Imagine how much of the cost of this vehicle will be in making it run 0-60 in 3 or 4 instead of 6 seconds or more. The additional amperage, cooling load, and other considerations from having a car that silently can slam at high speed into an object (wall, child, car, bike) without warning is immense. Given that the car is already @ $80k. It is virtually certain to cost more to get into production once they have all of the safety features including anything required to keep from electrocuting people when lots of metal gets bent and to recharge the cells individually so that none of the >6k cells get overcharged, you have a car that is designed for an specialized market segment that is very small. (High end Mercedes roadsters themselves are lucky to sell in the 10s of thousands, and they work in climates this car cannot. You don't want to get these batteries too hot or cold or use a lot of AC or heating)
Unless major care is used to charge or discharge only a portion of the theoretical range, battery replacement costs could overwhelm the purchase costs over a few years. In CA, the manufacturer must guarantee the batteries for at least 100k miles or 7 years (might be higher), so if they are replaced 3 or 4 times over that period, the manufacturer would have to add another ~$100k to the purchase price to cover the expected battery packs.
R.e. crashes, note that almost all nuclear reactor designs assume that the geometries of the fuel are intact. Good assumption most of the time.;) A car rearended by a semi or locomotive might need a VERY highly engineered containment system. With >6000 cells, you will need this to prevent fires caused by high-altitude, high temporatures, or manufacturing defects, or a short.
It would great for this to be super successful, but it sounds like a prototype and a great article in a magazine.
How providing multiple levels of performance in software shafting people?
CPU, car, and other vendors do this all of the time. It is a well known economic tradeoff that building a few number of product variations an enabling them in different market saves huge amounts of product development costs and improves product quality. Some examples include building only one version of a CPU family from a single die. This one die design can be very carefully tested without spreading the hundreds of millions of dollars of design validation work over a number of designs, avoiding a lot of end user defects. Then the design can be put into multiple markets, big caches, faster speeds etc, first based on speed and defect grading, then an the yield goes up, by disabling good parts. However, because the one design is being used for all of the parts, the errors that crop up an be incorporated very quickly into new stepping of the product, giving a much more robust product for even the specialty speeds that would otherwise remain buggy for far longer. Due to the volume, the die shinks can happen quicker and the once premium products can be propagated faster.
Same things happen in OSs, DBs etc, where often a lot of code is disabled but still present in simpler and lower costs configurations. Since the code base is often identical (such as in XP variations) for the most part, all of the bugs can be squeezed out faster. An 16 processor product may inherently be the same as a personal version and a tremendously higher price, but they all can share the same more rapid improvement of code and bug lists.
Yes, you theoretically have a crippled product due to all the features being built into the hardware or software, but both the company and the consumer have won through the resulting efficiency and quality.
This is not shafting, it's smart market differentiation and in many cases benefits everyone. (no guarantees, but what's new!)
This drive should be an absolute success. Who cares about 2 x transfer, by itself... the seek time is almost always more than time to transfer a large block of date. Plus, the time alone to rotate to the block's location is a huge amount also. At 5400 rpm, 1/2 of a rotation is 11ms, more than the time to transfer almost any block of data. The transfer speed is only good on hard drives when the head is positioned and the disk rotates into position--a 20-50% duty cycle. The flash drive can start transferring immediately and keep transferring--a huge win. Why can't Samsung tell us right now that this runs wickedly fast, or is their transfer rate really some effective speed accounting for the lack of seeks, otherwise we should be seeing 5x plus speeds. This is a trivial benchmark, but if they can make the specs without some odd adjustment this we should have a incredibly productive laptop in a lot of heavy disk based applications--it even would be able to go to sleep way faster. A large premium would be justified easier without any more hesitation for cost.
For most applications of the Ipod other than being a couch potato, the nice thing about using an Ipod is that you can use it jogging, riding a bike, in a gym (while intertwined with a midieval torture device repurposed as a very expensive gym machine), reading slashdot etc. In all of these cases, being able to push a button and having the device read you the current song info, next to play, the name of the album or song on a scan etc is way more useful than having to pull the player out of the jogging sleave or take you eyes off of your screen. Just one button and the very inexpensive text to speech technology available would let you have the player read the current song, coming song, or while skipping--set it to read the song / album prior to play so that you can keep navigating. this would be very useful--even if you had a screen, you would not want to take the player out of its holder.
There is no reason the shuffle should not have a feature like this. Why not?
With an exclusive deal, every other phone vendor other than Cingular MUST use a non-Itunes solution. To add music to a phone is trivial, the processing power to decompress music is low speed processing compared to radio and all you need is dual channel audio chip and more flash and your good to go.
Almost by definition, these player will be compatible with WMV (with many vendors of product) or Real formats in addition to MP3. A streaming download will be availabe from at least one music vendor (it will just be a bunch of songs stringed together that expires.) Also trivial and could be many "channel"... Apple disses streaming or rent all you want models. You can have any of these three.
And, no reason why videos and cameras can't be supported... they are already in almost every phone anyway.
They are going to be blown out of the market due to their very closed and limited purchased-only music and no video model.
too bad.
Internet improving faster than transport costs
on
NYT On Flying Cars
·
· Score: 1
When the internet is allowing more cardiologists, engineers, architects, accountant, and wall street analysts to work remotely, either from home or across the world, the target market of people who can spend huge amounts on commuting is vastly diminishing.
Not to say, that if you want to vacation where you can't get from here to there in first class, you "might" need but you can't afford this, because like virtually all vehicles, if they are not used frequently, capital costs exceed operating costs by a huge margin--you are better off chartering!
It would be far simpler to have 1, 3, 10, and 30�
on
Making Change
·
· Score: 1
The math is easy, it reduces the number of denominations, and results in virtually the same number of average coins at the 18 cent piece.
Also, with only 4 types of coins, it makes for extra room in cash drawers, few coins types to reconcile, and the math is within the bounds of the average person.
it would totally remove any incentive to make music or other content that has a high cost per sale ratio, i.e. small bands, any kind of music using a large orchestra for non-classical music, ancient music variations with instruments that may only last a few songs in their remaining lifetime, other audio/music contect that is of limited interest but high value, experimental or foreign music, and certainly almost any short story delivered through music.
much music will never achieve 10,000 copies sold through any means (that means you are not in the top 300 in the US)... why does there need to be price fix. You absolutely can't tour on this, much of this is not club material (the audience might not stay up this late) At best, if you have a nationwide audience of, say, 10,000 who would buy a CD or CD track, you might have a club audience in, say a club in a particular location in LA on a given (non-Laker, etc, etc.) night at best of 1 or 0 fans! Exciting. You need electronic distribution.
As a separate if not equally difficult issue, is that you need to have pricing that makes sense for the market, i.e. if you have Indian, NZ, and US fans, you likely can't charge any of them a day's pay to listen to your track as much as they might like to buy your track. The reverse of region control is pricing for the average market, i.e. much less than the US market. It will still exclude most of the low end of the world market, because food is more important. Do I keep my rights if i move (or travel) or sell my collection?
Also, in any licensing, I want to be able to release in multi unit sizes--a 30 second ditty, a 2 minute song, a 4 minute song, and maybe a 10 minute song or along with a bulk buy of my best songs for a bulk price. My 10 minute song may be the superior version, my 4 minute song a standard version, but maybe the 2 minute version is the really great version of the short ones that should get the real premium! Also, I might have a 2 minute version which is a computerized sped up version of the 3 minute version that has the same content -- which is licenses at the high price -- the one that preserves my time or the one that lags? And how did you set a "fixed price for all the works?"
US Toll-free routing is very complex == $$$
on
U.S. Endorses ENUM
·
· Score: 2, Interesting
Note that when you look up a US local phone number, it has only one destination (subject to local number portability) However, with US toll-free numbers (i.e. 800, 888, 877, 866, 855), the destination location is dependent on the time of day, the calling party id, and the relative amount of traffic sent to one location versus another. A toll-free number can have a separate destination for a given calling party 10-digit US phone number, area code or other criteria so that a call you place and one I place may end up on different carriers and different parts of the country. Also, a toll free number often absolutely prohibit connections altogether from some portions of the US (or Canada etc). Routing these calls properly is not simple.
For reasons such as this, I think we can expect a lot of hands out looking for money for putting small records in databases (registration) and for looking such up. Look at the business models of the heavy sponsors of ENUM.
Since the license for G723.1 is around ~$1, hardware seems overkill in an era where p3s and p4s abount. G723.1 is a stretch for slow processors, but luckily in an era where video compression is feasible, audio compression no longer requires hardware.
Is this hardware purchase a scam?
Anyone log the times for the best Ads and Plays
on
Superbowl XXXVII
·
· Score: 1
I'd sure like to have a time line for the best ads, best plays, and other memorable moments. I always tape the superbowl, but only want to watch the best parts!
Such a log would allow for chapter markers and section markets and would make a TIVO type capability so much more valuable. It isn't just which program, but which part of each program that is worth watching. Would make "ad skip" look primitive--besides, I'm watching for the ads!
I can't understand why everyone thinks IDEs have to be low reliabilty. One year warranties are a result of the fact that OEMs who buy 99% of these drives have warranties = 1 year, often far less and won't pay a penny for anything not essential to make a sale. A warranty is a very expensive item.
Maxtor builds drives with a mean time to failure of 1,000,000 hours --- 114 years - is that enough?
See http://www.maxtor.com/en/products/ata/enterprise_a pplications/index.htm
I wonder if the demo on the digitwireless site is rigged. T9 claims that 95% of the time the first guess is correct- my experience and pretty good, but the digitwireless shows a 14 key entry for their method, a 23 key entry for T9, and 27 for the conventional entry with multiple taps to get a second or third letter. Also, i cannot believe that the key entry could be a quite as fast per key when the keyboard has the extra complexity.
If T9's claim is good, digitwireless is not going to be a better solution for monoliqual western language users.
Agreed. You save the most money by not buying Office. Switching OS's saves far less, but at the cost of tremendous end-user issues. Most machines will come with a no extra cost windows license the will work acceptably without upgrades to newer versions for many many years.
Most office users end up only using Word. Buy Works which includes Word 2002 for $100. Don't bother with upgrades. Use Office or any replacement for those users who need other components.
You can save 60% of more of the cost at almost no hassle. Don't worry about upgrades, you'll get that when the useful life of the machines are over.
If paid placement is found not to work well in movies and TV's, then maybe the trend to finance movies this way won't take off. Otherwise, we'll eventually end up watching really long commercials with a hidden movie plot!
Virtually all of the cost of new drugs is in finding or creating the one out of 1000s that has significant beneficial effects vs damage and then going through exhaustive trials to make sure you didn't miss something.
Actually manufacturing the resulting drugs is sometimes expensive depending on the process, but it usually nearly free. In third world countries there are often identical drugs that are 1/10 or 1/100 the blockbuster price in the US. And generics are often drastically cheaper even in the US.
The raw materials are often virtually free, aspirin, codeine etc in bulk powder form went for at most a dollar or two per KG, when I last checked about ten years ago.
Not unlike the cost of your homemade copy of windows on a CD vs from the manufacturer or the cost of the truly high quality plug and play fully functional "Rolex" knockoff vs the one that the Rolex company makes, or YSL dress or Gucci bag... The IP of knowing how to make it is only value due to IP protection.
If you scale something like this out you have two big expenses other than marketing--that is paying for the streams and the bandwidth. The disk storage and serving itself is relatively small.
You have to have enough bandwidth for your peak times--usually your peak usage hour is has ten percentage of 24 hour day usage. It is really important to understand your peak ratios -- outside that peak, it is wasted bandwidth (like empty airline seats.)
Say your stream is 350k bps - the sample movie. If you can multicast (not likely... ) then you only need to have a stream for every simultaneous movie and format (a low bandwidth version also.)
If you don't can't multicast, your high bandwidth movie users are many times the cost of your low bandwidth users. Let's assume we don't have many low bandwidth users. So, at 350k that's only 3 movies per 1mbps of capacity or about 4 per T1. Using the peak to average ratio and assuming we have at least 1.5 hours per movie you only get two movies of capacity per 1 mbps of capacity. On a 24 hour basis that 20 movies and 600 movies per month per 1 mbps. In bulk, internet access can get to $200 per 1 mpbs. That's your biggest cost: $200 of serving per 600 movies. WHOOPS - you can't do DVD rates, even MPEG1 - your costs get into many dollars and your better off buying / renting from the local store--the clerk can hand you 9++GIGABYTES with your receipt and popcorn.
Multicast and / or lower bandwidth costs are needed to make this kind of thing work. The MPAA shouldn't worry too much, yet -- Moore's laws keeps working.
The real media cost is significant but pales versus bandwidth.
One dollar is a lot for 80% of the world, about right for a lot of the far east, and "too cheap" in the US. This would be the same even if it is DVD quality.
The nice thing about buying items from the rest of the world is that it is often at a much lower price point overseas. Importing IP into the US is far easier than buying other IP such as drugs in Mexico.
You get a movie for free and a five dollar (5 movie credit) just for signing up. You can watch - dont pontification and see it go down or get slashdotted -- regardless of whether you feel it should stay up.
Even thought it is real streamed at 300k bps, you'll get an idea of what the future could look like if we really could get our film libraries live.
Remember that many US concepts of copyright, fair use, etc don't translate into equivalent laws in other countries. This may be legal now and forever for agreements executed under the laws of Taiwan (this site). Note that some countries consider region coding to be unlawful (NZ?.
Note that the fair use concept in the US is stronger than in many others.
US owned a lot of IP and is considered to be unfair in its licensing practices in other countries -- they don't like embargoes on content, restrictive format licensing on contects, copy protection, delayed release dates in other countries and other US centric concepts.
If you have heavily hit database indexes, i.e. google, then you may need 100-1000x fewer machines. The cost of the disks is not the important cost, it is the far fewer number of machines for an equivalent query rate. However, you want to have far more than 2gb of directly addressed ram per machine--in fact at current prices it is probably cost effective to put 100's of gb per machine if you need to keep the query ram based--even if the CPUs are dwarfed in cost by the ram.
This is one of the reasons that we need 64 bit addressability on commodity IA architecture ASAP -- Ram drives using an IO subsystem adds a huge overhead compared to indexing in arrays and natural data organization as opposed to fixed blocks of byte that have to be retrieved as a unit with 100s++ of instructions and security models in the way of access!
Right now the typical voice cell user is paying $30 / month for a 200 minutes of prime time usage. At around 10,000 bits / second, thats 120 mb of data capacity in the form of voice bits. In another words, 4 megabytes of data on a good plan costs $1! This yields marginal profits on the oligopolies' multi-billion networks and spectrum. That's good pricing--many plans are far more expensive. Other ways to say this is that an MP3 is about a $1 of bandwidth and a 56k bps download (let say actually at 50kbps in actuality) uses $10 worth of bandwith per minute. Therefore, you don't need faster, you need a cheaper network. At least, you want your cell phone to use 802.11b when in range!
The importance of "64 bits" is in the ability for programs to seemlessly handle a flat (i.e. directly addresses) memory. This permits large arrays, databases, in memory files, etc to be handled. Even 10 more bits of addressing (i.e. 42 bits) would give you a huge benefit, as it would allow direct addressing of 4,000 gigabytes--more than the largest disk drives. The problem first recognized by Digital Equipment in the VAX design was that due to Moore's law you essentially had to add a bit to memory space each year - that doubles memory. Memory has gotten very inexpensive, even more so than density has increased - now 4 gigabytes of memory would be $1000 and yields incredible advantages, as disk access is perhaps 1000 times or more slower. Not being able to directly address memory is very difficult. If you want to manipulate 128 bit or larger object however, is takes a corresponding more instructions or more fetches, but it can be done.
These day large databases can be used by individuals that can be manipulated in memory, in the past, this would have cost millions of dollars of memory--you would have been using a mainframe or not even attempting to tackle the problem.
As for Risc vs Cisc, you may be able to operate somewhat faster for any given (RISC/CISC, IA/SPARC... ) instruction set, but it is the bit density and fetch pattern to the offchip world that mainly limits performance as the CPU ALUs are so much faster than the access times of the attached main memory. CPUs have gone from 8mhz to 2gmhz-250x, memory has sped up by 5-10x in the last 20 years.
Slashdot Mirror
only works when the saucepan is not in use. otherwise, you'll need a spare ...
Imagine how much of the cost of this vehicle will be in making it run 0-60 in 3 or 4 instead of 6 seconds or more. The additional amperage, cooling load, and other considerations from having a car that silently can slam at high speed into an object (wall, child, car, bike) without warning is immense. Given that the car is already @ $80k. It is virtually certain to cost more to get into production once they have all of the safety features including anything required to keep from electrocuting people when lots of metal gets bent and to recharge the cells individually so that none of the >6k cells get overcharged, you have a car that is designed for an specialized market segment that is very small. (High end Mercedes roadsters themselves are lucky to sell in the 10s of thousands, and they work in climates this car cannot. You don't want to get these batteries too hot or cold or use a lot of AC or heating)
;) A car rearended by a semi or locomotive might need a VERY highly engineered containment system. With >6000 cells, you will need this to prevent fires caused by high-altitude, high temporatures, or manufacturing defects, or a short.
Unless major care is used to charge or discharge only a portion of the theoretical range, battery replacement costs could overwhelm the purchase costs over a few years. In CA, the manufacturer must guarantee the batteries for at least 100k miles or 7 years (might be higher), so if they are replaced 3 or 4 times over that period, the manufacturer would have to add another ~$100k to the purchase price to cover the expected battery packs.
R.e. crashes, note that almost all nuclear reactor designs assume that the geometries of the fuel are intact. Good assumption most of the time.
It would great for this to be super successful, but it sounds like a prototype and a great article in a magazine.
Are you going to put a huge fireproof red sticker on the car warning firefighters not to douse the flames with water so as not to create an explosion?
How providing multiple levels of performance in software shafting people?
CPU, car, and other vendors do this all of the time. It is a well known
economic tradeoff that building a few number of product variations an
enabling them in different market saves huge amounts of product development
costs and improves product quality. Some examples include building only
one version of a CPU family from a single die. This one die design can be very
carefully tested without spreading the hundreds of millions of dollars of
design validation work over a number of designs, avoiding a lot of end user
defects. Then the design can be put into multiple markets, big caches,
faster speeds etc, first based on speed and defect grading, then an the
yield goes up, by disabling good parts. However, because the one design is
being used for all of the parts, the errors that crop up an be incorporated
very quickly into new stepping of the product, giving a much more robust
product for even the specialty speeds that would otherwise remain buggy
for far longer. Due to the volume, the die shinks can happen quicker and the
once premium products can be propagated faster.
Same things happen in OSs, DBs etc, where often a lot of code is disabled but
still present in simpler and lower costs configurations. Since the code base
is often identical (such as in XP variations) for the most part, all of the
bugs can be squeezed out faster. An 16 processor product may inherently be
the same as a personal version and a tremendously higher price, but they all
can share the same more rapid improvement of code and bug lists.
Yes, you theoretically have a crippled product due to all the features being
built into the hardware or software, but both the company and the consumer
have won through the resulting efficiency and quality.
This is not shafting, it's smart market differentiation and in many cases
benefits everyone. (no guarantees, but what's new!)
This drive should be an absolute success. Who cares about 2 x transfer, by itself ... the seek time is almost always more than time to transfer a large block of date. Plus, the time alone to rotate to the block's location is a huge amount also. At 5400 rpm, 1/2 of a rotation is 11ms, more than the time to transfer almost any block of data. The transfer speed is only good on hard drives when the head is positioned and the disk rotates into position--a 20-50% duty cycle. The flash drive can start transferring immediately and keep transferring--a huge win. Why can't Samsung tell us right now that this runs wickedly fast, or is their transfer rate really some effective speed accounting for the lack of seeks, otherwise we should be seeing 5x plus speeds. This is a trivial benchmark, but if they can make the specs without some odd adjustment this we should have a incredibly productive laptop in a lot of heavy disk based applications--it even would be able to go to sleep way faster. A large premium would be justified easier without any more hesitation for cost.
For most applications of the Ipod other than being a couch potato, the nice thing about using an Ipod is that you can use it jogging, riding a bike, in a gym (while intertwined with a midieval torture device repurposed as a very expensive gym machine), reading slashdot etc. In all of these cases, being able to push a button and having the device read you the current song info, next to play, the name of the album or song on a scan etc is way more useful than having to pull the player out of the jogging sleave or take you eyes off of your screen. Just one button and the very inexpensive text to speech technology available would let you have the player read the current song, coming song, or while skipping--set it to read the song / album prior to play so that you can keep navigating. this would be very useful--even if you had a screen, you would not want to take the player out of its holder.
There is no reason the shuffle should not have a feature like this. Why not?
With an exclusive deal, every other phone vendor other than Cingular MUST use a non-Itunes solution. To add music to a phone is trivial, the processing power to decompress music is low speed processing compared to radio and all you need is dual channel audio chip and more flash and your good to go.
... Apple disses streaming or rent all you want models. You can have any of these three.
... they are already in almost every phone anyway.
Almost by definition, these player will be compatible with WMV (with many vendors of product) or Real formats in addition to MP3. A streaming download will be availabe from at least one music vendor (it will just be a bunch of songs stringed together that expires.) Also trivial and could be many "channel"
And, no reason why videos and cameras can't be supported
They are going to be blown out of the market due to their very closed and limited purchased-only music and no video model.
too bad.
When the internet is allowing more cardiologists, engineers, architects, accountant, and wall street analysts to work remotely, either from home or across the world, the target market of people who can spend huge amounts on commuting is vastly diminishing.
Not to say, that if you want to vacation where you can't get from here to there in first class, you "might" need but you can't afford this, because like virtually all vehicles, if they are not used frequently, capital costs exceed operating costs by a huge margin--you are better off chartering!
The math is easy, it reduces the number of denominations, and results in virtually the same number of average coins at the 18 cent piece.
Also, with only 4 types of coins, it makes for extra room in cash drawers, few coins types to reconcile, and the math is within the bounds of the average person.
The same advantage accrues in currency also.
this is a can of worms ...
... why does there need to be price fix. You absolutely can't tour on this, much of this is not club material (the audience might not stay up this late) At best, if you have a nationwide audience of, say, 10,000 who would buy a CD or CD track, you might have a club audience in, say a club in a particular location in LA on a given (non-Laker, etc, etc.) night at best of 1 or 0 fans! Exciting. You need electronic distribution.
it would totally remove any incentive to make music or other content that has a high cost per sale ratio, i.e. small bands, any kind of music using a large orchestra for non-classical music, ancient music variations with instruments that may only last a few songs in their remaining lifetime, other audio/music contect that is of limited interest but high value, experimental or foreign music, and certainly almost any short story delivered through music.
much music will never achieve 10,000 copies sold through any means (that means you are not in the top 300 in the US)
As a separate if not equally difficult issue, is that you need to have pricing that makes sense for the market, i.e. if you have Indian, NZ, and US fans, you likely can't charge any of them a day's pay to listen to your track as much as they might like to buy your track. The reverse of region control is pricing for the average market, i.e. much less than the US market. It will still exclude most of the low end of the world market, because food is more important. Do I keep my rights if i move (or travel) or sell my collection?
Also, in any licensing, I want to be able to release in multi unit sizes--a 30 second ditty, a 2 minute song, a 4 minute song, and maybe a 10 minute song or along with a bulk buy of my best songs for a bulk price. My 10 minute song may be the superior version, my 4 minute song a standard version, but maybe the 2 minute version is the really great version of the short ones that should get the real premium! Also, I might have a 2 minute version which is a computerized sped up version of the 3 minute version that has the same content -- which is licenses at the high price -- the one that preserves my time or the one that lags? And how did you set a "fixed price for all the works?"
Note that when you look up a US local phone number, it has only one destination (subject to local number portability) However, with US toll-free numbers (i.e. 800, 888, 877, 866, 855), the destination location is dependent on the time of day, the calling party id, and the relative amount of traffic sent to one location versus another. A toll-free number can have a separate destination for a given calling party 10-digit US phone number, area code or other criteria so that a call you place and one I place may end up on different carriers and different parts of the country. Also, a toll free number often absolutely prohibit connections altogether from some portions of the US (or Canada etc). Routing these calls properly is not simple.
For reasons such as this, I think we can expect a lot of hands out looking for money for putting small records in databases (registration) and for looking such up. Look at the business models of the heavy sponsors of ENUM.
Since the license for G723.1 is around ~$1, hardware seems overkill in an era where p3s and p4s abount. G723.1 is a stretch for slow processors, but luckily in an era where video compression is feasible, audio compression no longer requires hardware.
Is this hardware purchase a scam?
I'd sure like to have a time line for the best ads, best plays, and other memorable moments. I always tape the superbowl, but only want to watch the best parts!
Such a log would allow for chapter markers and section markets and would make a TIVO type capability so much more valuable. It isn't just which program, but which part of each program that is worth watching. Would make "ad skip" look primitive--besides, I'm watching for the ads!
I can't understand why everyone thinks IDEs have to be low reliabilty. One year warranties are a result of the fact that OEMs who buy 99% of these drives have warranties = 1 year, often far less and won't pay a penny for anything not essential to make a sale. A warranty is a very expensive item.
a pplications/index.htm
Maxtor builds drives with a mean time to failure of 1,000,000 hours --- 114 years - is that enough?
See http://www.maxtor.com/en/products/ata/enterprise_
I wonder if the demo on the digitwireless site is rigged. T9 claims that 95% of the time the first guess is correct- my experience and pretty good, but the digitwireless shows a 14 key entry for their method, a 23 key entry for T9, and 27 for the conventional entry with multiple taps to get a second or third letter. Also, i cannot believe that the key entry could be a quite as fast per key when the keyboard has the extra complexity.
If T9's claim is good, digitwireless is not going to be a better solution for monoliqual western language users.
Agreed. You save the most money by not buying Office. Switching OS's saves far less, but at the cost of tremendous end-user issues. Most machines will come with a no extra cost windows license the will work acceptably without upgrades to newer versions for many many years.
Most office users end up only using Word. Buy Works which includes Word 2002 for $100. Don't bother with upgrades. Use Office or any replacement for those users who need other components.
You can save 60% of more of the cost at almost no hassle. Don't worry about upgrades, you'll get that when the useful life of the machines are over.
If paid placement is found not to work well in movies and TV's, then maybe the trend to finance movies this way won't take off. Otherwise, we'll eventually end up watching really long commercials with a hidden movie plot!
Virtually all of the cost of new drugs is in finding or creating the one out of 1000s that has significant beneficial effects vs damage and then going through exhaustive trials to make sure you didn't miss something.
... The IP of knowing how to make it is only value due to IP protection.
Actually manufacturing the resulting drugs is sometimes expensive depending on the process, but it usually nearly free. In third world countries there are often identical drugs that are 1/10 or 1/100 the blockbuster price in the US. And generics are often drastically cheaper even in the US.
The raw materials are often virtually free, aspirin, codeine etc in bulk powder form went for at most a dollar or two per KG, when I last checked about ten years ago.
Not unlike the cost of your homemade copy of windows on a CD vs from the manufacturer or the cost of the truly high quality plug and play fully functional "Rolex" knockoff vs the one that the Rolex company makes, or YSL dress or Gucci bag
worked for me - 2 minutes after being posted in slashdot. Might not scale.
If you scale something like this out you have two big expenses other than marketing--that is paying for the streams and the bandwidth. The disk storage and serving itself is relatively small.
... ) then you only need to have a stream for every simultaneous movie and format (a low bandwidth version also.)
You have to have enough bandwidth for your peak times--usually your peak usage hour is has ten percentage of 24 hour day usage. It is really important to understand your peak ratios -- outside that peak, it is wasted bandwidth (like empty airline seats.)
Say your stream is 350k bps - the sample movie. If you can multicast (not likely
If you don't can't multicast, your high bandwidth movie users are many times the cost of your low bandwidth users. Let's assume we don't have many low bandwidth users. So, at 350k that's only 3 movies per 1mbps of capacity or about 4 per T1. Using the peak to average ratio and assuming we have at least 1.5 hours per movie you only get two movies of capacity per 1 mbps of capacity. On a 24 hour basis that 20 movies and 600 movies per month per 1 mbps. In bulk, internet access can get to $200 per 1 mpbs. That's your biggest cost: $200 of serving per 600 movies. WHOOPS - you can't do DVD rates, even MPEG1 - your costs get into many dollars and your better off buying / renting from the local store--the clerk can hand you 9++GIGABYTES with your receipt and popcorn.
Multicast and / or lower bandwidth costs are needed to make this kind of thing work. The MPAA shouldn't worry too much, yet -- Moore's laws keeps working.
The real media cost is significant but pales versus bandwidth.
One dollar is a lot for 80% of the world, about right for a lot of the far east, and "too cheap" in the US. This would be the same even if it is DVD quality.
The nice thing about buying items from the rest of the world is that it is often at a much lower price point overseas. Importing IP into the US is far easier than buying other IP such as drugs in Mexico.
You get a movie for free and a five dollar (5 movie credit) just for signing up. You can watch - dont pontification and see it go down or get slashdotted -- regardless of whether you feel it should stay up.
Even thought it is real streamed at 300k bps, you'll get an idea of what the future could look like if we really could get our film libraries live.
Remember that many US concepts of copyright, fair use, etc don't translate into equivalent laws in other countries. This may be legal now and forever for agreements executed under the laws of Taiwan (this site). Note that some countries consider region coding to be unlawful (NZ?.
Note that the fair use concept in the US is stronger than in many others.
US owned a lot of IP and is considered to be unfair in its licensing practices in other countries -- they don't like embargoes on content, restrictive format licensing on contects, copy protection, delayed release dates in other countries and other US centric concepts.
If you have heavily hit database indexes, i.e. google, then you may need 100-1000x fewer machines. The cost of the disks is not the important cost, it is the far fewer number of machines for an equivalent query rate. However, you want to have far more than 2gb of directly addressed ram per machine--in fact at current prices it is probably cost effective to put 100's of gb per machine if you need to keep the query ram based--even if the CPUs are dwarfed in cost by the ram.
This is one of the reasons that we need 64 bit addressability on commodity IA architecture ASAP -- Ram drives using an IO subsystem adds a huge overhead compared to indexing in arrays and natural data organization as opposed to fixed blocks of byte that have to be retrieved as a unit with 100s++ of instructions and security models in the way of access!
Right now the typical voice cell user is paying $30 / month for a 200 minutes of prime time usage. At around 10,000 bits / second, thats 120 mb of data capacity in the form of voice bits. In another words, 4 megabytes of data on a good plan costs $1! This yields marginal profits on the oligopolies' multi-billion networks and spectrum. That's good pricing--many plans are far more expensive. Other ways to say this is that an MP3 is about a $1 of bandwidth and a 56k bps download (let say actually at 50kbps in actuality) uses $10 worth of bandwith per minute. Therefore, you don't need faster, you need a cheaper network. At least, you want your cell phone to use 802.11b when in range!
The importance of "64 bits" is in the ability for programs to seemlessly handle a flat (i.e. directly addresses) memory. This permits large arrays, databases, in memory files, etc to be handled. Even 10 more bits of addressing (i.e. 42 bits) would give you a huge benefit, as it would allow direct addressing of 4,000 gigabytes--more than the largest disk drives. The problem first recognized by Digital Equipment in the VAX design was that due to Moore's law you essentially had to add a bit to memory space each year - that doubles memory. Memory has gotten very inexpensive, even more so than density has increased - now 4 gigabytes of memory would be $1000 and yields incredible advantages, as disk access is perhaps 1000 times or more slower. Not being able to directly address memory is very difficult. If you want to manipulate 128 bit or larger object however, is takes a corresponding more instructions or more fetches, but it can be done.
... ) instruction set, but it is the bit density and fetch pattern to the offchip world that mainly limits performance as the CPU ALUs are so much faster than the access times of the attached main memory. CPUs have gone from 8mhz to 2gmhz-250x, memory has sped up by 5-10x in the last 20 years.
These day large databases can be used by individuals that can be manipulated in memory, in the past, this would have cost millions of dollars of memory--you would have been using a mainframe or not even attempting to tackle the problem.
As for Risc vs Cisc, you may be able to operate somewhat faster for any given (RISC/CISC, IA/SPARC