Or better yet, copy some data from/dev/random to a file.
The beauty of good encryption is that it generates seemingly random output. In theory, almost any random stream can generate any message so it would be possible for the authorities to mistakenly identify random data for an encrypted file.
And such a random block might be the encryption pad for something else. Here is an idea: use encrypted file as encryption pads for other files... this way, authorities would have no way to prove wether or not the pad has any meaning but you can at least prove that the file itself is needed to decrypt some other file.
It is open season for abuse. There are too many ways for people to hide meaning and too many for authorities to want meaning where there is none.
And what are they supposed to do with ephemeral/intermediate session keys that are not saved anywhere and usually never shown to the user?
And what about the case where the user simply and honestly forgot the encryption password? Does this mean that bad memory will become a crime?
If this could mean better, cheaper XP Pro/Server and the demise of Windows XP Crippled Editions (Starter and Home), I would be fairly happy.
Now, I wonder why large companies would have such a large proportion of Mac users. My guess is this is largely thanks to large corporations' overweight marketing departments. With companies ditching engineering and manufacturing, marketing is the only thing still going strong and the graphists they employ typically love Apple stuff.
They started rebranding Luxeon LEDs and Tektronix/Fluke/etc. instruments?
It does suck that most Dell, HP and others' stuff is now made by someone else. What is the point of buying any American brand now that all branded laptops (for example) are actually made by Compal&friends? The day sub-contracting OEMs start gaining traction, HP&all will be one business model short from being able to generate any profits.
It is probably only a matter of time. Shareholders demand profits, profits "demand" the cheapest Chinese/Taiwanese/etc. OEMs and cheaper parts.
It sucks that everything has been headed that way for most of the last 15+ years but established players in competitive markets do not want to risk being priced out of their respective markets. If the final OSX-x86 runs on generic PCs, Apple will have to compete with PC OEMs for system sales.
Add a copper mesh/film layer to your kevlar body armour and you should be fine. Tin foil should also work.
The reason they asked their test subjects to remove their glasses and contacts probably is that 95GHz has practically no penetration and would cause contacts to dry up in-place and might shatter glass glasses.
My guess is that being exposed to this thing would feel somewhat like being burned alive... without a flame. Let the 21st century witch hunting games beigin!
Cooling efficiency is expressed with the "Coefficient of Performance" (COP), and is the ratio of displaced energy and input energy.
If you need 1kW (or 1kJ/s) to displace 10kW (10kJ/s), the COP is 10. This means the displaced energy is ten times as much as the energy used to move it.
Now, the catch with Peltier elements is that they have high COP only at very low power and small temperature difference, usually around 5-10% of the power rating and 10C temperature difference. Unless they are operated under these optimal constraints, their COP quickly drops under five. So, to produce a highly efficient TEC AC able to handle 1000W, one would need a 1kW TEC bank operated at ~80W. Considering that a TEC costs about $0.25 per rated watt, this efficient solution would cost over $250, roughly twice the price of an average room AC... and it gets worse: 1kW is barely enough to cool one square meter worth of solar heating, car and house windows have a much larger surface area total than that.
Note: a TEC's 100% rating is where the TEC barely manages to pump its own heat away from the cold side. TECs used for thermal regulation usually operate in the 30-50% range. The high-efficiency range is usually somewhere around 5% with COPs sometimes reaching over 15. For comparison, the theoretical limit for freon (and many substitutes) is around 16 but the best practical implementations only reach around 12.
Now, a typical room AC pumps from 5kW to 12kW with a temperature delta around 20C with a COP around 10. So, to beat the phase-change system's efficiency, the TEC solution would have to be beefed up by about 20X (10X the load, 2X the delta), bringing the cost around $5k, which is 20X as expensive as good classic AC.
Until they find materials that offer both better electrical conductivity and better thermal insulation to improve their overall performance (widen their sweet spot and move it up the power curve), TECs will remain a somewhat marginal cooling solution.
One cable channel can serve up to 40Mbps but a single upstream channel is limited to 8Mbps. So for cable (DOCSIS) systems, this would typically be around 1/5, which is approximately what my ISP is sticking with, at around 1/6. (I know each serving group could include an arbitrary number of upstream and downstream channels but I suspect most cable ISPs, including my own, play cheap.)
BTW, there are places where 100Mbps and FTTH are already common, the catch is that these are shared networks. IIRC, in Sweden, a company basically puts everyone on a 100Mbps switched network with a single 100Mbps uplink. There is also Verizon's "Fios" where every house appears to become a node that injects and extracts traffic from an optical ring network. (Why else would the base installation require an SLA battery to specifically power the optical network tap?)
Because FireFox starts downloads before I pick the place where I want to put them, they are usually done by the time I click "save" so I do not have the "opportunity" to cancel most of the time.
I was wondering if there was an option to tell FF NOT to start "pre-downloading" stuff... I hate wasting 10+MB by the time I find out I already hag the file. (I discovered this when 300+MB downloads instantaneously finished upon clicking 'Save' because I went away for a while between clicking the link and picking the destination. With a 20GB/month downstream quota, this can be problematic/costly.)
BT for ~10MB files seems somewhat overkill and inconvenient. If Mozilla's mirrors all ran BT to seed every project, people would still end up getting 99% of the files from the servers. When I downloaded the FC4 DVD ISO, there were at least two IPs belonging to known RH mirrors that were feeding me 150-300KB/s each and pushed 2.6GB out of 2.9GB.
For small files with few fast servers, the download would end before ordinary peers start sending or requesting anything... ordinary peers will not request data from one another once their download bandwidth is maxed out thanks to 2-3 mirrors. A good and fairly simple load-balancing redirect script hidden behind the download link would be just as good as BT, without the inconvenience.
The litigation and software patent business models are other reasons that make programming, much of CS and IT much less interesting.
Who wants to get involved in an industry where legal exposure lurks around every corner? This hurts SMEs the most and SMEs employ the majority of people.
While spring contacts would be the simplest, they have a hard time with friction. With a lifetime in the 100k-1M cycles, the keyboard would have a useful service life of 2000-5000h before the most frequently used keys start to glitch/fail, not quite acceptable for a >$200 keyboard.
Flexible film (and wires) would work but require a number of extra steps - one end can be bonded to the PCB but the other needs some form of connector. An even better alternative would probably be to make the key's travel distance short enough to directly use spring-loaded pins.
Using a RFIDesque model to deliver 10-100mW to each key would be possible since the keys could be magnetically coupled with the keyboard to minimize dispersion... but the losses would be big enough to bring this quite a bit beyond USB's guaranteed power delivery capacity, even more so considering that even the wired implementation would already push the limit when all keys are white.
What I find even funnier is that assuming the "kids" in question are 17+ (or more like 15+), they all had some lectures on the topic in highschool and many already have first-hand experience in real life.
One could simply include only code-free header files (with adequate comments to let people know what these functions should do) for the modules that contain unwilling third-party code.
Function declarations should be safe, otherwise we risk seeing patents and other nonsense on the likes of "int (func*)(void*, int)"
Integration into OSS still leaves the code in plain sight for everyone to see. Integration into closed source makes the code and any modifications disappear from the public.
This is a little like GPL vs BSD: GPL forces code to remain free and public in every way while BSD allows everybody to freely "steal" and integrate the code into commercial closed-source software.
Farming costs less than $1/h per person, a small programmer team would cost well over $10/h. The bot's development would most likely take many weeks (more likely months without prior knowledge of the game's inner workings for tapping) and game updates might add extra checks to detect cheating devices.
In any case, I do not play multiplayer games so I should be safe from such exploits, real life does not (unfortunately?) have rollbacks.
7-9 is the number of routing (metallization) layers. The 32-40 figure is probably the number of masks required for the whole fabbing process, including substrate doping masks, insulation masks, metallization masks, etc.
In any case, doing "cubic" chips is not really going to be practical: volume increases faster than surface (heat transfer) area. If the power density increases faster than the transfer surface, the core will be even more likely to overheat unless the extra circuitry is low-power and can serve as a sort of heat-spreader like caches do in current CPUs.
Also, there is the matter of IO density, the core needs to be large enough to place all these IO and power pads.
Large, rectangular, thin chips provide plenty of heat spreading and IO bonding area thanks to large caches and all the transistors being on the same layer.
Also, going cubic (adding semiconducting layers) would add many extra masks, at least two or three per layer. With vertical transistor spanning three layers, using these would require between eight and 12 extra masks which in turn becomes more than 50 extra processing steps. This could substantially increase failure rates by multiplying the risk of one layer contaminating another, mask misallignment and other small process variations.
If adding layers was easy and cheap, AMD, Intel and the others would not go so far out of their way to fit their designs into the fewest layers possible. The same generally applies to PCBs.
When I said RFID-powered/programmed, I was referring to the keyboard-key interface.
Each key would have an integrated RFID-OLED driver and the RFID transmitter would be in the keyboard for minimal range. The keyboard itself would indeed be USB.
The reason for RFID (or equivalent) power and programming would be to avoid otherwise unavoidable wire breakage from repetitive key hits. The alternative would a tactile flat panel since wired keys probably would not live long.
The FAA also says "It is in the initial stages of production"
The real question here is: what is their definition of "initial stage". Is it initial as in "market/feasibility research", initial as in evaluating implementation technologies, or initial as in currently working on or debugging the first prototypes?
If they hope to release in 2006, they probably need prototypes for SDK and firmware validation this year.
I wonder what their keyboard-key interface would be... RFID? Serial? (If it was serial, I would certainly like to have a few keys to use as display devices for other projects.)
Alternatively, a flat tactile surface with configurable areas would be much simpler to manufacture and be much more polyvalent.
I think a single key for this keyboard would cost at least $1 USD to build. After all, each key would have a 32x32-64x64 OLED display, matching driver circuitry, probably RFID-powered and programmed. This would probably cost at least $2 USD/key.
If you consider the 133t gamer wow-factor appeal of this thing, I would not be surprised if such a keyboard came with an introductory price tag around $400 USD, with the price progressively dropping towards $300 as the initial surge disperses, then towards $200 as mass production kicks in.
At $200, I would start thinking about it. At $100, I would probably have no second thoughts about it but I do not think this will ever drop under $150 in this form.
A flat panel keyboard would be simpler to manufacture and more flexible but requires giving up the tactile response... one could configure a standard keyboard area, a trackpad area, volume slider area, etc. anywhere on the surface and even use it as a secondary/extended display, this would be great for laptops.
On very old hardware, the 'Turbo' button operated a clock signal multiplexer.
On slightly newer hardware, the 'Turbo' button was often wired to the clock multiplier or FSB frequency selection jumbers. Switch position changes for these were often taking effect only after a reset or power cycling.
Newer computers do not have any such functionality since everything is handled by the chipset, CPU, BIOS and software.
Does the fact that modern software requires Win2k or newer impedes the use of Win3.xx? Huh... never mind.
I did try to install Win3.11 on a P3 just for the heck of it and it crashed during boot. I did sort of like the way it booted in less than two seconds when I installed it for fun on one of my P200s after years of >30 seconds with Win9x though.
Does older stuff get in the way of newer stuff's sales? More or less. Stuff eventually fails or is obsolesced and needs to be fixed, replaced or upgraded. There will always be a replacement market and this market is growing with the installed base.
Instead of worrying about market growth, it would be about time to start worrying more seriously about the e-Waste this replacement economy will generate.
[...] if not for the right reasons. [...] With the expiration of the LZW patent it's not really a "GIF replacement" anymore
Looks like you did not notice the past tense in "its main aim was to provide a better, Compuserve-free GIF alternative"... back when PNG first came along, this _was_ one of the reasons, even if it no longer is _now_.
Doh! (I should avoid speed-reading so many things at once. BTW, it seems IBM also has a LZW patent due to expire next year.)
In any case, GIF is still the most common generic site image format on the sites I visit so I do not quite see this alledged favourism in action. Maybe there is a study somewhere with actual online image format statistics that would say otherwise.
And with GIF's liberation (or even throughout the Unisys/Compuserve episode), most site designers/maintainers simply applied the good old "If it ain't broke, don't fix it."
I think the fact that JPEG (1985) has been around for many more years than PNG (1995) and the fact that Unisys was collecting LZW (used by PNG) licenses from 1995 to 1999 also have something to do about that.
In any case, the majority of sites I visit still use GIFs (1987) for generic elements, like the rounded end on separators and story icons here.
AFAIK, PNG was never aimed at replacing JPEG... its main aim was to provide a better, Compuserve-free GIF alternative.
Slashdot Mirror
Or better yet, copy some data from /dev/random to a file.
The beauty of good encryption is that it generates seemingly random output. In theory, almost any random stream can generate any message so it would be possible for the authorities to mistakenly identify random data for an encrypted file.
And such a random block might be the encryption pad for something else. Here is an idea: use encrypted file as encryption pads for other files... this way, authorities would have no way to prove wether or not the pad has any meaning but you can at least prove that the file itself is needed to decrypt some other file.
It is open season for abuse. There are too many ways for people to hide meaning and too many for authorities to want meaning where there is none.
And what are they supposed to do with ephemeral/intermediate session keys that are not saved anywhere and usually never shown to the user?
And what about the case where the user simply and honestly forgot the encryption password? Does this mean that bad memory will become a crime?
If this could mean better, cheaper XP Pro/Server and the demise of Windows XP Crippled Editions (Starter and Home), I would be fairly happy.
Now, I wonder why large companies would have such a large proportion of Mac users. My guess is this is largely thanks to large corporations' overweight marketing departments. With companies ditching engineering and manufacturing, marketing is the only thing still going strong and the graphists they employ typically love Apple stuff.
Like what?
They started rebranding Luxeon LEDs and Tektronix/Fluke/etc. instruments?
It does suck that most Dell, HP and others' stuff is now made by someone else. What is the point of buying any American brand now that all branded laptops (for example) are actually made by Compal&friends? The day sub-contracting OEMs start gaining traction, HP&all will be one business model short from being able to generate any profits.
It is probably only a matter of time. Shareholders demand profits, profits "demand" the cheapest Chinese/Taiwanese/etc. OEMs and cheaper parts.
It sucks that everything has been headed that way for most of the last 15+ years but established players in competitive markets do not want to risk being priced out of their respective markets. If the final OSX-x86 runs on generic PCs, Apple will have to compete with PC OEMs for system sales.
Add a copper mesh/film layer to your kevlar body armour and you should be fine. Tin foil should also work.
The reason they asked their test subjects to remove their glasses and contacts probably is that 95GHz has practically no penetration and would cause contacts to dry up in-place and might shatter glass glasses.
My guess is that being exposed to this thing would feel somewhat like being burned alive... without a flame. Let the 21st century witch hunting games beigin!
Cooling efficiency is expressed with the "Coefficient of Performance" (COP), and is the ratio of displaced energy and input energy.
If you need 1kW (or 1kJ/s) to displace 10kW (10kJ/s), the COP is 10. This means the displaced energy is ten times as much as the energy used to move it.
Now, the catch with Peltier elements is that they have high COP only at very low power and small temperature difference, usually around 5-10% of the power rating and 10C temperature difference. Unless they are operated under these optimal constraints, their COP quickly drops under five. So, to produce a highly efficient TEC AC able to handle 1000W, one would need a 1kW TEC bank operated at ~80W. Considering that a TEC costs about $0.25 per rated watt, this efficient solution would cost over $250, roughly twice the price of an average room AC... and it gets worse: 1kW is barely enough to cool one square meter worth of solar heating, car and house windows have a much larger surface area total than that.
Note: a TEC's 100% rating is where the TEC barely manages to pump its own heat away from the cold side. TECs used for thermal regulation usually operate in the 30-50% range. The high-efficiency range is usually somewhere around 5% with COPs sometimes reaching over 15. For comparison, the theoretical limit for freon (and many substitutes) is around 16 but the best practical implementations only reach around 12.
Now, a typical room AC pumps from 5kW to 12kW with a temperature delta around 20C with a COP around 10. So, to beat the phase-change system's efficiency, the TEC solution would have to be beefed up by about 20X (10X the load, 2X the delta), bringing the cost around $5k, which is 20X as expensive as good classic AC.
Until they find materials that offer both better electrical conductivity and better thermal insulation to improve their overall performance (widen their sweet spot and move it up the power curve), TECs will remain a somewhat marginal cooling solution.
Not for the faster ADSL and cable deals.
One cable channel can serve up to 40Mbps but a single upstream channel is limited to 8Mbps. So for cable (DOCSIS) systems, this would typically be around 1/5, which is approximately what my ISP is sticking with, at around 1/6. (I know each serving group could include an arbitrary number of upstream and downstream channels but I suspect most cable ISPs, including my own, play cheap.)
BTW, there are places where 100Mbps and FTTH are already common, the catch is that these are shared networks. IIRC, in Sweden, a company basically puts everyone on a 100Mbps switched network with a single 100Mbps uplink. There is also Verizon's "Fios" where every house appears to become a node that injects and extracts traffic from an optical ring network. (Why else would the base installation require an SLA battery to specifically power the optical network tap?)
Because FireFox starts downloads before I pick the place where I want to put them, they are usually done by the time I click "save" so I do not have the "opportunity" to cancel most of the time.
I was wondering if there was an option to tell FF NOT to start "pre-downloading" stuff... I hate wasting 10+MB by the time I find out I already hag the file. (I discovered this when 300+MB downloads instantaneously finished upon clicking 'Save' because I went away for a while between clicking the link and picking the destination. With a 20GB/month downstream quota, this can be problematic/costly.)
BT for ~10MB files seems somewhat overkill and inconvenient. If Mozilla's mirrors all ran BT to seed every project, people would still end up getting 99% of the files from the servers. When I downloaded the FC4 DVD ISO, there were at least two IPs belonging to known RH mirrors that were feeding me 150-300KB/s each and pushed 2.6GB out of 2.9GB.
For small files with few fast servers, the download would end before ordinary peers start sending or requesting anything... ordinary peers will not request data from one another once their download bandwidth is maxed out thanks to 2-3 mirrors. A good and fairly simple load-balancing redirect script hidden behind the download link would be just as good as BT, without the inconvenience.
The litigation and software patent business models are other reasons that make programming, much of CS and IT much less interesting.
Who wants to get involved in an industry where legal exposure lurks around every corner? This hurts SMEs the most and SMEs employ the majority of people.
While spring contacts would be the simplest, they have a hard time with friction. With a lifetime in the 100k-1M cycles, the keyboard would have a useful service life of 2000-5000h before the most frequently used keys start to glitch/fail, not quite acceptable for a >$200 keyboard.
Flexible film (and wires) would work but require a number of extra steps - one end can be bonded to the PCB but the other needs some form of connector. An even better alternative would probably be to make the key's travel distance short enough to directly use spring-loaded pins.
Using a RFIDesque model to deliver 10-100mW to each key would be possible since the keys could be magnetically coupled with the keyboard to minimize dispersion... but the losses would be big enough to bring this quite a bit beyond USB's guaranteed power delivery capacity, even more so considering that even the wired implementation would already push the limit when all keys are white.
What I find even funnier is that assuming the "kids" in question are 17+ (or more like 15+), they all had some lectures on the topic in highschool and many already have first-hand experience in real life.
Alarmists sure like barking up the wrong tree.
One could simply include only code-free header files (with adequate comments to let people know what these functions should do) for the modules that contain unwilling third-party code.
Function declarations should be safe, otherwise we risk seeing patents and other nonsense on the likes of "int (func*)(void*, int)"
I wonder how extensive the breakage would be.
There is a distinction.
Integration into OSS still leaves the code in plain sight for everyone to see. Integration into closed source makes the code and any modifications disappear from the public.
This is a little like GPL vs BSD: GPL forces code to remain free and public in every way while BSD allows everybody to freely "steal" and integrate the code into commercial closed-source software.
I thought that was sarcasm.
Irony goes like: "Bad stuff like that never happens to me... oh crap."
Sarcasm is saying/wording things in a way that implies the opposite.
And while we're at it, oxymorons are the association of two opposite ideas, one popular (joke) example being "Microsoft Works."
Farming costs less than $1/h per person, a small programmer team would cost well over $10/h. The bot's development would most likely take many weeks (more likely months without prior knowledge of the game's inner workings for tapping) and game updates might add extra checks to detect cheating devices.
In any case, I do not play multiplayer games so I should be safe from such exploits, real life does not (unfortunately?) have rollbacks.
7-9 is the number of routing (metallization) layers. The 32-40 figure is probably the number of masks required for the whole fabbing process, including substrate doping masks, insulation masks, metallization masks, etc.
In any case, doing "cubic" chips is not really going to be practical: volume increases faster than surface (heat transfer) area. If the power density increases faster than the transfer surface, the core will be even more likely to overheat unless the extra circuitry is low-power and can serve as a sort of heat-spreader like caches do in current CPUs.
Also, there is the matter of IO density, the core needs to be large enough to place all these IO and power pads.
Large, rectangular, thin chips provide plenty of heat spreading and IO bonding area thanks to large caches and all the transistors being on the same layer.
Also, going cubic (adding semiconducting layers) would add many extra masks, at least two or three per layer. With vertical transistor spanning three layers, using these would require between eight and 12 extra masks which in turn becomes more than 50 extra processing steps. This could substantially increase failure rates by multiplying the risk of one layer contaminating another, mask misallignment and other small process variations.
If adding layers was easy and cheap, AMD, Intel and the others would not go so far out of their way to fit their designs into the fewest layers possible. The same generally applies to PCBs.
When I said RFID-powered/programmed, I was referring to the keyboard-key interface.
Each key would have an integrated RFID-OLED driver and the RFID transmitter would be in the keyboard for minimal range. The keyboard itself would indeed be USB.
The reason for RFID (or equivalent) power and programming would be to avoid otherwise unavoidable wire breakage from repetitive key hits. The alternative would a tactile flat panel since wired keys probably would not live long.
The FAA also says "It is in the initial stages of production"
The real question here is: what is their definition of "initial stage". Is it initial as in "market/feasibility research", initial as in evaluating implementation technologies, or initial as in currently working on or debugging the first prototypes?
If they hope to release in 2006, they probably need prototypes for SDK and firmware validation this year.
I wonder what their keyboard-key interface would be... RFID? Serial? (If it was serial, I would certainly like to have a few keys to use as display devices for other projects.)
Alternatively, a flat tactile surface with configurable areas would be much simpler to manufacture and be much more polyvalent.
I think a single key for this keyboard would cost at least $1 USD to build. After all, each key would have a 32x32-64x64 OLED display, matching driver circuitry, probably RFID-powered and programmed. This would probably cost at least $2 USD/key.
If you consider the 133t gamer wow-factor appeal of this thing, I would not be surprised if such a keyboard came with an introductory price tag around $400 USD, with the price progressively dropping towards $300 as the initial surge disperses, then towards $200 as mass production kicks in.
At $200, I would start thinking about it. At $100, I would probably have no second thoughts about it but I do not think this will ever drop under $150 in this form.
A flat panel keyboard would be simpler to manufacture and more flexible but requires giving up the tactile response... one could configure a standard keyboard area, a trackpad area, volume slider area, etc. anywhere on the surface and even use it as a secondary/extended display, this would be great for laptops.
On very old hardware, the 'Turbo' button operated a clock signal multiplexer.
On slightly newer hardware, the 'Turbo' button was often wired to the clock multiplier or FSB frequency selection jumbers. Switch position changes for these were often taking effect only after a reset or power cycling.
Newer computers do not have any such functionality since everything is handled by the chipset, CPU, BIOS and software.
Does the fact that modern software requires Win2k or newer impedes the use of Win3.xx? Huh... never mind.
I did try to install Win3.11 on a P3 just for the heck of it and it crashed during boot. I did sort of like the way it booted in less than two seconds when I installed it for fun on one of my P200s after years of >30 seconds with Win9x though.
Does older stuff get in the way of newer stuff's sales? More or less. Stuff eventually fails or is obsolesced and needs to be fixed, replaced or upgraded. There will always be a replacement market and this market is growing with the installed base.
Instead of worrying about market growth, it would be about time to start worrying more seriously about the e-Waste this replacement economy will generate.
[...] if not for the right reasons. [...] With the expiration of the LZW patent it's not really a "GIF replacement" anymore
Looks like you did not notice the past tense in "its main aim was to provide a better, Compuserve-free GIF alternative"... back when PNG first came along, this _was_ one of the reasons, even if it no longer is _now_.
Doh! (I should avoid speed-reading so many things at once. BTW, it seems IBM also has a LZW patent due to expire next year.)
In any case, GIF is still the most common generic site image format on the sites I visit so I do not quite see this alledged favourism in action. Maybe there is a study somewhere with actual online image format statistics that would say otherwise.
And with GIF's liberation (or even throughout the Unisys/Compuserve episode), most site designers/maintainers simply applied the good old "If it ain't broke, don't fix it."
I think the fact that JPEG (1985) has been around for many more years than PNG (1995) and the fact that Unisys was collecting LZW (used by PNG) licenses from 1995 to 1999 also have something to do about that.
In any case, the majority of sites I visit still use GIFs (1987) for generic elements, like the rounded end on separators and story icons here.
AFAIK, PNG was never aimed at replacing JPEG... its main aim was to provide a better, Compuserve-free GIF alternative.