Most countries, including the US, are signatories to the Berne convention and have aligned their copyritght law to the convention. That means that a work is owned by its author even if there is no explicit copyright notification or author's name: that in turn means that if you copy an anonymous work, the actual author may later sue you for copyright violation.
And wrong again: if you somehow get confused and think the work is owned by the FSF, then you would need to get permission from the FSF before copying it under a license different than the GPL. When you contacted them they would send you a nice letter telling you how wrong you are.
Scenario:
EoSCo (Evil or Stupid Company) copies GPL'ed code from SRP (Some Random Project) and delivers in in binary-only form.
RI (Random Investigator) finds fingerprints of SRP in the binary and tells the workd.
JRC ( J. Random Coder,) a contributor to SRP, notices RI's blog entry and sends a letter to EoSCo: "You are violating my copyright. Please stop."
EoSCo ignores Letter.
JRC Sues EoSCo for copyright violation.
Judge: Do you have a license from JRC to copy this stuff?
EoSCo: Your Honor, we thought it belonged to FSF
Judge: So you have a license from FSF?
EoSCo: well no...
Judge: You have violated JRC's Copyright. Cease using this code at once and pay damages.
Note: each contributor to SRP owns his own copyrights, unless he assigns them to the SRP. Thus, EoSCO is vulnerable to each of the contributors. It only takes one single contributor to kill EoSCo's illegal product.
I just measured my current cell phone:
it is 86x47x22 mm. This is an 3-year-old Samsung, not a big old qualcomm.
The ROAD is 128x60x25mm
I abandoned my belt holster three years ago when I shifted to the Samsung. Qualitatively, the ROAD is on the bleeding edge of being too large to keep in your pocket if you wear guy-type casual clothes. I guess it's back to wearing the phone in a belt holster if I move to the ROAD.
On the other hand, 128mm looks like an absolute minimum width for a QWERTY keyboard, and 47mm is really short for a usable screen, so unless we go virtual, it looks like a belt holster is mandatory.
My service provider is Verizon. No GSM. Will ROAD eventually have CDMA, or will the US eventually have GSM?
Based on the links yoiu provide, the 10K drives have a bulk transfer rate of 59 to 118MB/sec and the 7200RPM drives have a bulk transfer rate of 760Mb/s, equivalent to 95MB/s. That is, the 10K drives are at best marginally faster than the 7200RPM drives.
Applications that are truly latency-sensitive can generally best be served by using large RAM cache.
The "big" 10K disks in your links are still 250GB.
1) Disk perforamnce gains have outpaced CPU performance gains for at least the last decade.
2) The author simply does not understand HD design constraints. For a given RPM, the data transfer speed increases as the density per platter increases. This is constrained by the Magical electronics that read and write the bits on the disk. So, twice the density also implies twice the bulk data transfer rate (not the burst rate.)
3) SATA. SATA is now being sold at (or very near) the price of EIDE. Last A year ago SATA sold at a premium of $20-$30/drive. By the end of 2005, SATA will be cheaper than EIDE for otherwise-equal drives.
4) Price. Price/gig went from $1.00 at the beginning of 2004 to $.50 at the beginning of 2005, at the "sweet spot." The current "sweet spot" is 250GB. There is not reason to doubt that the price/Gig will reach $.25 by the end of the year.
5) interest in 10K and 15K RPM is misplaced for most applications. Speed affects rotational delay and nothing else. Bulk transfer rate is more important in most applications (point 2 above.)
If it spins twice as fast but has half the density, it has the same bulk transfer rate.
6) interest in SCSI is outdated. SATA with one (competent) controller per disk has better characeristics.
Optical is unreliable for stratellite-to-ground connections, due to weather. However, there is effectively no weather at 65K feet, so the stratellites can talk to each other using lasers. These guys are claiming a 500-mile line-of-sight visiblity to the ground, but if you read the fine print, the effective radius to the ground is really only 75 miles. Of coures, 75 Miles is truly impressive. The 500-mile number is still important, because it is a good approximation for the distance at which a stratellite can "see" another stratellite using a laser, above the highest clouds.
Consider a grid of these beasties to cover a region of 3000mi x 2000 mi (i.e., the contenental US.) This will require 6x4= 24 Stratellites. Using lasers, this grid can carry all long-haul traffic. Local loop (WiFi, 75-mi radius) would require 3000x2000/(75x75) = (30x20x4/3x4/3)= 600x16/9)= (3200/3) = about 1000 Stratellites. However, the population density is massively skewed, so we can use expensive antennas in rural areas and cheap antennas in urban areas. Using a rural antenna, we can "see" a stratellite that is 500mi away, so we need 24 stratellites to handle rural antennas. we now add "urban' stratellites as needed.
Conclusion: 24 Stratellites will cover the US, giving base coverage of (say) 20 Kbps/sq mi for non-urban areas. We then add additional stratellites based on population density at (say) 100Kbps/person. Worst-case we need 1000 stratellites, but we probably end up with about 100 stratellites to cover the population.
The fundamental enabler for using an optical transition as a time stand is a device called the Femtosecond comb laser. This is buried way down in the article.
The reason we currently use the cesium transition is that it was the best precise atomic transition we know about that ws slow enough to count using the available counting technology. the oscillation speed is near 10Ghz.
We can now count this directly rather than through an elaborate divider chain, which is why the accuracy has improved by five orders of magnitude over the past forty years.
Now, we can use the femtosecond comb as a pahse-locked divider of extreme precision. THis lets us "count" optical transitions by mode-locking a pulsed lasser to the optical reference. This in turn permit us to use a mercury transition or a strontiou transition that is 3 orders of magnigude faster than the cesium transition.
The difference between the mercury transitins and the strontium transition (a factor of three) is utterly trivial by comparison to the factor of 1000 gain provided by the femtosecond comb.
Dr. Hawking is perhaps the most briliant living physicist. From all accounts he has many sterling qualities, and I admire him. However, his various disabilities prevent him from being "graceful" by most meanings of the word.
In his chosen environment, we can ignore his physical impairments and evaluate him objectively. In this environment, at this time, he was "gracious", not "graceful."
ISP should shut off port 25, because it defends the rest of us from the clueless.
However, if your ISP blocks prot 25 and you have a legitimate reason to use a different MTA, you can still do so by having the administrator of the MTA open a port other than 25. for example, you and several of your friends can get together and rent a cheap server somewhere on the internet (e.g., www.linode.com, $20/mo) and run your own MTA (sendmail or postfix.) You can either set up a VPN connection via SSH, or simply open a separate port and then change the settings on your e-mail clients to send to that port instead of port 25.
As the administrator of the MTA, you will of course restrict the use of this port to only you any your friends. Note that your e-mail will no longer originate from the blocked ISP, but from your own tiny little home on the net. OF course you will need to rent your server from an organization that enforces a serious anit-apam policy, or they may get black-holed themselves.
True. Somewhere beteen my brain and the "submit" button, I dropped the assumption that there would be ten data centers, each with a router. Note that
the equivalent of these routers already exists in a distributes fashion to accomodate the Akamai nodes. THe basic point remains: Akamai (or a competitor) could provide its true service (dynamic, automatic expansion of server capacity) more efficiently if it would jsut get past its focus on "locality." "Locality" solves a non-problem.
Everything you say is completely valid, but you still miss my point. Akamai could provide all of these services from a small number of large data centers rather than a large numger of small data centers. This is why Akamai's costs are higher than Google's costs, and will remain so.
When you are in a large number of places, you cannot take advantage of economies of scale in CPU upgrade, etc.
Sorry, but Google is a case in point, and Google serves more data than download.microoft.com, I think. Akami has 14K servers. If you put them all in one place and give them 100Mbps each, and it's all used at the same time, it is 1400Gpbs. That's fourteen 10Gpbs ports on a core router facing the data center plus 14 ports facing the internet. That's not enough to fill up a single Juniper T640, or a single DWDM fiber.
Akamai's fundamental premis is flawed. The premis is that core bandwidth is scarce, so high-hit web pages should be replicated "locally." Therefore, Akamai servers are scattered all over the place.
By contrast, Google has a whole bunch of computers in each of a very few places. This completely changes the economics.
The reason Akamai's premis is flawed is simple: core bandwidth is cheap, because the core was overbuilt during the bubble and because of the incredible advances in core technologies. By contrast, the last mile is still constrained, primarily because of monopolies and politics.
The effect of this is that once your packet gets from your house to the first router, the rest of the internet is all effectively an equal cost from you.
I just did a google for
programmer salary.
A 25th percentile proigrammer in the US makes 42K/yr.
Ther appear to be thousands of worthwhile resources on the web to answer this question.
Why ask Slashdot?
There are two parts to this technology: the very low power to and from the phones relatively immobile phones (i.e.,the pico-cell,) and the sophisticated uplink. This particular uplink is sophisticated and is useful for aircraft. Forget it and concentrate on the pico-cell.
I want one of these in my house, with an internet uplink via my broadband connection. Then, when someone calls me on my cell phone at home, I will have a consiten and reliable cell connection that does not depend on my (distant, overloaded) cell tower. Furthermore, I will be using pico-power, reducing the amount of environmental radio clutter in my house and in my neighborhood.
Same at the office. My conpany could put in a pico-cell wiht more channels and support all the phones in the office. The wireless company could install pico-cells in hundreds of offices for the cost of a single cell tower, thus reducing demand on the cell towers. Major win-win for everybody.
Of course, we may be able to get the saem effect with WIFI-enabled cell phones and sophisticated call forwarding.
SUre. the same pico-cell technology that works on planes can work on trains, busses, and subways. The advantages to the system operator are big enough that they need not charge more for this service. Baseically, throws a fairly large burden on the system when a bunch of people move between two cells at the same time. The pico-cell uses a different up-link (varying with type of pico-cell) to connect the entire passenger load to a single location. The phone owner benefits by better call quality, much longer battery life, and less radio energy in the environment.
I just did a google for T-rays. "T-rays" (terahertz rays) are longer than infrared, shorter than microwave. They have nothing whatsoever to do with photonics. Cross-check: C=300Mm/s divide by 1Thz:
wavelength = 300 micrometers, or 300,000nm, or about one thousand times longer than a visible photon, 2000 times bigger than the wavelengths used in lithography.
I think you may have confused the transmission bandwidth of the transistor (1Thz) with its switching speed.
The electronics are the bottleneck because switching is the bottleneck, and the electronic switches are the fastest available. Photonics is only useful for transmission, not switching. This is not to disparage the incredible advances that were made in photonics over the last decade.
Currently, photonics is used primarily for communications. Long-haul optical fiber operates o at 1500nm and 1300nm. Short-reach optics (VCSELs) operate at 800nm. Optics in silicon dioxide glasses can operate from these infra-red ranges down to about 200nm. To go below that, you will need to find materials that are transparent (for "conductors") and optically active (for switches and amplifiers) at these smaller wavelengths. Finding such materials for use at the macro scale has been a formidable task. Finding such materials for use as a lithographic substrate is likely to be a lot harder.
So yes, I was making the assumption that micro-photonics would be based on a silicon dioxide (glass) substrate, which is constrained to operate above 200nm. Current "integrated optics" (AWGs and SOAs) are made on a silicon dioxide substrate.
One alternative might be diamond. I don't know what its transparent cutoff is.
The industry has been experimenting with liquid interfaces for the last several years. I don't know what's supposed to be new about the RIT setup.
Each size reduction calls for the use of higher-frequency light. This calls for lenses that can operate at those frequencies. We are already in the ultraviolet, where ordinary glass is opaque. The industry is currently using 193nm light to create 90nm features, in violation of simple optics. The industry planned to go to 153nm light, but the only transparent lens material at that wavelength is CaF (calcium flouride) and CaF lenses were simply too hard to make reliably. By the rules of simple optics, you cannot create an image with higher resolution than the wavelength of the light you are using. The industry therefore uses "phase shift" masks instead of "binary" masks. In a "binary"mask, each point on the mask is either opaque or clear. A phase-shift mask has other levels of transparency. A phase-shift mask causes the light to interfere with itself to create the required pattern on the chip, and developing a phase-shift mask to create your desired pattern is an extremely complex and expensive proposition: a mask set can cost a million dollars, so unless you make a lot of identical chips, the cost of hte mask is a noticable chunk of the cost of the chip.
Lenses:
A lens focuses light because of differences in the index of refraction between the lens material and the material surrounding the lens. The higher the difference, the higher the magnification. The system with the liquid (water, oil, whatever) Also has a lens. The water allows higher magnification without higher curvature. High curvature causes lots of bad things.
Chemistry:
Picking hte right liquid is hard. The liquid must be transparent to the photons you are using.It must not react witht he material of the lens, mask, or any of the substrate materials (resist, metal, orsemiconductor, chamber) and it must be reasonably environmentally benign. Fortunately, water meets the requirement for most systems.
Alternatives:
The industry has been trying to get our of this mess for at least 20 years. The alternatives are e-beam and X-ray. E-beam is too slow, X-ray has a host of truly complex problems.
Sorry, no.
Optical chips are less dense than electronic chips, for the simple reason that photons are "bigger" than electrons, A visible photon has a wavelength of about 800nm, so a waveguide for this photon must have roughly equivalent dimensions.
From a "speed" standpoint, light in glass moves at C/I (where I is the index of refraction of the material. For optical glass, I=1.5 or thereabouts, so light moves at two thirds the speed of light in a vacuum. electrical pulses in copper move at about.87 C, while radio frequencies in a "transmission line" such as a coaxial cable move at nearly C. Thus, an optical signal on a fiber travels at about 200Km/ms, while a radio signal on a coaxial cable travels at about 300Km/ms.
Electronic switches (transistors) can operate up to about 150Ghz. (in the lab, bleeding-edge.) These decvices are 90nm or so. Photonic switches can operate at perhaps 40Ghz. These devices are big discrete components, and even if they are miniaturized they will need to be the at least the size of a photon wavelength, ten times bigger than the electronic switch in each dimension.
Photonics is a wonderful science, and will lead further dramatic decreases in cost of communications, but the physics is all wrong for replacing electrons with photons as a way to miniaturize a computer. Look to nanotech for that, but that's another story.
Is it just me, or has sun lost its credibility in the open source community? They seem to be very ambivalent about Open Source, and I no longer trust them.
Linux: Sun says yes/no/maybe
Java License: Sun can't decide
SCO: Sun has a relationship with these scuzballs, or maybe not.
I'll just wait for an unambiguously open 3D desktop.
Costs $80.00 on e-bay, has web browser, and is hackable. Just wire it to your LAN and feed it from any random conputer in your house.
It can also do a whole bunch of other stuff.
Nope. A Cisco 12000 has 16 ports of 10Gbps. It receives 160Gbps, and transmits 160Gbps, so it "can move" only 160Gbps. Cisco calls this "320Gbps." the internet industry calls this "Cisco math."
A Juniper T540 has 32 lines of 10Gbps, so it can actually "move" 320Gbps.
A big Avici router can do better.
You are correct: your post is a gross mischaracterization. SCO's latest motion is in oposition to IBM's motion to compel. IBM had to file the motion to compel because SCO ws being unresponsive. It is SCO that is dragging their feet. IBM has moved this case along as briskly as possible.
This is the very first time that SCO has aledged that IBM is delaying, and it is clearly a cynical ploy on SCO's part. If they really thought IBM was delaying or that IBM's responses to discovery were inadequate, SCO would file their own motion to compel. They have not done this, beause IBM has in fact been responsive.
In a statement in late May, quoted in a CNET article on 28 May, Darryl McBride of SCO said that the SCO lawyers are are working on a contingency fee basis. SCO is not paying the lawyers anything. IFthey win, the lawyers get a big cut of the proceeds.
If SCO gets counter-sued for barratry, I hope their lawyers are required to pay an equivalent percentage of the penalty.
Slashdot Mirror
Gee, that was really cheap. Most of us use more than 600Mbit in a few seconds. OHHHH! perhaps you meant Mbps, not Mbit......
Scenario:
EoSCo (Evil or Stupid Company) copies GPL'ed code from SRP (Some Random Project) and delivers in in binary-only form.
RI (Random Investigator) finds fingerprints of SRP in the binary and tells the workd.
JRC ( J. Random Coder,) a contributor to SRP, notices RI's blog entry and sends a letter to EoSCo: "You are violating my copyright. Please stop."
EoSCo ignores Letter.
JRC Sues EoSCo for copyright violation.
Judge: Do you have a license from JRC to copy this stuff?
EoSCo: Your Honor, we thought it belonged to FSF
Judge: So you have a license from FSF?
EoSCo: well no...
Judge: You have violated JRC's Copyright. Cease using this code at once and pay damages.
Note: each contributor to SRP owns his own copyrights, unless he assigns them to the SRP. Thus, EoSCO is vulnerable to each of the contributors. It only takes one single contributor to kill EoSCo's illegal product.
The ROAD is 128x60x25mm
I abandoned my belt holster three years ago when I shifted to the Samsung. Qualitatively, the ROAD is on the bleeding edge of being too large to keep in your pocket if you wear guy-type casual clothes. I guess it's back to wearing the phone in a belt holster if I move to the ROAD.
On the other hand, 128mm looks like an absolute minimum width for a QWERTY keyboard, and 47mm is really short for a usable screen, so unless we go virtual, it looks like a belt holster is mandatory.
My service provider is Verizon. No GSM. Will ROAD eventually have CDMA, or will the US eventually have GSM?
Based on the links yoiu provide, the 10K drives have a bulk transfer rate of 59 to 118MB/sec and the 7200RPM drives have a bulk transfer rate of 760Mb/s, equivalent to 95MB/s. That is, the 10K drives are at best marginally faster than the 7200RPM drives. Applications that are truly latency-sensitive can generally best be served by using large RAM cache. The "big" 10K disks in your links are still 250GB.
1) Disk perforamnce gains have outpaced CPU performance gains for at least the last decade. 2) The author simply does not understand HD design constraints. For a given RPM, the data transfer speed increases as the density per platter increases. This is constrained by the Magical electronics that read and write the bits on the disk. So, twice the density also implies twice the bulk data transfer rate (not the burst rate.) 3) SATA. SATA is now being sold at (or very near) the price of EIDE. Last A year ago SATA sold at a premium of $20-$30/drive. By the end of 2005, SATA will be cheaper than EIDE for otherwise-equal drives. 4) Price. Price/gig went from $1.00 at the beginning of 2004 to $.50 at the beginning of 2005, at the "sweet spot." The current "sweet spot" is 250GB. There is not reason to doubt that the price/Gig will reach $.25 by the end of the year. 5) interest in 10K and 15K RPM is misplaced for most applications. Speed affects rotational delay and nothing else. Bulk transfer rate is more important in most applications (point 2 above.) If it spins twice as fast but has half the density, it has the same bulk transfer rate. 6) interest in SCSI is outdated. SATA with one (competent) controller per disk has better characeristics.
Optical is unreliable for stratellite-to-ground connections, due to weather. However, there is effectively no weather at 65K feet, so the stratellites can talk to each other using lasers. These guys are claiming a 500-mile line-of-sight visiblity to the ground, but if you read the fine print, the effective radius to the ground is really only 75 miles. Of coures, 75 Miles is truly impressive. The 500-mile number is still important, because it is a good approximation for the distance at which a stratellite can "see" another stratellite using a laser, above the highest clouds. Consider a grid of these beasties to cover a region of 3000mi x 2000 mi (i.e., the contenental US.) This will require 6x4= 24 Stratellites. Using lasers, this grid can carry all long-haul traffic. Local loop (WiFi, 75-mi radius) would require 3000x2000/(75x75) = (30x20x4/3x4/3)= 600x16/9)= (3200/3) = about 1000 Stratellites. However, the population density is massively skewed, so we can use expensive antennas in rural areas and cheap antennas in urban areas. Using a rural antenna, we can "see" a stratellite that is 500mi away, so we need 24 stratellites to handle rural antennas. we now add "urban' stratellites as needed. Conclusion: 24 Stratellites will cover the US, giving base coverage of (say) 20 Kbps/sq mi for non-urban areas. We then add additional stratellites based on population density at (say) 100Kbps/person. Worst-case we need 1000 stratellites, but we probably end up with about 100 stratellites to cover the population.
The fundamental enabler for using an optical transition as a time stand is a device called the Femtosecond comb laser. This is buried way down in the article. The reason we currently use the cesium transition is that it was the best precise atomic transition we know about that ws slow enough to count using the available counting technology. the oscillation speed is near 10Ghz. We can now count this directly rather than through an elaborate divider chain, which is why the accuracy has improved by five orders of magnitude over the past forty years. Now, we can use the femtosecond comb as a pahse-locked divider of extreme precision. THis lets us "count" optical transitions by mode-locking a pulsed lasser to the optical reference. This in turn permit us to use a mercury transition or a strontiou transition that is 3 orders of magnigude faster than the cesium transition. The difference between the mercury transitins and the strontium transition (a factor of three) is utterly trivial by comparison to the factor of 1000 gain provided by the femtosecond comb.
Dr. Hawking is perhaps the most briliant living physicist. From all accounts he has many sterling qualities, and I admire him. However, his various disabilities prevent him from being "graceful" by most meanings of the word. In his chosen environment, we can ignore his physical impairments and evaluate him objectively. In this environment, at this time, he was "gracious", not "graceful."
ISP should shut off port 25, because it defends the rest of us from the clueless. However, if your ISP blocks prot 25 and you have a legitimate reason to use a different MTA, you can still do so by having the administrator of the MTA open a port other than 25. for example, you and several of your friends can get together and rent a cheap server somewhere on the internet (e.g., www.linode.com, $20/mo) and run your own MTA (sendmail or postfix.) You can either set up a VPN connection via SSH, or simply open a separate port and then change the settings on your e-mail clients to send to that port instead of port 25. As the administrator of the MTA, you will of course restrict the use of this port to only you any your friends. Note that your e-mail will no longer originate from the blocked ISP, but from your own tiny little home on the net. OF course you will need to rent your server from an organization that enforces a serious anit-apam policy, or they may get black-holed themselves.
True. Somewhere beteen my brain and the "submit" button, I dropped the assumption that there would be ten data centers, each with a router. Note that the equivalent of these routers already exists in a distributes fashion to accomodate the Akamai nodes. THe basic point remains: Akamai (or a competitor) could provide its true service (dynamic, automatic expansion of server capacity) more efficiently if it would jsut get past its focus on "locality." "Locality" solves a non-problem.
Everything you say is completely valid, but you still miss my point. Akamai could provide all of these services from a small number of large data centers rather than a large numger of small data centers. This is why Akamai's costs are higher than Google's costs, and will remain so. When you are in a large number of places, you cannot take advantage of economies of scale in CPU upgrade, etc.
Sorry, but Google is a case in point, and Google serves more data than download.microoft.com, I think. Akami has 14K servers. If you put them all in one place and give them 100Mbps each, and it's all used at the same time, it is 1400Gpbs. That's fourteen 10Gpbs ports on a core router facing the data center plus 14 ports facing the internet. That's not enough to fill up a single Juniper T640, or a single DWDM fiber.
By contrast, Google has a whole bunch of computers in each of a very few places. This completely changes the economics.
The reason Akamai's premis is flawed is simple: core bandwidth is cheap, because the core was overbuilt during the bubble and because of the incredible advances in core technologies. By contrast, the last mile is still constrained, primarily because of monopolies and politics.
The effect of this is that once your packet gets from your house to the first router, the rest of the internet is all effectively an equal cost from you.
I just did a google for programmer salary. A 25th percentile proigrammer in the US makes 42K/yr. Ther appear to be thousands of worthwhile resources on the web to answer this question. Why ask Slashdot?
There are two parts to this technology: the very low power to and from the phones relatively immobile phones (i.e.,the pico-cell,) and the sophisticated uplink. This particular uplink is sophisticated and is useful for aircraft. Forget it and concentrate on the pico-cell. I want one of these in my house, with an internet uplink via my broadband connection. Then, when someone calls me on my cell phone at home, I will have a consiten and reliable cell connection that does not depend on my (distant, overloaded) cell tower. Furthermore, I will be using pico-power, reducing the amount of environmental radio clutter in my house and in my neighborhood. Same at the office. My conpany could put in a pico-cell wiht more channels and support all the phones in the office. The wireless company could install pico-cells in hundreds of offices for the cost of a single cell tower, thus reducing demand on the cell towers. Major win-win for everybody. Of course, we may be able to get the saem effect with WIFI-enabled cell phones and sophisticated call forwarding.
SUre. the same pico-cell technology that works on planes can work on trains, busses, and subways. The advantages to the system operator are big enough that they need not charge more for this service. Baseically, throws a fairly large burden on the system when a bunch of people move between two cells at the same time. The pico-cell uses a different up-link (varying with type of pico-cell) to connect the entire passenger load to a single location. The phone owner benefits by better call quality, much longer battery life, and less radio energy in the environment.
I think you may have confused the transmission bandwidth of the transistor (1Thz) with its switching speed.
The electronics are the bottleneck because switching is the bottleneck, and the electronic switches are the fastest available. Photonics is only useful for transmission, not switching. This is not to disparage the incredible advances that were made in photonics over the last decade.
So yes, I was making the assumption that micro-photonics would be based on a silicon dioxide (glass) substrate, which is constrained to operate above 200nm. Current "integrated optics" (AWGs and SOAs) are made on a silicon dioxide substrate.
One alternative might be diamond. I don't know what its transparent cutoff is.
The industry has been experimenting with liquid interfaces for the last several years. I don't know what's supposed to be new about the RIT setup. Each size reduction calls for the use of higher-frequency light. This calls for lenses that can operate at those frequencies. We are already in the ultraviolet, where ordinary glass is opaque. The industry is currently using 193nm light to create 90nm features, in violation of simple optics. The industry planned to go to 153nm light, but the only transparent lens material at that wavelength is CaF (calcium flouride) and CaF lenses were simply too hard to make reliably. By the rules of simple optics, you cannot create an image with higher resolution than the wavelength of the light you are using. The industry therefore uses "phase shift" masks instead of "binary" masks. In a "binary"mask, each point on the mask is either opaque or clear. A phase-shift mask has other levels of transparency. A phase-shift mask causes the light to interfere with itself to create the required pattern on the chip, and developing a phase-shift mask to create your desired pattern is an extremely complex and expensive proposition: a mask set can cost a million dollars, so unless you make a lot of identical chips, the cost of hte mask is a noticable chunk of the cost of the chip. Lenses: A lens focuses light because of differences in the index of refraction between the lens material and the material surrounding the lens. The higher the difference, the higher the magnification. The system with the liquid (water, oil, whatever) Also has a lens. The water allows higher magnification without higher curvature. High curvature causes lots of bad things. Chemistry: Picking hte right liquid is hard. The liquid must be transparent to the photons you are using.It must not react witht he material of the lens, mask, or any of the substrate materials (resist, metal, orsemiconductor, chamber) and it must be reasonably environmentally benign. Fortunately, water meets the requirement for most systems. Alternatives: The industry has been trying to get our of this mess for at least 20 years. The alternatives are e-beam and X-ray. E-beam is too slow, X-ray has a host of truly complex problems.
Sorry, no. Optical chips are less dense than electronic chips, for the simple reason that photons are "bigger" than electrons, A visible photon has a wavelength of about 800nm, so a waveguide for this photon must have roughly equivalent dimensions. From a "speed" standpoint, light in glass moves at C/I (where I is the index of refraction of the material. For optical glass, I=1.5 or thereabouts, so light moves at two thirds the speed of light in a vacuum. electrical pulses in copper move at about .87 C, while radio frequencies in a "transmission line" such as a coaxial cable move at nearly C. Thus, an optical signal on a fiber travels at about 200Km/ms, while a radio signal on a coaxial cable travels at about 300Km/ms.
Electronic switches (transistors) can operate up to about 150Ghz. (in the lab, bleeding-edge.) These decvices are 90nm or so. Photonic switches can operate at perhaps 40Ghz. These devices are big discrete components, and even if they are miniaturized they will need to be the at least the size of a photon wavelength, ten times bigger than the electronic switch in each dimension.
Photonics is a wonderful science, and will lead further dramatic decreases in cost of communications, but the physics is all wrong for replacing electrons with photons as a way to miniaturize a computer. Look to nanotech for that, but that's another story.
Is it just me, or has sun lost its credibility in the open source community? They seem to be very ambivalent about Open Source, and I no longer trust them. Linux: Sun says yes/no/maybe Java License: Sun can't decide SCO: Sun has a relationship with these scuzballs, or maybe not. I'll just wait for an unambiguously open 3D desktop.
Costs $80.00 on e-bay, has web browser, and is hackable. Just wire it to your LAN and feed it from any random conputer in your house. It can also do a whole bunch of other stuff.
Nope. A Cisco 12000 has 16 ports of 10Gbps. It receives 160Gbps, and transmits 160Gbps, so it "can move" only 160Gbps. Cisco calls this "320Gbps." the internet industry calls this "Cisco math." A Juniper T540 has 32 lines of 10Gbps, so it can actually "move" 320Gbps. A big Avici router can do better.
You are correct: your post is a gross mischaracterization. SCO's latest motion is in oposition to IBM's motion to compel. IBM had to file the motion to compel because SCO ws being unresponsive. It is SCO that is dragging their feet. IBM has moved this case along as briskly as possible. This is the very first time that SCO has aledged that IBM is delaying, and it is clearly a cynical ploy on SCO's part. If they really thought IBM was delaying or that IBM's responses to discovery were inadequate, SCO would file their own motion to compel. They have not done this, beause IBM has in fact been responsive.
In a statement in late May, quoted in a CNET article on 28 May, Darryl McBride of SCO said that the SCO lawyers are are working on a contingency fee basis. SCO is not paying the lawyers anything. IFthey win, the lawyers get a big cut of the proceeds. If SCO gets counter-sued for barratry, I hope their lawyers are required to pay an equivalent percentage of the penalty.