In a server based product, a development team can get a lot of mileage out of having a couple of reasonably beefy machines to do integration and testing on. What is the complaint that leads you to want to replace machines? Do builds take too long? Are the tools too slow? Instead of spending $600 on each developer, which as you know, won't really fix all their woes, how about spending $100 on memory for each person, as several people have suggested. Then buy three or four $2000 systems, and parcel them out to subprojects.
If you were a lead developer, I would suggest a slightly different tactic for those machines, and that perhaps a review of the tools the developers use might be in order, but as a sysadmin, no one is going to want to hear from you that the build process needs to be seriously overhauled.
I say this, because if your project is small enough to be considered a candidate for being shelved, and developing has gotten painful on 400 Mhz machines, then I suspect that like many projects, the build environment has grown by accretion, and no one has bothered to get off their ass and do anything about how long a build takes. Each change only adds another five to ten seconds to the build process, but over the course of a year, you get nickeled and dimed to death on these little slowdowns. And while some tools might be good enough to produce a working prototype, a lot of the IDEs end up falling to pieces as soon as you introduce source control, and serious project requirements. Sometimes it's better to just pick a good compiler, a good source control system, and a decent build tool, and chuck the rest out the window.
At my company, a coworker and I were recently pulled off of our project and onto the company's main project. For the last few weeks, we've been allocating a good bit of time to tuning the build process. What used to take 35 minutes for a build now takes just over 8 minutes on a 1Ghz machine, and can be trimmed to less than 3 minutes by turning off some more exhaustive building features. The time we spent will pay for itself in a couple of months. The last two place I worked at, I achieved similar results with less effort, but the development teams were smaller.
Currently, I'm working on a change to drop the build world time to 7 minutes, while he's working the kinks out of setting up a continuous integration machine. Once completed, people won't have to run the optional tests on their machines (most already don't). They can just rely on a 5 minute build on one beefy workstation (in this case, a Solaris box, as it's a Java/Ant/Perforce project) to tell them if they screwed up, instead of a disruptively long build cycle on their more meager development boxes.
Wasn't NASA recently bragging how one of their satellites recently designed (launched?) had a directional high gain antenna that was made out of hundreds of tiny nondirectional antennas. To achieve a directional signal, each antenna broadcasts at a slight time skew to achieve constructive interference in the direction you wished to broadcast (and presumably, in the opposite direction as well).
In theory, this should trickle down to consumer devices in 5 years or so. A portable device should be able to analyze the direction of incoming signals from the base station (by trangulation on its microantenna array) to figure out which way to broadcast the response.
When the mail browser doesn't appear to have been written by a ham-fisted clod.
The.8x mail reader is hands down faster than the.6x reader, but honestly, I could write a better performing mail reader with my big toe after a case of Guinness (and I'm a lightweight). The mail shouldn't gum up with only a couple hundred messages in my inbox.
Mr Goldman is really pushing his luck here. He's bet, at 50:1 odds, that someone can't compress truly random data. That's great. Except Random.org -isn't- truly random data.
If you look at the stats, you'll see that the ent program suggests that the data being generated is at least superficially compressible, in the neighborhood of one quarter of one percent. All Mike said was that you have to compress it 1 byte. Given a large enough file, and a good statistical compressor, it's entirely possible that he'd have to cough up the dough.
It may still be possible to honestly 'trick the trickster', here.
All of the benchmarks I've seen suggest that something like UCL (whose reference implementation is also under the GPL) beats the pants off of LZ77, and hence zlib, for data throughput. I wonder why they chose zlib. Out of habit?
-
Reality-based interfaces are inefficient
on
3D GUI Project
·
· Score: 1
When's the last time you were writing a letter, and flipped windows to check a web page?
When's the last time you were brushing your teeth and 'flipped windows' to hang a picture on a wall in the basement, and then return to brushing your teeth?
Or did you do like most of us, and put down the toothbrush, walk out of the bathroom, out into the garage for a hammer, back inside and down into the basement, over to the wall, nail in the hook, put up the painting, walk back upstairs, out into the garage to deposit the hammer, back inside and to the bathroom to finish brushing your teeth?
We do certain tasks in the computer because their real world counterparts take to long. Typing instead of using a pencil. Sending email instead of finding an envelope, getting that foul-tasting goo on your tongue, and putting the thing in the mailbox. Did you remember a stamp?
In short, don't think for a moment that modeling 3-D productivity tools after the real world is going to get you any sort of respect from your users.
Do anything else, so long as you remember that the human brain, as stated before, doesn't really deal well with 3D. Any arboreal skills our brains had for navigating in 3 dimensions fled some time ago, after we adapted to walking. You or I may retain this ability, but the average person doesn't.
I believe the proposed cause for this phenomenon is that breaking down sugar produces free radicals, which damage the cells. These are counteracted by antioxidants (assuming you're consuming enough, which most of us apparently don't) which are supposed to capture the free radicals before they do damage, most but not all of the time.
So if you break down less fuel, you create fewer free radicals, which means less cell damage, which means longer time between cell generations, which means the time bombs built into the cells goes off later, and the organism lasts longer.
If you think I have nothing better to do with my free time then whore myself out for karma points, you need to get a grip. It's been my experience that people who denegrate technologies that aren't part of the Linux Core Philosophy are far more likely to get karma than those who attempt to be the voice of reason.
Yes, I posted fairly commonly known information. At the time of my posting, several people had alluded vaguely to this information, true enough. But no one bothered to spell it out for those who haven't used BeOS, so they could stop making snide, ill-informed comments about the Capitalist Pig-Dogs.
This is not a portable unit. It's a home unit.
Not counting any potential for pirating software, it's safe to assume that a music lover could have upward of ten thousand songs on their hands. That's quite a few song titles and artists to search from, when you have an urge to listen to some song by a female artist that came out in 95 or 96, with something about icicles in it.
Without even trying, you could have that from a BeFS. (you'd have to add a 'lyrics' metadata to the song, of course, but that's hardly a stretch)
Journaling, Schmournaling. Be's filesystem keeps information about files in a metadata database. You can add arbitrary information types to this metadata, and you can search the filesystem for files.
For instance, you can add Artist, title, album, live/recorded, and bitrate to the information on the file. Then you could search for all live recordings by Tori Amos or Limp Bizkit, at a datarate above 128 kbps, sorted by title.
All without writing but one line of code (the search query)
do you envision the Plex86 project as a useful platform for modeling and testing new Operating Systems? If so, what effort have you taken, or do you plan to take, to support this use of the software?
Remember when McCarthy made Communist mean "daughter-raping baby killer"?
Charlie Chaplin lived out his golden years in the south of France because of Mister McCarthy and his little campaign.
Not that this refutes your core argument in any fashion. But the old joke about the difference between being British and American is that the British think a hundred miles is a long way, and the Americans think a hundred years is a long time, is fitting here.
Some will see an example from the fifties as more compelling than one from the early forties, less able to shrug it off.
Everyone is lauding the Crusoe for power consumption, but that's not the only important feature of the Transmeta architecture. The other, arguably more important feature is the divorce from backward compatibility. If they start encouraging you and I to write code to the Crusoe Instruction Set, then the next processor has to support that as well. Just like Intel still supports i8086 instructions.
Their only option for giving you what you want would be to provide the Crusoe analog to the Intel interpreter when they start producing 2nd generation processors, but that devolves into having to ship massive ROMS and memory-hungry code with their hardware. Does anyone really want that?
At my highschool, they had a week long hazing to get into NHS. It essentially amounted to spending a week acting like an idiot to prove how smart you are.
Suffice it to say, I declined filling out an application, and told people that I didn't approve of the initiation ceremony.
At my last job, I was the "Are you fucking nuts? No way in hell!" guy. (I'd like to first state that people were either upset or at least faked sadness when I left). As time passed, it became somewhat less effective, and quite tiring, so I had to start trying to force other people to do the bitching, which of course starts to feel a bit like office politics of the least interesting sort rather quickly.
If I'm going to play office politics, I'd rather be using it to improve my position of authority, not trying to make up for weakness above.
The company I just left just did this. No one at the top was materially affected by the fact that we lost a round of investment money. Fifteen percent of my peers were canned, and everyone's bonuses were canceled (they only reason, btw, that I stayed past July).
Did the management who were fucking up get their overinflated salaries cut? No. But it looks like they're considering killing off my entire office, even though we were doing about 75% of their actual R&D. All they see is a bunch of 'whiners'.
Hell yes, we're whining. We're whining because you guys have your heads up your asses, and you haven't bothered to send people from the head office to find out what our concerns are (or rather, you didn't until after the layoffs started). OF COURSE we're going to bitch about it.
While it's true that the local aggregate bandwidth will be unaffected, any P2P protocol with a concept of distance between peers, and file cloning will use less aggregate bandwidth from the perspective of the entire internet.
I'm sure someone else has coined a clever phrase for this already, but let's refer to this notion of bandwidth * hops in units of MBit-hops (akin to a kilowatt hour).
If I download a file from my ISP, am I not using fewer megabit hops than if I download the same file from Outer Mongolia? Of course I am. This reduces the overall congestion between my ISP and Outer Mongolia, freeing those megabit hops to either be wasted, or used by someone else.
Let's say Mr Merkle designs a chip that's 99.9999% efficient at reversing the computation, including factoring in the extra gate counts for the circuit.
Now, throw a generously tight 10000 bit twiddles at each key.
That's 1% of his original figure, for ~84 days of the full output of the sun. Yeah, we should be able to manage that, no problem.
MacOS?
You've got to be fucking kidding me.
MacOS X is the first modern operating system (and I'm talking 1994 definition, here) to come out of Apple, and it hasn't shipped yet. Why don't you take your toy fifteen year old OS and go outside and play?
Stop imagining that you're the Great Altruist, providing access to your Great and Bountiful Resources, for the Greater Good, with no personal gain involved.
It's a Co-op, dummy. It only works if you 'pay' for what you get out of it. You're no more special than every other member of the group. You want to be able to pull files from Freenet at a reasonable clip? Then you put up a Freenet node, and join the co-op. You want to upload files implicating a former employee with a modicum of anonymity? Then you put up a Freenet node, and join the co-op.
If you don't, don't. The world won't end without you. It'll be okay.
If you have one of these units on site, I would hope that you had an AC system as well. If the AC is properly sized to drive a 25 degree delta-t with the microgen running at full blast (hint: cool the intake, but not the exhaust pipe), then you should be fine in 99% of the inhabited areas in the US. In the other 1%, you might need a geothermal heat exchanger based cooling system in order to work it out, but IIRC, those pay for themselves in about 3 years, anyway.
you've defended someone's right to say something you morally oppose.
Until then, it's just posturing.
I think he's entirely in line. It is you who isn't.
If you don't want people to be able to say what they want to, then don't run a Freenet node. It's that simple.
(oh, and also, please don't run for public office. We've got enough of your sort already)
-
Then keep reading it, and wait for the Clue Stick
on
Freenet 0.3 Released
·
· Score: 1
It's a distributed caching architecture. All copies are created as a reaction to user-driven interaction with the network. Joe clicks a button, sending a request to Cache1, which sends a request to Cache4, which sends a request to Cache314, which sends a request to Cache1425, which has the actual document, creating up to three new copies of the file as it travels to Joe.
This is the essence of Freenet, and it's a pull architecture, bucko.
The only exceptions to this would be initial insertion (which would be a tiny fraction of all requests), and any potential future optimisations to the system, which would assuredly take into account scalability concerns.
Slashdot Mirror
In a server based product, a development team can get a lot of mileage out of having a couple of reasonably beefy machines to do integration and testing on. What is the complaint that leads you to want to replace machines? Do builds take too long? Are the tools too slow? Instead of spending $600 on each developer, which as you know, won't really fix all their woes, how about spending $100 on memory for each person, as several people have suggested. Then buy three or four $2000 systems, and parcel them out to subprojects.
If you were a lead developer, I would suggest a slightly different tactic for those machines, and that perhaps a review of the tools the developers use might be in order, but as a sysadmin, no one is going to want to hear from you that the build process needs to be seriously overhauled.
I say this, because if your project is small enough to be considered a candidate for being shelved, and developing has gotten painful on 400 Mhz machines, then I suspect that like many projects, the build environment has grown by accretion, and no one has bothered to get off their ass and do anything about how long a build takes. Each change only adds another five to ten seconds to the build process, but over the course of a year, you get nickeled and dimed to death on these little slowdowns. And while some tools might be good enough to produce a working prototype, a lot of the IDEs end up falling to pieces as soon as you introduce source control, and serious project requirements. Sometimes it's better to just pick a good compiler, a good source control system, and a decent build tool, and chuck the rest out the window.
At my company, a coworker and I were recently pulled off of our project and onto the company's main project. For the last few weeks, we've been allocating a good bit of time to tuning the build process. What used to take 35 minutes for a build now takes just over 8 minutes on a 1Ghz machine, and can be trimmed to less than 3 minutes by turning off some more exhaustive building features. The time we spent will pay for itself in a couple of months. The last two place I worked at, I achieved similar results with less effort, but the development teams were smaller.
Currently, I'm working on a change to drop the build world time to 7 minutes, while he's working the kinks out of setting up a continuous integration machine. Once completed, people won't have to run the optional tests on their machines (most already don't). They can just rely on a 5 minute build on one beefy workstation (in this case, a Solaris box, as it's a Java/Ant/Perforce project) to tell them if they screwed up, instead of a disruptively long build cycle on their more meager development boxes.
Wasn't NASA recently bragging how one of their satellites recently designed (launched?) had a directional high gain antenna that was made out of hundreds of tiny nondirectional antennas. To achieve a directional signal, each antenna broadcasts at a slight time skew to achieve constructive interference in the direction you wished to broadcast (and presumably, in the opposite direction as well).
In theory, this should trickle down to consumer devices in 5 years or so. A portable device should be able to analyze the direction of incoming signals from the base station (by trangulation on its microantenna array) to figure out which way to broadcast the response.
Apparently you and I are the same sort of 'idiot', since your inbox is about the same size as mine.
I'm glad to hear that this issue has been addressed, finally. Color me happier.
-
When the mail browser doesn't appear to have been written by a ham-fisted clod.
.8x mail reader is hands down faster than the .6x reader, but honestly, I could write a better performing mail reader with my big toe after a case of Guinness (and I'm a lightweight). The mail shouldn't gum up with only a couple hundred messages in my inbox.
The
-
Mr Goldman is really pushing his luck here. He's bet, at 50:1 odds, that someone can't compress truly random data. That's great. Except Random.org -isn't- truly random data.
If you look at the stats, you'll see that the ent program suggests that the data being generated is at least superficially compressible, in the neighborhood of one quarter of one percent. All Mike said was that you have to compress it 1 byte. Given a large enough file, and a good statistical compressor, it's entirely possible that he'd have to cough up the dough.
It may still be possible to honestly 'trick the trickster', here.
-
All of the benchmarks I've seen suggest that something like UCL (whose reference implementation is also under the GPL) beats the pants off of LZ77, and hence zlib, for data throughput. I wonder why they chose zlib. Out of habit?
-
When's the last time you were writing a letter, and flipped windows to check a web page?
When's the last time you were brushing your teeth and 'flipped windows' to hang a picture on a wall in the basement, and then return to brushing your teeth?
Or did you do like most of us, and put down the toothbrush, walk out of the bathroom, out into the garage for a hammer, back inside and down into the basement, over to the wall, nail in the hook, put up the painting, walk back upstairs, out into the garage to deposit the hammer, back inside and to the bathroom to finish brushing your teeth?
We do certain tasks in the computer because their real world counterparts take to long. Typing instead of using a pencil. Sending email instead of finding an envelope, getting that foul-tasting goo on your tongue, and putting the thing in the mailbox. Did you remember a stamp?
In short, don't think for a moment that modeling 3-D productivity tools after the real world is going to get you any sort of respect from your users.
Do anything else, so long as you remember that the human brain, as stated before, doesn't really deal well with 3D. Any arboreal skills our brains had for navigating in 3 dimensions fled some time ago, after we adapted to walking. You or I may retain this ability, but the average person doesn't.
-
I believe the proposed cause for this phenomenon is that breaking down sugar produces free radicals, which damage the cells. These are counteracted by antioxidants (assuming you're consuming enough, which most of us apparently don't) which are supposed to capture the free radicals before they do damage, most but not all of the time.
So if you break down less fuel, you create fewer free radicals, which means less cell damage, which means longer time between cell generations, which means the time bombs built into the cells goes off later, and the organism lasts longer.
Or so went the theory.
-
If you think I have nothing better to do with my free time then whore myself out for karma points, you need to get a grip. It's been my experience that people who denegrate technologies that aren't part of the Linux Core Philosophy are far more likely to get karma than those who attempt to be the voice of reason.
Yes, I posted fairly commonly known information. At the time of my posting, several people had alluded vaguely to this information, true enough. But no one bothered to spell it out for those who haven't used BeOS, so they could stop making snide, ill-informed comments about the Capitalist Pig-Dogs.
-
This is not a portable unit. It's a home unit.
Not counting any potential for pirating software, it's safe to assume that a music lover could have upward of ten thousand songs on their hands. That's quite a few song titles and artists to search from, when you have an urge to listen to some song by a female artist that came out in 95 or 96, with something about icicles in it.
Without even trying, you could have that from a BeFS. (you'd have to add a 'lyrics' metadata to the song, of course, but that's hardly a stretch)
-
Journaling, Schmournaling. Be's filesystem keeps information about files in a metadata database. You can add arbitrary information types to this metadata, and you can search the filesystem for files.
For instance, you can add Artist, title, album, live/recorded, and bitrate to the information on the file. Then you could search for all live recordings by Tori Amos or Limp Bizkit, at a datarate above 128 kbps, sorted by title.
All without writing but one line of code (the search query)
-
do you envision the Plex86 project as a useful platform for modeling and testing new Operating Systems? If so, what effort have you taken, or do you plan to take, to support this use of the software?
-
Remember when McCarthy made Communist mean "daughter-raping baby killer"?
Charlie Chaplin lived out his golden years in the south of France because of Mister McCarthy and his little campaign.
Not that this refutes your core argument in any fashion. But the old joke about the difference between being British and American is that the British think a hundred miles is a long way, and the Americans think a hundred years is a long time, is fitting here.
Some will see an example from the fifties as more compelling than one from the early forties, less able to shrug it off.
-
Everyone is lauding the Crusoe for power consumption, but that's not the only important feature of the Transmeta architecture. The other, arguably more important feature is the divorce from backward compatibility. If they start encouraging you and I to write code to the Crusoe Instruction Set, then the next processor has to support that as well. Just like Intel still supports i8086 instructions.
Their only option for giving you what you want would be to provide the Crusoe analog to the Intel interpreter when they start producing 2nd generation processors, but that devolves into having to ship massive ROMS and memory-hungry code with their hardware. Does anyone really want that?
-
At my highschool, they had a week long hazing to get into NHS. It essentially amounted to spending a week acting like an idiot to prove how smart you are.
Suffice it to say, I declined filling out an application, and told people that I didn't approve of the initiation ceremony.
-
At my last job, I was the "Are you fucking nuts? No way in hell!" guy. (I'd like to first state that people were either upset or at least faked sadness when I left). As time passed, it became somewhat less effective, and quite tiring, so I had to start trying to force other people to do the bitching, which of course starts to feel a bit like office politics of the least interesting sort rather quickly.
If I'm going to play office politics, I'd rather be using it to improve my position of authority, not trying to make up for weakness above.
-
The company I just left just did this. No one at the top was materially affected by the fact that we lost a round of investment money. Fifteen percent of my peers were canned, and everyone's bonuses were canceled (they only reason, btw, that I stayed past July).
Did the management who were fucking up get their overinflated salaries cut? No. But it looks like they're considering killing off my entire office, even though we were doing about 75% of their actual R&D. All they see is a bunch of 'whiners'.
Hell yes, we're whining. We're whining because you guys have your heads up your asses, and you haven't bothered to send people from the head office to find out what our concerns are (or rather, you didn't until after the layoffs started). OF COURSE we're going to bitch about it.
-
While it's true that the local aggregate bandwidth will be unaffected, any P2P protocol with a concept of distance between peers, and file cloning will use less aggregate bandwidth from the perspective of the entire internet.
I'm sure someone else has coined a clever phrase for this already, but let's refer to this notion of bandwidth * hops in units of MBit-hops (akin to a kilowatt hour).
If I download a file from my ISP, am I not using fewer megabit hops than if I download the same file from Outer Mongolia? Of course I am. This reduces the overall congestion between my ISP and Outer Mongolia, freeing those megabit hops to either be wasted, or used by someone else.
-
Sun has NOT comitted to parameterized types for JDK 1.4. I see this as a serious step backward in their progress toward polishing Java.
-
I'm not quite sure I buy your conjecture.
Let's say Mr Merkle designs a chip that's 99.9999% efficient at reversing the computation, including factoring in the extra gate counts for the circuit.
Now, throw a generously tight 10000 bit twiddles at each key.
That's 1% of his original figure, for ~84 days of the full output of the sun. Yeah, we should be able to manage that, no problem.
He says, tongue firmly planted in cheek.
-
MacOS?
You've got to be fucking kidding me.
MacOS X is the first modern operating system (and I'm talking 1994 definition, here) to come out of Apple, and it hasn't shipped yet. Why don't you take your toy fifteen year old OS and go outside and play?
Put up or shut up, indeed.
-
Stop imagining that you're the Great Altruist, providing access to your Great and Bountiful Resources, for the Greater Good, with no personal gain involved.
It's a Co-op, dummy. It only works if you 'pay' for what you get out of it. You're no more special than every other member of the group. You want to be able to pull files from Freenet at a reasonable clip? Then you put up a Freenet node, and join the co-op. You want to upload files implicating a former employee with a modicum of anonymity? Then you put up a Freenet node, and join the co-op.
If you don't, don't. The world won't end without you. It'll be okay.
-
If you have one of these units on site, I would hope that you had an AC system as well. If the AC is properly sized to drive a 25 degree delta-t with the microgen running at full blast (hint: cool the intake, but not the exhaust pipe), then you should be fine in 99% of the inhabited areas in the US. In the other 1%, you might need a geothermal heat exchanger based cooling system in order to work it out, but IIRC, those pay for themselves in about 3 years, anyway.
-
you've defended someone's right to say something you morally oppose.
Until then, it's just posturing.
I think he's entirely in line. It is you who isn't.
If you don't want people to be able to say what they want to, then don't run a Freenet node. It's that simple.
(oh, and also, please don't run for public office. We've got enough of your sort already)
-
It's a distributed caching architecture. All copies are created as a reaction to user-driven interaction with the network. Joe clicks a button, sending a request to Cache1, which sends a request to Cache4, which sends a request to Cache314, which sends a request to Cache1425, which has the actual document, creating up to three new copies of the file as it travels to Joe.
This is the essence of Freenet, and it's a pull architecture, bucko.
The only exceptions to this would be initial insertion (which would be a tiny fraction of all requests), and any potential future optimisations to the system, which would assuredly take into account scalability concerns.
-