Actually, JINI does not require a JVM, at least for network devices. Each device that wants to be identified by the JINI system somehow 'broadcasts' itself to the network, and contains a small Java 'driver'. Any devices with a user interface that want to use JINI devices must have a JVM, though. Any device that has a JVM can use any JINI device. Sun has been doing some pretty good work squeezing Java into embedded devices, and new Palms should be coming with an embedded JVM, as well as many future cell-phones, etc.
The brother of one of the guys I work with works with military satellite technology. While debugging the software on one of the sats, he took a picture of his motorcycle. Resolution was good enough to read the license plate! And this wasn't a blown-up picture, either.
If you are using 160/m and spending thousands on 10k drives, Fibre Channel is the competitor in that market, not FireWire. FW is Good Enough(tm) to replace SCSI substems up SCS2 UW. Since it promises to attack the consumer market, FW hard drives shouldn't command the high premium that SCSI drives command.
Novell has a similar technology called DigitalMe. It does not carry 'e-cash', but allows you to enter any information in any site. I haven't tried using credit card information with it, but I'm sure it would work. Much more secure (and interesting) than the Microsoft offering.
Optical computer are not that far off (closer than quantum computers, anyway). Simple optical circuits have been created using modified current technology. Optical wells have beed demonstrated that allow seperation and amplification seperate wavelengths of light on a single waveguide. And of course, nobody knows what the military is up to, although optical computing is doubtless of great interest to them (resistant to EMP).
As far as Russian stuff goes, their equipment is very fault tolerant (even if a bit 'primitive'). They used vacuum tubes in their radar for a long time, both because they were immune to EMP, but also because they could handle higher power rates. The ELBRUS 2 (I think) was incredibly unreliable, it crashed every ten minutes. Of course, it rebooted the affected part, with little or no problem to the operating software. Military wise, it probably would not have been difficult for NATO forces to invade Kosovo. A diversionary strike against Vojvodina probably wouldn't be very costly, either. In Vojvodina, the plains are perfect for tank warfare, and since the plains spread into Hungary, there would be few natural obstacles. A tank battle between NATO M1A2 and Leopard tanks against Russian T-80's would probably be in favor of NATO. In Kosovo, troops on the ground would increase the hit rate exponentially. One other thing that isn't mentioned in the news very often, is that the majority of successful hits on Yugoslav forces were called in by the KLA (indirectly). Air attacks against mobile ground targets is not very effective unless you have a spotter on the ground to call in close air support. Ground forces also decrease the worry of AA, since the ground that is controlled by ground forces is (supposedly) free of enemy AA. Yugoslav comapnies selling weapons probably isn't a great idea to get an idea of the capabilities of the JNA, since many weapons were supplied by Russia (of course, Yugoslav companies could be offering surplus Russian weaponry, too). But yeah, the JNA was in better shape than the Iraqi army, and in a better defensive position. Of course, it things got too dirty, we could of just given the Ko sovars heavy weapons and let them do the dirty work. Cyberwar in Serbia took the form of killing TV stations, radio stations, power subsystems, etc. The primary weapon for taking out power stations was a carbon-fiber bomb. The carbon fiber would drift in the equipment, causing shorts everywhere. Non-destructive, but it takes time to correct. It disables it for the enemy, causing discomfort to the civilian population, but is easy to repair when reconstructing the country. Other uses would be extensions of old tactics like disinformation and propaganda. Overall, perhaps not as effective as hoped, but Milosevic did effectively lose Kosovo, and Serbia's infrastructure isn't exactly in great shape.
In a K-12 environment, you'd have to have a REALLY stupid sysadmin to allow any student to crack Solaris. It's much more secure than the Netware/DOS solutions that many schools use today. Anyways, most schools wouldn't store student data on the Ray server, but on seperate admin servers.
IEEE803.11 specs at 11Mb/s, fast enough for pretty much any application. The balls server as I/O points, with a large central computer on the space station that can handle voice recognition, and other high bandwidth/processing applications.
If you read back a few months, SGI signed an agreement with NVidia to cooperate on consumer graphics chipsets. I'll betcha that a good part of the NV10 (esp. the GPU) is SGI tech.
Hopefully they have something really kick-ass high-end for the own high-end graphics.
1. Don't know. They're probably smaller tha 3', so may not be subjected to 'eyesore' laws.
2. A signal has to get to the tower, and vice-versa. If you get a good bounce, line-of-sight isn't needed. (Line-of-sight in microwave isn't exactly the same as "ooh, I can see the tower from here!")
3. No. I doubt it very much.
4. Look at the map on their web site. Find a map of Tucson with distances on it. Compare the two. The farther away you are, the less bandwidth you get. They probably make sure that anybody within their 'service range' gets half-way decent bandwidth, though.
5. Depending on the backbone, very well. If you too many subscribers fora tower, you just add another tower. The only problem I see is that it looks like they use 8mbps uplins for the towers. This will get easily overwhelmed with >15 users. I imagine that they will upgrade to 45mbps links in the future.
6. Like cable. But all traffic is encrypted. Not the best crypto, but good enough for everybody except the government.
The primary speed problems with microkernels have had to do primarily with implementation. Don't forget that macrokernels have have decades to develop, while microkernels have primarily been in the experimental stage for the last ten years.
MkLinux (and NeXtStep) were based on Mach2.5, which uses several tricks to decrease the performance impact of microkernels. For the most part, it made Mach resemble a macrokernel more than a microkernel, but it was fast enough to allow NeXtStep to run on a 68040 25Mhz with an advanced GUI (with very good performance, I might add).
Mach 2.0 was an awful dog, probably because it was the first complete microkernel design for Mach. Mach 3.0 introduced several new concepts, but wasn't optimized for performance. Mach 4, although still slower than macrokernels, brought the performance difference to within a few percentage ponts.
The Mach design, however, was primarily experimental (despite its usage in many products), and the project was cancelled in the mid-90's (forget exactly when, though) Mach 4 continued to be developed by team at the UofUtah??, but dropped that in favor of developing a new microkernel based on the lessons learned from Mach.
The L4 microkernel has been designed with performance in mind (as well as easier programming), and is faster than Linux right now. However, it still doesn't handle multiple servers now (AFAIK), but does have a Linux server so it can emulate a Linux machine. Hurd will probably adapt it in place of Mach once the kernel becomes more stable and they feel comfortable mucking around with the internals.
Coda servers are available for Linux, tarballs, SRPMs and RPMs.
There are LDAP PAM modules so that you can use LDAP for authentication, as well as NISLDAP gateways.
I'm not at home, so I don't have any URLs, but I plan on trying to set up an integrated authentication/security system this fall at school.
Speaking of NIS+, I think that glibc2.1 has support for it, making it easier to configure. There is a server for Linux, but I'm not sure how stable it is (still beta, AFAIK).
Umm, actually, no. The new USB speakers (not FireWire, no such thing as FireWire speakers) are COMPLETELY digital. the D/A converts is in the speaker. You no longer need a sound card, since all of the conversion is done off-system, and software can emulate the rest (MIDI sampling, certain sound effects, etc.). It seems pretty cool.
Kosovo was an autonomous province of Serbia, with a relatively well integrated population. Somewhat poorer than the rest of yugoslavia, but not much. Serb nationalism started to scare some Kosovars, who began agitating for independence (actually greater autonomy until Slovenia, Croatia, et al. declared independence). In 1991?? (Not sure of exact date) Milosevich withdrew the autonomy of Kosovo, and began replacing Kosovars with Serbs in positions of power. Needless to say, such actions, combined with increasingly nationalistic Serb propaganda, increased the appeal of the KLA, which supported a violent conflict for Kosovar independence (earlier independence movements had ben centered around non-violent confrontation).
Milosevich used the NATO bombing as an excuse and cover for his plans of decreasing the Kosovar population. It was already planned and all set to go, all he needed was a cover, and the NATO bombing was perfect. Had NATO not bombed, Milosevich's plans would've been delayed for a bit, but the flow of refugees would happened eventually, with or without bombing.
I've never used one, but I've heard that Cyrix has really been helpful in getting the video working (full support in XFree) and theoretically the sound works too. There's probably some wierd option that needs to be set somewhere, though. I'd imagine that 2.2 would give you the best support, plus OSS commercial or ALSA.
Not if you use HTML formatting in the newsgroup. Sure, it sucks for people using primitive newsreaders, but there are hacks to view HTML for most newsreaders.
And the military-industrial complex is becoming an endangered species. Many have gone out of business or have had to merge to remain viable. With the increasing use of 'commercial' hardware, the military-idustrial complex has to compete with 'civilian' companies, and most companies have had to diversify out of pure military applications, or at least their mother companies have.
If anything, the military projects of Europe are more 'socialistic.' How many projects have been supported by France, GB, and Germany, even when costs have spiraled completely out of control? The governments funds inital research, as well as the cost of building, just to make sure the state-owned enterprise doesn't look too badly managed?
It would be fairly simple for them to build SMP machines that support pretty much any processor. The OpenPIC technology supported by AMD, Cyrix (and PPC?) supports multiple CPUs with no modifications. The problems lie in the bus, but even those can be worked out.
The big problem is that there is no OS that supports x86 OpenPIC. Although Linux could be modified, it would dramatically decrease the potential market. Windroids couldn't run NT (the whole reason for x86, right?), and the Mac people wouldn't be able to run SheepShaver, so they'd be left out, too.
OpenPIC, a SMP technology developed by Cyrix and AMD, supports up to 32 processors (theoretically), and all scheduling hardware is off-chip, so any x86 CPU can be used. If you were a masochist, it would bentirely possible to build a 32-processor SMP 8086 box!!! It looked farily easy to port to other CPU architectures, so other 'non-SMP' CPUs may have worked with OpenPIC (i.e. StrongARM).
It needed specific SMP support in the OS, and although several chipsets included full OpenPIC support, no board manufacturers included multiple sockets on their boards (can you say Intel and monopolistic pressures??). Since there wasn't any hardware support, it was kindof difficult for anyone to put the support into any OS.
*sigh* Sometimes I just hate the computer industry.
Already done. What do you think the SGI VW basically does? Memory, video, and CPUs are all one the same switched memory bus. 1.6Gbps (may be 3.2Gbps) directly from memory to video. Same for Sun machines and SGI Unix workstations.
The processors used are Alpha 21264 500Mhz CPUs. Approximately twice the SpecInt performance of a 400Mhz Xeon. There are two of them.
Memory bandwidth is twice as much as Intel.
66Mhz, 64-bit PCI slots. There is now way you can use a duplex GbEthernet in a PC, unless it is very low traffice (let alone the fact that there are no 32-bit 33Mhz adapters available)
Processor/memory bus is based on a three-way switch, making DMA transfers faster, and optimizing memory access and SMP.
They are built. The prices you have quoted are for components only, you didn't include the cost of assembly, testing, and burnin. The machine you listed does not have a 3-year, enterprise-level, on-site service contract. How much is that worth?
You're paying for the Compaq name.
Again, these are SERVERS, not workstations. The components are usually much higher quality than the components you purchase for your PC.
The Alpha CPU is 64-bit, great for huge databases and other memory-oriented or transaction-oriented applications.
FP kicks ass. You might by these for heavy-duty scientific applications.
For some applications, the PC you just built might be good, but for other applications the Compaq Alpha servers are better.
For a better comparison:
AMI MegaPlex Quad Xeon m/b 2MB built-in ATI video 8 32-bit PCI, 4 64-bit PCI, 2 ISA Includes custom chassis 3 600-watt power supplies 3 boards total
That is the rough equivalent of the DS120 system, which appears to be a bare-bones system. This board has more PCI slots, and can physically hold more memory, but the Xeon is pretty much the end of the line for the P6 core, while the 21264 has a much longer life ahead of it.
This is the core system, and I'm already beyond US$14,000. This is the very rough equivalent of a DS20
Mach4 has fixed most of the performance problems with Mach 2, 2.5, and 3. The new derivative of Mach, L4, is actually faster than macrokernels in many operations, and is just as fast in most other operations.
Having used NeXTstep on a 68040, the performance of the kernel really isn't going to be that important. If a rich GUI and apps can all run on a 68040, without lag, and realatively crisp response, performance on a G3 350 will not be a problem.
Kick-ass server performance probably won't scale as high as well as Linux, performance will still be good enough (and a lot better than NT).
Slashdot Mirror
Actually, JINI does not require a JVM, at least for network devices. Each device that wants to be identified by the JINI system somehow 'broadcasts' itself to the network, and contains a small Java 'driver'. Any devices with a user interface that want to use JINI devices must have a JVM, though. Any device that has a JVM can use any JINI device. Sun has been doing some pretty good work squeezing Java into embedded devices, and new Palms should be coming with an embedded JVM, as well as many future cell-phones, etc.
The brother of one of the guys I work with works with military satellite technology. While debugging the software on one of the sats, he took a picture of his motorcycle. Resolution was good enough to read the license plate! And this wasn't a blown-up picture, either.
If you are using 160/m and spending thousands on 10k drives, Fibre Channel is the competitor in that market, not FireWire. FW is Good Enough(tm) to replace SCSI substems up SCS2 UW. Since it promises to attack the consumer market, FW hard drives shouldn't command the high premium that SCSI drives command.
Novell has a similar technology called DigitalMe. It does not carry 'e-cash', but allows you to enter any information in any site. I haven't tried using credit card information with it, but I'm sure it would work. Much more secure (and interesting) than the Microsoft offering.
Optical computer are not that far off (closer than quantum computers, anyway). Simple optical circuits have been created using modified current technology. Optical wells have beed demonstrated that allow seperation and amplification seperate wavelengths of light on a single waveguide. And of course, nobody knows what the military is up to, although optical computing is doubtless of great interest to them (resistant to EMP).
As far as Russian stuff goes, their equipment is very fault tolerant (even if a bit 'primitive'). They used vacuum tubes in their radar for a long time, both because they were immune to EMP, but also because they could handle higher power rates. The ELBRUS 2 (I think) was incredibly unreliable, it crashed every ten minutes. Of course, it rebooted the affected part, with little or no problem to the operating software.
Military wise, it probably would not have been difficult for NATO forces to invade Kosovo. A diversionary strike against Vojvodina probably wouldn't be very costly, either. In Vojvodina, the plains are perfect for tank warfare, and since the plains spread into Hungary, there would be few natural obstacles. A tank battle between NATO M1A2 and Leopard tanks against Russian T-80's would probably be in favor of NATO. In Kosovo, troops on the ground would increase the hit rate exponentially. One other thing that isn't mentioned in the news very often, is that the majority of successful hits on Yugoslav forces were called in by the KLA (indirectly). Air attacks against mobile ground targets is not very effective unless you have a spotter on the ground to call in close air support. Ground forces also decrease the worry of AA, since the ground that is controlled by ground forces is (supposedly) free of enemy AA.
Yugoslav comapnies selling weapons probably isn't a great idea to get an idea of the capabilities of the JNA, since many weapons were supplied by Russia (of course, Yugoslav companies could be offering surplus Russian weaponry, too). But yeah, the JNA was in better shape than the Iraqi army, and in a better defensive position. Of course, it things got too dirty, we could of just given the Ko
sovars heavy weapons and let them do the dirty work.
Cyberwar in Serbia took the form of killing TV stations, radio stations, power subsystems, etc. The primary weapon for taking out power stations was a carbon-fiber bomb. The carbon fiber would drift in the equipment, causing shorts everywhere. Non-destructive, but it takes time to correct. It disables it for the enemy, causing discomfort to the civilian population, but is easy to repair when reconstructing the country. Other uses would be extensions of old tactics like disinformation and propaganda.
Overall, perhaps not as effective as hoped, but Milosevic did effectively lose Kosovo, and Serbia's infrastructure isn't exactly in great shape.
In a K-12 environment, you'd have to have a REALLY stupid sysadmin to allow any student to crack Solaris. It's much more secure than the Netware/DOS solutions that many schools use today. Anyways, most schools wouldn't store student data on the Ray server, but on seperate admin servers.
IEEE803.11 specs at 11Mb/s, fast enough for pretty much any application. The balls server as I/O points, with a large central computer on the space station that can handle voice recognition, and other high bandwidth/processing applications.
If you read back a few months, SGI signed an agreement with NVidia to cooperate on consumer graphics chipsets. I'll betcha that a good part of the NV10 (esp. the GPU) is SGI tech.
Hopefully they have something really kick-ass high-end for the own high-end graphics.
1. Don't know. They're probably smaller tha 3', so may not be subjected to 'eyesore' laws.
2. A signal has to get to the tower, and vice-versa. If you get a good bounce, line-of-sight isn't needed. (Line-of-sight in microwave isn't exactly the same as "ooh, I can see the tower from here!")
3. No. I doubt it very much.
4. Look at the map on their web site. Find a map of Tucson with distances on it. Compare the two. The farther away you are, the less bandwidth you get. They probably make sure that anybody within their 'service range' gets half-way decent bandwidth, though.
5. Depending on the backbone, very well. If you too many subscribers fora tower, you just add another tower. The only problem I see is that it looks like they use 8mbps uplins for the towers. This will get easily overwhelmed with >15 users. I imagine that they will upgrade to 45mbps links in the future.
6. Like cable. But all traffic is encrypted. Not the best crypto, but good enough for everybody except the government.
PReP and CHRP are different names for the same thing.
And according to one of my German friends, Indonesian is incrdibly similar to German. (Or at least the pronunciation, etc.)
The primary speed problems with microkernels have had to do primarily with implementation. Don't forget that macrokernels have have decades to develop, while microkernels have primarily been in the experimental stage for the last ten years.
MkLinux (and NeXtStep) were based on Mach2.5, which uses several tricks to decrease the performance impact of microkernels. For the most part, it made Mach resemble a macrokernel more than a microkernel, but it was fast enough to allow NeXtStep to run on a 68040 25Mhz with an advanced GUI (with very good performance, I might add).
Mach 2.0 was an awful dog, probably because it was the first complete microkernel design for Mach. Mach 3.0 introduced several new concepts, but wasn't optimized for performance. Mach 4, although still slower than macrokernels, brought the performance difference to within a few percentage ponts.
The Mach design, however, was primarily experimental (despite its usage in many products), and the project was cancelled in the mid-90's (forget exactly when, though) Mach 4 continued to be developed by team at the UofUtah??, but dropped that in favor of developing a new microkernel based on the lessons learned from Mach.
The L4 microkernel has been designed with performance in mind (as well as easier programming), and is faster than Linux right now. However, it still doesn't handle multiple servers now (AFAIK), but does have a Linux server so it can emulate a Linux machine. Hurd will probably adapt it in place of Mach once the kernel becomes more stable and they feel comfortable mucking around with the internals.
Coda servers are available for Linux, tarballs, SRPMs and RPMs.
There are LDAP PAM modules so that you can use LDAP for authentication, as well as NISLDAP gateways.
I'm not at home, so I don't have any URLs, but I plan on trying to set up an integrated authentication/security system this fall at school.
Speaking of NIS+, I think that glibc2.1 has support for it, making it easier to configure. There is a server for Linux, but I'm not sure how stable it is (still beta, AFAIK).
Umm, actually, no. The new USB speakers (not FireWire, no such thing as FireWire speakers) are COMPLETELY digital. the D/A converts is in the speaker. You no longer need a sound card, since all of the conversion is done off-system, and software can emulate the rest (MIDI sampling, certain sound effects, etc.). It seems pretty cool.
Kosovo was an autonomous province of Serbia, with a relatively well integrated population. Somewhat poorer than the rest of yugoslavia, but not much. Serb nationalism started to scare some Kosovars, who began agitating for independence (actually greater autonomy until Slovenia, Croatia, et al. declared independence). In 1991?? (Not sure of exact date) Milosevich withdrew the autonomy of Kosovo, and began replacing Kosovars with Serbs in positions of power. Needless to say, such actions, combined with increasingly nationalistic Serb propaganda, increased the appeal of the KLA, which supported a violent conflict for Kosovar independence (earlier independence movements had ben centered around non-violent confrontation).
Milosevich used the NATO bombing as an excuse and cover for his plans of decreasing the Kosovar population. It was already planned and all set to go, all he needed was a cover, and the NATO bombing was perfect. Had NATO not bombed, Milosevich's plans would've been delayed for a bit, but the flow of refugees would happened eventually, with or without bombing.
I've never used one, but I've heard that Cyrix has really been helpful in getting the video working (full support in XFree) and theoretically the sound works too. There's probably some wierd option that needs to be set somewhere, though. I'd imagine that 2.2 would give you the best support, plus OSS commercial or ALSA.
Not if you use HTML formatting in the newsgroup. Sure, it sucks for people using primitive newsreaders, but there are hacks to view HTML for most newsreaders.
And the military-industrial complex is becoming an endangered species. Many have gone out of business or have had to merge to remain viable. With the increasing use of 'commercial' hardware, the military-idustrial complex has to compete with 'civilian' companies, and most companies have had to diversify out of pure military applications, or at least their mother companies have.
If anything, the military projects of Europe are more 'socialistic.' How many projects have been supported by France, GB, and Germany, even when costs have spiraled completely out of control? The governments funds inital research, as well as the cost of building, just to make sure the state-owned enterprise doesn't look too badly managed?
It would be fairly simple for them to build SMP machines that support pretty much any processor. The OpenPIC technology supported by AMD, Cyrix (and PPC?) supports multiple CPUs with no modifications. The problems lie in the bus, but even those can be worked out.
The big problem is that there is no OS that supports x86 OpenPIC. Although Linux could be modified, it would dramatically decrease the potential market. Windroids couldn't run NT (the whole reason for x86, right?), and the Mac people wouldn't be able to run SheepShaver, so they'd be left out, too.
OpenPIC, a SMP technology developed by Cyrix and AMD, supports up to 32 processors (theoretically), and all scheduling hardware is off-chip, so any x86 CPU can be used. If you were a masochist, it would bentirely possible to build a 32-processor SMP 8086 box!!! It looked farily easy to port to other CPU architectures, so other 'non-SMP' CPUs may have worked with OpenPIC (i.e. StrongARM).
It needed specific SMP support in the OS, and although several chipsets included full OpenPIC support, no board manufacturers included multiple sockets on their boards (can you say Intel and monopolistic pressures??). Since there wasn't any hardware support, it was kindof difficult for anyone to put the support into any OS.
*sigh* Sometimes I just hate the computer industry.
Sun workstations from $2500, but w/ no SCSI
Already done. What do you think the SGI VW basically does? Memory, video, and CPUs are all one the same switched memory bus. 1.6Gbps (may be 3.2Gbps) directly from memory to video. Same for Sun machines and SGI Unix workstations.
The processors used are Alpha 21264 500Mhz CPUs. Approximately twice the SpecInt performance of a 400Mhz Xeon. There are two of them.
Memory bandwidth is twice as much as Intel.
66Mhz, 64-bit PCI slots. There is now way you can use a duplex GbEthernet in a PC, unless it is very low traffice (let alone the fact that there are no 32-bit 33Mhz adapters available)
Processor/memory bus is based on a three-way switch, making DMA transfers faster, and optimizing memory access and SMP.
They are built. The prices you have quoted are for components only, you didn't include the cost of assembly, testing, and burnin. The machine you listed does not have a 3-year, enterprise-level, on-site service contract. How much is that worth?
You're paying for the Compaq name.
Again, these are SERVERS, not workstations. The components are usually much higher quality than the components you purchase for your PC.
The Alpha CPU is 64-bit, great for huge databases and other memory-oriented or transaction-oriented applications.
FP kicks ass. You might by these for heavy-duty scientific applications.
For some applications, the PC you just built might be good, but for other applications the Compaq Alpha servers are better.
For a better comparison:
AMI MegaPlex Quad Xeon m/b
2MB built-in ATI video
8 32-bit PCI, 4 64-bit PCI, 2 ISA
Includes custom chassis
3 600-watt power supplies
3 boards total
~ US$10,600
2 Intel Xeon 400Mhz 1MB cache
2 @ ~ US$1,995 == ~US$3,390
128MB ECC PC100 SDRAM DIMM
~ US$200
Total: ~US$14,190
That is the rough equivalent of the DS120 system, which appears to be a bare-bones system. This board has more PCI slots, and can physically hold more memory, but the Xeon is pretty much the end of the line for the P6 core, while the 21264 has a much longer life ahead of it.
This is the core system, and I'm already beyond US$14,000. This is the very rough equivalent of a DS20
Mach4 has fixed most of the performance problems with Mach 2, 2.5, and 3. The new derivative of Mach, L4, is actually faster than macrokernels in many operations, and is just as fast in most other operations.
Having used NeXTstep on a 68040, the performance of the kernel really isn't going to be that important. If a rich GUI and apps can all run on a 68040, without lag, and realatively crisp response, performance on a G3 350 will not be a problem.
Kick-ass server performance probably won't scale as high as well as Linux, performance will still be good enough (and a lot better than NT).