But seriously, now that Hotel Pennsylvannia will be turned into millionaire's housing, maybe this thing can be moved to a cheaper and more accessible part of the country. Especially in light of the economy, we should focus on a place that is central and is a big Greyhound hub. Having everyone all camp out might be a bit of a stretch, So it's not cheap, but more accessible and more of a Greyhound hub than NYC? Does such a place actually exist? NY has 3+ airports, good connections by bus and train and is central for the East Coast (which has half of the US population). Even the center of US population is right around the Mississippi.
Perhaps some people could get together to organize a complementary conference, to be held in January (cheaper travel and most time separation from the real Hope). I suggest we put it in some place like Albequerque or Los Cruces, which might be bearable at that time of year, and is generally cheap. There are alot of hacker conferences in the US other than Defcon (Las Vegas) and HOPE (although those are the two biggest). Off the top of my head there is also Summercon (Atlanta), Notacon (Cleveland), Schmoocon (DC), Carolinacon (NC), and ToorCon (San Diego). Further abroad there is the Chaos Computer Camp in Germany. Obviously you can go off and start your own, but I'd take a careful look around first.
Right before The Last HOPE was announced, Vornado Realty Trust announced that they were tearing down the Hotel Pennsylvania before 2010 (when the next HOPE presumably would be). While there are certainly other venues in NYC, finding one the size of the space HOPE has in the Hotel Penn at a price that HOPE could be under $100 has appeared pretty hopeless.
I'm a reasonably heavy DSLR user who shoots on a Nikon D2H. I have shot for fashion and dance shows where I leave with over 1500 RAW photos (I attach my camera directly to a Powerbook which has a 250GB firewire drive attached). I've tried using iPhoto for managing my photos, as most of the professional workflow programs with databases are thousands of dollars to say hello. iPhoto essentially falls over and dies with those kind of numbers. iPhoto also doesn't actually handle RAW images, it converts them over to JPEG using a rather mediocre converter.
I used to use Photoshop CS for "developing" my raw images, but most of its capabilities are focused around working with the photo once you've imported it as a PSD, and not around manipulating the photo itself. Along with many other photographers I've discovered CaptureOne is incredibly useful for non destructive processing of RAW images, as well as doing a wonderful job on noise reduction, color noise, banding, white balance, exposure, and levels.
I was hoping Aperture could replace CaptureOne and iPhoto for me, while allowing me to contine to use Photoshop when I wanted to edit a photo rather than just process a RAW image. As far as I can tell, this is dead on what Apple intended Aperture for.
To start off, I imported 3 iPhoto libraries with a total of 45,000 images into Aperture. To my surprise, it also imported all album and roll data with it (I was expecting to end up with a flat photo space) as well as importing all NEFs and the jpegs iPhoto had created automatically as different versions of the same photo. It's clear that the upgrade path from iPhoto to Aperture was well thought out.
Aperture seems to be very good at handling a large image database. I now have 45,000 photos in a single Aperture library, and am not using more than 450MB of ram opening a window with all images in it (scrolling of course).
Aperture also claimed to be able to handle many of the non destructive RAW workflow duties I'd handled before with Capture One. That's a bit more of a mixed bag. The white balancing loupe doesn't work nearly as well as Capture One's and occasionally creates psychadelic white balances in the process. The sharpening and noise reduction algorithms are nowhere near as good as Capture One's, and color noise reduction seems to be almost non existant on high exposure shots. Before someone points out that this is what Photoshop or some other tool is for, Aperture only exports PSDs or TIFFs to other applications so it has to handle all RAW processing itself.
If Apple can figure out how to handle RAW images better, Aperature could really become an incredible product. As it is, the workflow management, versioning, and just plain dealing with tons of images seem to be really nice.
Given a standard bios these machines should have an incompatable partition structure to current Macs (as has already been noted). Have to wonder if the Intel Macs will be able to still handle slices and read old disks. There were also some rumors that the endianness may be different on HFS for "Mactel" for internal binary structures (and therefore incompatable filesystems). I'm assuming not as it would create a problem for iPods as well as other firewire hard drives, but boot support for different partitioning schemes worries me a bit more.
I don't care if I can't run Forth in my bios any more (in fact, Intel's new BIOS scheme they did for the Itanium is nice) but I really hope that ease of integration with intel chipsets doesn't get in the way of legacy support.
They also have more amusing risk factors than that one. (All text from SCO's current 10k)
We do not have a history of profitable operations.
We may not prevail in our SCO Litigation, which may adversely affect our business.
Our failure to timely file this Form 10-K, and our quarterly report on Form 10-Q for the quarter ended January 31, 2005 (as previously defined, the "Form 10-Q"), could result in the delisting of our common stock on The Nasdaq SmallCap Market.
Our Engagement Agreement with the Law Firms will require us to spend a significant amount of cash during fiscal year 2005 and could harm our liquidity position.
Our future SCOsource licensing revenue is uncertain.
We may lose the support of industry partners leading to an accelerated decline in our UNIX products and services revenue.
Our claims relating to our UNIX intellectual property may subject us to additional legal proceedings.
Fluctuations in our operating results or the failure of our operating results to meet the expectations of public market analysts and investors may negatively impact our stock price.
We operate in a highly competitive market and face significant competition from a variety of current and potential sources; many of our current and potential competitors have greater financial and technical resources than we do; thus, we may fail to compete effectively.
If the market for UNIX continues to contract, our business will be harmed.
We rely on our indirect sales channel for distribution of our products, and any disruption of our channel at any level could adversely affect the sales of our products.
Our Engagement Agreement with the Law Firms representing us to enforce our intellectual property rights may reduce our ability to raise additional financing.
Our foreign-based operations and sales create special problems, including the imposition of governmental controls and taxes and fluctuations in currency exchange rates that could hurt our results.
If we are unable to retain key personnel in an intensely competitive environment, our operations could be adversely affected.
We have issued shares and options under our Equity Compensation Plans that were not exempt from registration or qualification under federal and state securities laws, and, as a result, we may incur liability to repurchase such shares and options and may face additional potential claims under federal and state securities laws.
Our stock price is volatile.
There are risks associated with the potential exercise of our outstanding options.
The resale of common shares by BayStar may have an adverse impact on the market value of our stock and the existing holders of our common stock.
Our stock price could decline further because of the activities of short sellers.
The right of our board of directors to authorize additional shares of preferred stock could adversely impact the rights of holders of our common stock.
Our Stockholder Rights Plan could make it more difficult for a hostile bid for the Company or a change of control transaction to succeed at current market prices for our stock.
FYI, the full text of the risk that Groklaw is quoting from:
Our claims relating to our UNIX intellectual property may subject us to additional legal proceedings.
In August 2003, Red Hat brought a lawsuit against us asserting that the Linux operating system does not infringe on our UNIX intellectual property rights and seeking a declaratory judgment for non-infringement of copyrights and no misappropriation of trade secrets. In addition, Red Hat claims we have engaged in false advertising in violation of the Lanham Act, decept
If there's any group out there who is claiming that they are doing something that the SEC is going to have to investigate, they need to disclose the risk to make a shareholder lawsuit less likely.
If there is some risk that a company doesn't disclose, where there was some way for them to know that it was a posibility, they can be sued by shareholders if the stock goes down as a result of it comes true.
You often find really silly risks listed in safe harbor statement like that. For example, Walmart lists as a risk that they may not be able to buy from certain vendors if political instability takes place in their country. They also disclose that they won't do as well if they can't hire good employees.
Also, if they know about a SEC investigation or lawsuit against them, a company would usually give more information than that in a 10k (lest the SEC investigate them for the way they disclose their risks in their 10k).
You're confusing emulation with virtualization. Storage virtualization is a really big deal these days. Check out the book "Virtual Storage Redefined: Technologies and applications for storage virtualization" by Paul Massiglia (of Veritas).
Isolation, and performance guarintees on shared systems are often more important than raw performance in something like a datacenter environment.
I've seen several people now post sessions they've had with "Snubby". Snubby is assuming that people are ordering things in a specific order. A session I just had with it: telnet 64.94.110.11 25 Trying 64.94.110.11... Connected to 64.94.110.11. Escape character is '^]'. 220 snubby3-wceast Snubby Mail Rejector Daemon v1.3 ready
250 OK
250 OK
550 User domain does not exist.
250 OK
221 snubby3-wceast Snubby Mail Rejector Daemon v1.3 closing transmission channel Connection closed by foreign host. That's right. It doesn't parse the input at all (I just hit Enter a bunch of times). If you have multiple RCPT lines, or have an extra command in there anywhere, you will get an OK in the wrong place and it will look like you have succeeded.
I'm not claiming Linux doesn't have any problems, but take a closer look how each of the vulns you just posted effects overall system security. I think you'd find that the only one which could be exploited in most systems is the kdelibs problem, if you were using Konquerer.
I have to go back to August 1st to find a remote buffer overflow which effects a substantial portion of the Debian crowd. That would be the vulnerability in the X truetype font server. It runs as nobody, so not much can be done there either. Take a look at the Debian security advisories: for 2003. I think you'll find that the principles of least needed security level, and installing only what is asked for have served them well.
Additionally, look at how many of the packages which have had problems are a part of the core operating system. If Microsoft also put out advisories for many of the common programs run on its operating systems, the list might run a bit longer.
You also seem annoyed that all of these security updates need to be installed, but ignore the fact that you would have had to install many more of this number of upgrades to be running all current versions. Debian Stable is rock solid, but too boring for most of us.
I keep my Linux, OSX, Solaris and Windows systems equally up to date on security patches so I have an equal lack of problems on all of them, but you're comparing a kettle to a.... 50 gallon lobster pot I guess.
Adam
Still not a little hard drive replacement
on
4Gb CF Card Announced
·
· Score: 3, Informative
I've seen two misstatements repeated over and over again in this discussion. Folks have been suggesting that flash cards like this might represent the future of hard drives. They've also been mentioning that it has a higher speed than the IBM microdrive.
While this is true in a camera, where you tend to erase an entire card and then fill it in a linear fasion, this isn't true when you use it as a hard drive. Flash memory has two things which make it unique, slow erasures, and limited numbers of cycles. Unlike a hard drive, where you can simply overwrite data, in flash memory you have to erase a region of it first. Usually you also have to erase a much larger region than a filesystem block (64k vs 4k). These erasures can be as painful as.5 seconds as well. Typical UNIX filesystems like ext2 or ffs, keep their data structures in fixed locations. Most writes are to metadata, and they will cause the metadata parts of a CF card to be erased and overwritten over and over again. Unlike a hard drive which can survive almost unlimited cycles like this, you will only get a few thousand in flash memory. Copying a set of files might burn out some cells in a single operation.
The log-structured filesystem (lfs) presents a partial solution to this, by writing data in blocks, deleting it in blocks, and writing to the end of a disk before starting over again. Unfortunatly, lfs becomes unefficient once fragmentation starts to set in, as a "cleaner" is necessary to group data back into blocks.
I still think one of these would be cool in my camera, but I want a 4G microdrive for my computer.
They even seem to have said that new high-end calculators are coming out on HP's site. I'm curious where the development effort for this is coming from. I don't think the calculator folks from Australia or France survived the great Carly purge.
That's one reason why you haven't seen them in servers but there are others.
Testing - The test patterns a server drive and a desktop IDE drive go through are very different. IDE drives aren't run for more than a single pass or two, while SCSI drives can be subjected to a day or more of continuous testing
Firmware - SCSI firmware is made to do a better job at trying to save your data. In most cases it's 4-5x loc of IDE firmware. Much of this is different techniques of reading a block which contains an error.
Queueing - Now you can argue that SATA also brings this to the ATA platform, but it's only sort of true. SCSI drives get advantages from queueing for two reasons, greater queue depth leads to shorter overall write time, and the ability to reorder queues. If a drive can write the closest block to the head, first, it is going to perform better. Now it would be possible to do this on a SATA drive, but not at IDE costs. One of the biggest differences in chipsets is integration, usually SCSI drives will offload things like servo control to a separate processor. Unfortunatly on IDE, its one processor is at 80-90% load trying just to do servo control. It doesn't have time to reorder your queue. (This is of course fixable, but people would have to want to pay for the extra processor power)
Rotational error - SCSI drives are designed to handle the types of errors which come from putting several drives in the same case. That is each drive tries to transfer some of its rotational momentum into the case. The intersection of these forces is a case which torques in the direction of drive spin. In IDE drives this can cause drastic reductions in throughput as more and more retries are necessary. (There are some IDE RAID cases good enough to fix this completely, but most only do partially). I've actually seen conflicting research on this last pont, but only in how good the case has to be to prevent these effects.
I do have some experience hotswapping drives. Linux sort of handles it. echo `scsi remove-single-device 0 0 1 0` >/proc/scsi/scsi and then echo `scsi add-single-device 0 0 1 0` >/proc/scsi/scsi will in theory hot swap target 1. However, I've had about a 50-75% success rate with not crashing the machine doing it. Hotswappable IDE is even possible, but your controller has to support it.
We have (or maybe I should say had) a Roomba in our house. I believe it was a beta model. It did a much better job than any of us expected making it around college student rooms, around in a bathroom, and even our porch. The only real complaint we had with its operations was the small size of it's container for storing whatever it vacumed.
It had quite a few nifty features. The led on it slowly changed from green, to yellow, to red as the battery drained. It'd be nice to see that on a notebook computer! Being a house full of computer science majors, quite a bit of time was spent figuring out what its algorithm was for room coverage. While we didn't get it all quite worked out, it seemed to hit all of the room.
Unfortunatly, it met a fairly quick end. After about two days, we found it running in a circle. Opening it up, we discovered that one of the wheel motors had actually siezed. We still haven't been able to find the appropriate motor on mouser or digikey. It doesn't matter too much as the final version should be on its way to us soon enough.
This is not the case. SCSI drives have much more testing done on them, and much more predictible performance under certain types of workloads. If you test a drive with a very low queue depth, as the tests in the original article were, you aren't likely to see a difference. However, if you start allowing for queue reordering, and more seeking, the performance curves will change. Additionally, the number of attempts that a SCSI drive will make to recover after an error is much higher than an IDE. The amount of code in the firmware is actually an order of magnitude larger as well, partially due to more advanced error correction algorithms.
It was released, but cDc put all of their effort into press releases on it, and not what they said at the conference. It was hard to pick out the important stuff out of an hour of assinine skits. As an aside, I think Oxblood Ruffin (sp?) had some of the most important stuff to say, unfortunatly he was so awful at using a microphone that finally one of the other cDc folks went up on stage to pass on what he was saying.
Second, a quick definition of what this is all about: it appears to be a collection of great scientific and programming works to be used as a primer for new programmers.
A collection of great programming works? Yes. To be used as a primer for new beginners? Not really. TAOCP (The Art Of Computer Programming) is one of the best advanced computer science books ever written. It doesn't teach you how to write object-oriented code, or even "hello world". Instead it sets down algorithms, and techniques to create algorithms to solve problems elegently.
Giving this to a new programmer would be pretty much a worthless exercise, as the code in the book isn't even written in a real (as in used in the real world) language, but rather one made up in academia to be Turing Complete and demonstrate algorithms. I keep all three volumes on my desk, and find that they can be incredibly useful in assisting with some problems. However, they're never going to tell me why my program has a memory leak, or the difference between reference and value.
The previous volumes of TAOCP have had numberous errors (Knuth pays $2.56 for every one you find) just due to the complexity of the material they cover, combined with the changing world of computer science. Hopefylly getting out pre-prints of the book, bit by bit, will help get alot of these fixed before the 1st edition.
I have no idea how you were expecting to "allay" the confusion of others, when you don't even seem to understand it yourself.
Bah, even when they do create an English website, so often the Japanese has much more info. Check out Nikon or Yamaha for example, even if you can't read Japanese, the sites are much much more useful.
So that's cute, but it degrades in the city for reasons similar to why GPS fails... Big metalic objects. At many points in time, you are probably in sight of 1-2 of the cell towers you are communicating with. This does mean that you have an idea of the area you are in. As soon as your signal path is no longer direct propegation however (ie, bouncing off buildings) all you get is an upper bounds as to someone's distance from a tower. Timings, strength, and even if you were to try to calculate angles from the tower are all going to be inaccurate unless you are in a big open space. Even better, in a city towers are closely packed enough that you might have a path to a tower further away from you than another you cannot reach just due to position, making the possible area outside the cell.
So yeah, I know what city my "car/pet/wife/computer" is in, but if any of them is stolen/lost/having a heart attack/being broken into, the best you've got is a big square.
Carnegie Mellon University has had a wireless network for years now. A few years ago all of the academic buildings had full coverage, and in the past year this has been extended to dorms and most outdoor areas.
The computer science department at CMU as well as the Human Computer Interaction Institute (HCII) and the department of Electrical and Computer Engineering have been putting out papers on actual implementations of campus location systems. Most deal with its use for contextual/location aware computing (one of the more recent papers). Although some have dealt with the privacy implications (I should know, I was an author of one published at IEEE Wireless 2001). Project Aura deals with quite a bit of reasearch around what can be done positivly with this technology as well.
As one last thing, I wrote software to poll wavepoints and figure out a location over 1.5 years ago... It was less than 50 lines of C, so I have trouble being impressed by this.
Even in a current day micropayment system you need to take the money in chunks. This is due to the fact that there is overhead for each transaction, and expecially with things like credit cards. Unless a user is billed in certain minimum chunks, the overhead dominates the actual payment. In a full-blown micropayment system, you'd probably give some chunk of money to one place, and they'd give it to sites, a few fractions at a time. In this case the fraction is always $0.005/page, rounding up to the nearest $5.00
People have been saying forever that someone was going to do a site with some form of a micropayment system. It's interesting to see/. taking the lead here. I really have to wonder though, how many people are bothered by the ads on/. ? Quite a few of the people here read through a junkbuster proxy of one form or another, so such things never even appear. Hopefully there isn't an effort to push people in the way of paying (more annoying ads, etc). Because that would be truly unfortunate.
Best of luck to you, but I really don't expect to see that you will have much for sales.
Slashdot Mirror
Right before The Last HOPE was announced, Vornado Realty Trust announced that they were tearing down the Hotel Pennsylvania before 2010 (when the next HOPE presumably would be). While there are certainly other venues in NYC, finding one the size of the space HOPE has in the Hotel Penn at a price that HOPE could be under $100 has appeared pretty hopeless.
I'm a reasonably heavy DSLR user who shoots on a Nikon D2H. I have shot for fashion and dance shows where I leave with over 1500 RAW photos (I attach my camera directly to a Powerbook which has a 250GB firewire drive attached). I've tried using iPhoto for managing my photos, as most of the professional workflow programs with databases are thousands of dollars to say hello. iPhoto essentially falls over and dies with those kind of numbers. iPhoto also doesn't actually handle RAW images, it converts them over to JPEG using a rather mediocre converter.
I used to use Photoshop CS for "developing" my raw images, but most of its capabilities are focused around working with the photo once you've imported it as a PSD, and not around manipulating the photo itself. Along with many other photographers I've discovered CaptureOne is incredibly useful for non destructive processing of RAW images, as well as doing a wonderful job on noise reduction, color noise, banding, white balance, exposure, and levels.
I was hoping Aperture could replace CaptureOne and iPhoto for me, while allowing me to contine to use Photoshop when I wanted to edit a photo rather than just process a RAW image. As far as I can tell, this is dead on what Apple intended Aperture for.
To start off, I imported 3 iPhoto libraries with a total of 45,000 images into Aperture. To my surprise, it also imported all album and roll data with it (I was expecting to end up with a flat photo space) as well as importing all NEFs and the jpegs iPhoto had created automatically as different versions of the same photo. It's clear that the upgrade path from iPhoto to Aperture was well thought out.
Aperture seems to be very good at handling a large image database. I now have 45,000 photos in a single Aperture library, and am not using more than 450MB of ram opening a window with all images in it (scrolling of course).
Aperture also claimed to be able to handle many of the non destructive RAW workflow duties I'd handled before with Capture One. That's a bit more of a mixed bag. The white balancing loupe doesn't work nearly as well as Capture One's and occasionally creates psychadelic white balances in the process. The sharpening and noise reduction algorithms are nowhere near as good as Capture One's, and color noise reduction seems to be almost non existant on high exposure shots. Before someone points out that this is what Photoshop or some other tool is for, Aperture only exports PSDs or TIFFs to other applications so it has to handle all RAW processing itself.
If Apple can figure out how to handle RAW images better, Aperature could really become an incredible product. As it is, the workflow management, versioning, and just plain dealing with tons of images seem to be really nice.
If you only have one firewire drive that you save for emergencies, this actually is a very useful thing.
Given a standard bios these machines should have an incompatable partition structure to current Macs (as has already been noted). Have to wonder if the Intel Macs will be able to still handle slices and read old disks. There were also some rumors that the endianness may be different on HFS for "Mactel" for internal binary structures (and therefore incompatable filesystems). I'm assuming not as it would create a problem for iPods as well as other firewire hard drives, but boot support for different partitioning schemes worries me a bit more.
I don't care if I can't run Forth in my bios any more (in fact, Intel's new BIOS scheme they did for the Itanium is nice) but I really hope that ease of integration with intel chipsets doesn't get in the way of legacy support.
FYI, the full text of the risk that Groklaw is quoting from:
If there's any group out there who is claiming that they are doing something that the SEC is going to have to investigate, they need to disclose the risk to make a shareholder lawsuit less likely.
If there is some risk that a company doesn't disclose, where there was some way for them to know that it was a posibility, they can be sued by shareholders if the stock goes down as a result of it comes true.
You often find really silly risks listed in safe harbor statement like that. For example, Walmart lists as a risk that they may not be able to buy from certain vendors if political instability takes place in their country. They also disclose that they won't do as well if they can't hire good employees.
Also, if they know about a SEC investigation or lawsuit against them, a company would usually give more information than that in a 10k (lest the SEC investigate them for the way they disclose their risks in their 10k).
This was posted already this week in an attempt to karma whore.
You're confusing emulation with virtualization. Storage virtualization is a really big deal these days. Check out the book "Virtual Storage Redefined: Technologies and applications for storage virtualization" by Paul Massiglia (of Veritas).
Isolation, and performance guarintees on shared systems are often more important than raw performance in something like a datacenter environment.
Adam
I've seen several people now post sessions they've had with "Snubby". Snubby is assuming that people are ordering things in a specific order. A session I just had with it:
telnet 64.94.110.11 25
Trying 64.94.110.11...
Connected to 64.94.110.11.
Escape character is '^]'.
220 snubby3-wceast Snubby Mail Rejector Daemon v1.3 ready
250 OK
250 OK
550 User domain does not exist.
250 OK
221 snubby3-wceast Snubby Mail Rejector Daemon v1.3 closing transmission channel
Connection closed by foreign host.
That's right. It doesn't parse the input at all (I just hit Enter a bunch of times). If you have multiple RCPT lines, or have an extra command in there anywhere, you will get an OK in the wrong place and it will look like you have succeeded.
Adam
I'm not claiming Linux doesn't have any problems, but take a closer look how each of the vulns you just posted effects overall system security. I think you'd find that the only one which could be exploited in most systems is the kdelibs problem, if you were using Konquerer.
I have to go back to August 1st to find a remote buffer overflow which effects a substantial portion of the Debian crowd. That would be the vulnerability in the X truetype font server. It runs as nobody, so not much can be done there either. Take a look at the Debian security advisories: for 2003. I think you'll find that the principles of least needed security level, and installing only what is asked for have served them well.
Additionally, look at how many of the packages which have had problems are a part of the core operating system. If Microsoft also put out advisories for many of the common programs run on its operating systems, the list might run a bit longer.
You also seem annoyed that all of these security updates need to be installed, but ignore the fact that you would have had to install many more of this number of upgrades to be running all current versions. Debian Stable is rock solid, but too boring for most of us.
I keep my Linux, OSX, Solaris and Windows systems equally up to date on security patches so I have an equal lack of problems on all of them, but you're comparing a kettle to a.... 50 gallon lobster pot I guess.
Adam
I've seen two misstatements repeated over and over again in this discussion. Folks have been suggesting that flash cards like this might represent the future of hard drives. They've also been mentioning that it has a higher speed than the IBM microdrive.
.5 seconds as well. Typical UNIX filesystems like ext2 or ffs, keep their data structures in fixed locations. Most writes are to metadata, and they will cause the metadata parts of a CF card to be erased and overwritten over and over again. Unlike a hard drive which can survive almost unlimited cycles like this, you will only get a few thousand in flash memory. Copying a set of files might burn out some cells in a single operation.
While this is true in a camera, where you tend to erase an entire card and then fill it in a linear fasion, this isn't true when you use it as a hard drive. Flash memory has two things which make it unique, slow erasures, and limited numbers of cycles. Unlike a hard drive, where you can simply overwrite data, in flash memory you have to erase a region of it first. Usually you also have to erase a much larger region than a filesystem block (64k vs 4k). These erasures can be as painful as
The log-structured filesystem (lfs) presents a partial solution to this, by writing data in blocks, deleting it in blocks, and writing to the end of a disk before starting over again. Unfortunatly, lfs becomes unefficient once fragmentation starts to set in, as a "cleaner" is necessary to group data back into blocks.
I still think one of these would be cool in my camera, but I want a 4G microdrive for my computer.
Adam
They even seem to have said that new high-end calculators are coming out on HP's site. I'm curious where the development effort for this is coming from. I don't think the calculator folks from Australia or France survived the great Carly purge.
adam
IA-64 Allows for it as well. It also has region support, so that executable and stack regions of memory can be completely separate.
adam
I do have some experience hotswapping drives. Linux sort of handles it. echo `scsi remove-single-device 0 0 1 0` >
echo `scsi add-single-device 0 0 1 0` >
Adam
We have (or maybe I should say had) a Roomba in our house. I believe it was a beta model. It did a much better job than any of us expected making it around college student rooms, around in a bathroom, and even our porch. The only real complaint we had with its operations was the small size of it's container for storing whatever it vacumed.
It had quite a few nifty features. The led on it slowly changed from green, to yellow, to red as the battery drained. It'd be nice to see that on a notebook computer! Being a house full of computer science majors, quite a bit of time was spent figuring out what its algorithm was for room coverage. While we didn't get it all quite worked out, it seemed to hit all of the room.
Unfortunatly, it met a fairly quick end. After about two days, we found it running in a circle. Opening it up, we discovered that one of the wheel motors had actually siezed. We still haven't been able to find the appropriate motor on mouser or digikey. It doesn't matter too much as the final version should be on its way to us soon enough.
This is not the case. SCSI drives have much more testing done on them, and much more predictible performance under certain types of workloads. If you test a drive with a very low queue depth, as the tests in the original article were, you aren't likely to see a difference. However, if you start allowing for queue reordering, and more seeking, the performance curves will change. Additionally, the number of attempts that a SCSI drive will make to recover after an error is much higher than an IDE. The amount of code in the firmware is actually an order of magnitude larger as well, partially due to more advanced error correction algorithms.
adam
Unless of course it appears in a computer science paper... Quite a few papers on storage I've seen assume log base 2.
It was released, but cDc put all of their effort into press releases on it, and not what they said at the conference. It was hard to pick out the important stuff out of an hour of assinine skits. As an aside, I think Oxblood Ruffin (sp?) had some of the most important stuff to say, unfortunatly he was so awful at using a microphone that finally one of the other cDc folks went up on stage to pass on what he was saying.
A collection of great programming works? Yes. To be used as a primer for new beginners? Not really.
TAOCP (The Art Of Computer Programming) is one of the best advanced computer science books ever written. It doesn't teach you how to write object-oriented code, or even "hello world". Instead it sets down algorithms, and techniques to create algorithms to solve problems elegently.
Giving this to a new programmer would be pretty much a worthless exercise, as the code in the book isn't even written in a real (as in used in the real world) language, but rather one made up in academia to be Turing Complete and demonstrate algorithms. I keep all three volumes on my desk, and find that they can be incredibly useful in assisting with some problems. However, they're never going to tell me why my program has a memory leak, or the difference between reference and value.
The previous volumes of TAOCP have had numberous errors (Knuth pays $2.56 for every one you find) just due to the complexity of the material they cover, combined with the changing world of computer science. Hopefylly getting out pre-prints of the book, bit by bit, will help get alot of these fixed before the 1st edition.
I have no idea how you were expecting to "allay" the confusion of others, when you don't even seem to understand it yourself.
Adam Pennington
Bah, even when they do create an English website, so often the Japanese has much more info. Check out Nikon or Yamaha for example, even if you can't read Japanese, the sites are much much more useful.
So that's cute, but it degrades in the city for reasons similar to why GPS fails... Big metalic objects. At many points in time, you are probably in sight of 1-2 of the cell towers you are communicating with. This does mean that you have an idea of the area you are in. As soon as your signal path is no longer direct propegation however (ie, bouncing off buildings) all you get is an upper bounds as to someone's distance from a tower. Timings, strength, and even if you were to try to calculate angles from the tower are all going to be inaccurate unless you are in a big open space. Even better, in a city towers are closely packed enough that you might have a path to a tower further away from you than another you cannot reach just due to position, making the possible area outside the cell.
So yeah, I know what city my "car/pet/wife/computer" is in, but if any of them is stolen/lost/having a heart attack/being broken into, the best you've got is a big square.
The computer science department at CMU as well as the Human Computer Interaction Institute (HCII) and the department of Electrical and Computer Engineering have been putting out papers on actual implementations of campus location systems. Most deal with its use for contextual/location aware computing (one of the more recent papers). Although some have dealt with the privacy implications (I should know, I was an author of one published at IEEE Wireless 2001). Project Aura deals with quite a bit of reasearch around what can be done positivly with this technology as well.
As one last thing, I wrote software to poll wavepoints and figure out a location over 1.5 years ago... It was less than 50 lines of C, so I have trouble being impressed by this.
Even in a current day micropayment system you need to take the money in chunks. This is due to the fact that there is overhead for each transaction, and expecially with things like credit cards. Unless a user is billed in certain minimum chunks, the overhead dominates the actual payment. In a full-blown micropayment system, you'd probably give some chunk of money to one place, and they'd give it to sites, a few fractions at a time. In this case the fraction is always $0.005/page, rounding up to the nearest $5.00
People have been saying forever that someone was going to do a site with some form of a micropayment system. It's interesting to see /. taking the lead here. I really have to wonder though, how many people are bothered by the ads on /. ? Quite a few of the people here read through a junkbuster proxy of one form or another, so such things never even appear. Hopefully there isn't an effort to push people in the way of paying (more annoying ads, etc). Because that would be truly unfortunate.
Best of luck to you, but I really don't expect to see that you will have much for sales.