuff that works and is deployed today. Microsoft is going to have to offer something either cheaper or better if they want to take over the TV over IP market.
Various companies are working on exactly that, but I wonder if it's worth it. Cellular/Wi-Fi roaming is so complex that it's going to cost a lot, which means they will charge you a lot for it. So in the end it'll probably be cheaper just to use a regular cell phone.
The nForce chipsets don't support specific sockets -- they support HyperTransport. That means they support all K8 sockets. The chance of someone building a high-end Socket 754 motherboard is slim, though.
DUNDi and Skype are both P2P, but otherwise they are quite different. Skype uses supernodes to get through NATs; I don't think DUNDi even tries. Skype is designed to connect a huge number of PCs/phones; DUNDi looks like it's more of a server-to-server protocol (can it scale to millions of peers?). Skype users are mostly anonymous; DUNDi requires peers to negotiate contractual relationships.
And I can't resist picking on this...
Maybe someone will adapt the Speak Freely user interface to a P2P transport layer, maybe utilizing the Gnutella network itself?
Yeah, let's combine an obsolete, non-standards-compliant VoIP app with the least efficient P2P system available.
Apple chose Broadcom wireless chips because those were the first pre-802.11g chips available, and Apple stays with the same wireless chip vendor for an entire generation. Maybe when Apple moves to pre-802.11n they'll choose a more open vendor (assuming such a thing still exists).
Apple should decide whether they want Linux users using their hardware (and the resulting money) or whether having total control of their platform and product is more important to them. They can't have it both ways.
That is correct; Apple doesn't care about Linux. How much clearer do you want them to be?
The number of PowerBook Linux users is so small that you discredit yourself by threatening to switch to another platform.
Back in 2002, people were pitching IB as a replacement for PCI. Today, nobody tries to do that -- IB and PCI are used for different purposes (clustering and I/O expansion, respectively).
I know PowerBooks and obscure RISC laptops used to use SCSI drives, but I'm talking about present and future laptops: there's no way you're going to get SCSI, because it's too far out of the mainstream now.
Since MCE is only available preloaded on PCs, and those PCs must have TV tuner cards and remotes, it's hard to imagine than an MCE PC would be cheaper than a regular PC with XP Pro.
Slashdot Mirror
How come we haven't been using such schemes already?
Because they used to be considered CPU-intensive. Moore's Law took care of that aspect.
Also because there has been a lot of research in this area recently so FEC has gotten better.
Yeah, there's no room in the Internet for new protocols to be used along side the old protocols. No one adopted AIM, Gnutella, Real, or BitTorrent because they're not backwards compatible.
You are right that the people who designed TCP are smart, and error correction was understood at the time TCP was designed. But at that time computers weren't powerful enough to use FEC in TCP. TCP may have been the optimal protocol back in the 80s, but thanks to Moore's Law it may not be optimal today.
So you can't tolerate less than the original amount data being sent before retransmit?
Correct. That is dictated by information theory.
Why not just stick with TCP?
TCP only retransmits after it discovers that a packet has been dropped, which takes many milliseconds. Since forward error correction schemes are constantly transmitting check blocks, the lost data can be recovered much sooner.
When you make up numbers (40:1), you get wrong results. Over the air HDTV is 18 Mbps. WMV3 HDTV is under 10 Mbps.
If you want two x16 slots you just need to use two nForce 4 chips. Tyan is coming out with such a motherboard.
Companies like Minerva and Pace have TV over IP stuff that works and is deployed today. Microsoft is going to have to offer something either cheaper or better if they want to take over the TV over IP market.
I would expect that they'll provide you with a set top box -- no computer needed.
(Of course the box will be proprietary, but so are cable boxes.)
Various companies are working on exactly that, but I wonder if it's worth it. Cellular/Wi-Fi roaming is so complex that it's going to cost a lot, which means they will charge you a lot for it. So in the end it'll probably be cheaper just to use a regular cell phone.
The broadcast flag doesn't apply to TV over IP, so it's easy to predict that the broadcast flag will have no effect here.
Cable and satellite are totally DRM'ed today, so DRM'ed TV over IP is not really any worse.
Why would they? If Apple is violating patents, the patent holders aren't going to sue OS X users or application developers.
BlueGene/L is only at 36TFLOPS today, but by next year the full-size version is supposed to clock in at 180TFLOPS. The SX-8 has no chance to survive.
The nForce chipsets don't support specific sockets -- they support HyperTransport. That means they support all K8 sockets. The chance of someone building a high-end Socket 754 motherboard is slim, though.
Exactly. Packages are arbitrary, cores are not.
DUNDi and Skype are both P2P, but otherwise they are quite different. Skype uses supernodes to get through NATs; I don't think DUNDi even tries. Skype is designed to connect a huge number of PCs/phones; DUNDi looks like it's more of a server-to-server protocol (can it scale to millions of peers?). Skype users are mostly anonymous; DUNDi requires peers to negotiate contractual relationships.
And I can't resist picking on this...
Maybe someone will adapt the Speak Freely user interface to a P2P transport layer, maybe utilizing the Gnutella network itself?
Yeah, let's combine an obsolete, non-standards-compliant VoIP app with the least efficient P2P system available.
Apple chose Broadcom wireless chips because those were the first pre-802.11g chips available, and Apple stays with the same wireless chip vendor for an entire generation. Maybe when Apple moves to pre-802.11n they'll choose a more open vendor (assuming such a thing still exists).
Apple should decide whether they want Linux users using their hardware (and the resulting money) or whether having total control of their platform and product is more important to them. They can't have it both ways.
That is correct; Apple doesn't care about Linux. How much clearer do you want them to be?
The number of PowerBook Linux users is so small that you discredit yourself by threatening to switch to another platform.
If you look at the PDF, NASA is 4th with 8 nodes. If they want to reach #1 they'll need to get 8 more nodes online pronto.
Back in 2002, people were pitching IB as a replacement for PCI. Today, nobody tries to do that -- IB and PCI are used for different purposes (clustering and I/O expansion, respectively).
I know PowerBooks and obscure RISC laptops used to use SCSI drives, but I'm talking about present and future laptops: there's no way you're going to get SCSI, because it's too far out of the mainstream now.
Actually, now you can fit 10 drives in 1U instead of 4.
SCSI in laptops? Keep dreaming.
You mean the "frequency isn't everything, it's the only thing" Alpha team?
If you want a PDA, you know where to find one. The point of the OQO is that a PDA is not enough for some people.
Since MCE is only available preloaded on PCs, and those PCs must have TV tuner cards and remotes, it's hard to imagine than an MCE PC would be cheaper than a regular PC with XP Pro.
As a fellow Opteron owner, I know the feeling. But this article is about A64s, which definitely do have Cool 'n' quiet.
OpenDarwin sucks, so no one uses it.