I have no idea how FEAD works, but here's how I'd do something similar:
A large portion of shipped executables are blocks of standard code from the compiler. If you're using a Microsoft compiler, you can strip out the standard chunks and pull those chunks in from the binaries that are already in Windows.
If you're using another compiler, you can still probably do the same kind of thing: some intelligent block compression with the included code that'll do better than the "dumb" compression from "zip" or other algorithms.
You could combine that with compiler space optimization tricks, too: loop RE-rolling, for example. A lot of compilers do tricks to make code faster. Not many do things to explicitly make code smaller. A "shriking" compiler (or a disassembler/reassembler) that produced small code combined with an "expander" app that made the code bigger but faster could make very small apps.
I don't know the first think about working with 3-d molecular apps, but when I went to the PyMOL site listed in the submission, I saw the following on the main page. It seems to answer the question.
(4/18/2003) Accelerated OpenGL Hardware Recommendations (updated 4/30/2003): As we recently updated our development systems, we had a chance to obtain and test some exciting new graphics hardware! Thus, we can now provide the following endorsements:
For monocular visualization under Windows, we recommend the ATI Radeon 9700 Pro (approx. $400). This is by far the fastest OpenGL engine we have ever seen PyMOL run on!
For stereo visualization under Windows, we are re-evaluting our recommendation of the nVidia Quadro4 750XGL (approx. $500) because we and other users have encountered image stability problems (on Windows only!).
For monocular visualization under Linux, we recommend the ATI Radeon 8500 (approx. $200). This high-performance card is now supported out-of-the-box with full OpenGL acceleration under RedHat 9.0 (XFree86 4.3) -- no manual configuration is required!
For stereo visualization under Linux, we recommend the ATI FireGL2 (approx. $600). It is not the fastest card in the world, but IT SIMPLY WORKS. If you need better performance, go with the nVidia Quadro4 750XGL, but be prepared for driver updates.
For stereo visualization under Mac OSX, we recommend the nVidia GeForce4 Titanium (approx. $400). However, we would expect even better performance from the ATI Radeon 9700 (not yet tested).
For monocular visualization on a laptop, we recommend the nVidia GeForce4 Go 460 chip running under Windows. However, we also expect excellent performance from the ATI Mobility Radeon 9000 chip, which will likely do better with smooth (antialiased) lines.
If you're running Big Expensive Software (Oracle or DB2, for example), you want to run a configuration that your Big Expensive Software vendor supports, and you don't want to upgrade any more than necessary. Same thing's true for in-house apps. Upgrades can take a lot of manpower (thus money). In cases like that, plunking down a few hundred (or even thousand) for a support contract and/or a configuration that's guaranteed to be supported for five years is a great deal.
On the other hand, if you're running "commodity" services and servers (i.e. the kind of stuff that's installed by default when you click the "server" box during a RH install), you probably have failover anyway, so there's no harm in running the cheap/free version of RH with some sort of internal RPM update service (AutoRPM, rpm-update, or the like) and upgrading all of those machines every year. On the flip side, it might still be cheaper for your organization to buy Enterprise or Workstation and put your "upgrade time" to use somewhere else.
The big case for the free version, IMO, is software development and desktop clients. If you're WRITING software for RH, your developers should be using the platform that they're developing for. Also, if you have a whole lot of desktop linux users, the per-seat price for the Workstation version can add up quickly.
I've not encountered a machine in ages where the extra percentage of CPU or disk usage involved in journalling a normally read-only partition mattered much. Or, more accurately, when load matters, I either make the disk read-only (if things don't change) or journalling is even more important than usual (if things do change).
If your disks are 99% full and journalling overhead matters, the solution isn't "turn off journalling", it's "buy a bigger disk". Same with CPU. Sure, there are quick tricks like "noatime" for heavily-used filesystems, but most of the time, spending money on people who will optimize the system is more expensive than just buying more hardware and making a bigger system.
My question'd be "Why bother installing any FS that's not the default for your OS?" Unless the answer's "I'm doing something so specialized that the default FS won't work", you're probably doing something wrong.
I have a 12" Powerbook that gets warm enough in the left front corner to cause extreme discomfort. I'm too much of a wuss to leave my hand there long enough to see if it'll actually cause physical burns, but I suspect that it would.
You can download the source, untar, change to the directory, and "./btdownloadheadless.py --url http://foo.bar.blah/goatse.cx.torrent" This requires Python, which is either built-in or comes with the (free) developer tools.
If you argue "that's too complicated!", you probably don't want an iso image of a compile-from-source linux distro in the first place.
I have a.torrent set up for the KDE version of the Gentoo Live CD here (host is fnord.andrew.cmu.edu. Used IP to avoid DNS slashdotting).
The md5 checks out, but I haven't yet burned it or tested it.
Re:You sir, are obviously not a drinker
on
Absolute OpenBSD
·
· Score: 1
Smirnoff distillery was backing the movies so they modified to drink to add vodka to it
Ah, that explains it. I wondered why the hell anyone would drink a vodka martini. Gin and vermouth go together nicely (esp. with the tiny bit of olive flavor). Vodka and vermouth just tastes like really crappy vodka (Gilby's, anyone?)
I noticed one of the recent Bond movies finally stopped pretending that they were martinis. You just see Bond drinking vodka shots. Now if only he'd drink good vodka...
The problem with qmail is that the author is a screaming loony (albeit a very smart one).
We all know how support works for open source products: you use the mailing lists, IRC channels, mail to the author, etc. Woe be unto you if you ask a qmail question that was answered 3 years ago (and is thus archived somewhere), or worse yet, you ask for functionality that qmail doesn't have and DJB deems "inessential".
I've not even asked questions, but I've read the archives. A surlier, more user-hostile community would be hard to find.
I work at a tech university (CMU) and went to a public university down the road (Pitt) a few years ago.
If you're going to a private "geek" university, pick up a decent laptop (with wireless if your school supports it) and a pile of extra batteries (and an external charger if possible). Laptops are damned near fashion accessories at many private universities. Bonus style points for a Powerbook. Go ahead and use it to take notes: everyone else will, and the professors (in the tech classes, at least) will often be wearing microphones to be heard over the typing. Really.
If you're going to a public university, the rules are different. Spent the money on a good monitor and an okay PC. Got an awesome custom case? Leave it at home. See if you can ram your PC into a POS Packard-Bell case scavanged from the neighbor's trash, and throw a brick in the bottom for good measure. Cover the monitor (not the vents!) and PC case with stickers to crap it up. You want this thing to scream "too crappy to steal!" Use a composition book (the kind with stitched pages) for each class. No laptops: the clicking will piss off your classmates. No spiral-bound notebooks: you will crush the spiral wire and end up switching notebooks mid-semester.
Regardless of university: get a small, cheap PDA. Don't use it for notes. Use it as an alarm/calendar and carry it everywhere. Set an alarm for 15 minutes before each class and recitation, and an alarm for two hours before each homework assignment is due: you want to allow time for the "Oh, shit!" followed by an emergency trip to your dorm/apartment to pick it up or (if you completely forgot) time to put something quick and shoddy together so that you can get partial credit. For major projects, put a reminder in for a week before it's due so that you can ask questions, get missing materials, etc. from the professor. Oh, yeah: sync religiously. Since you'll be packing this thing everywhere, you're going to lose or destroy at least one of them (beer and PDAs are mortal enemies).
Get a printer. Speed doesn't matter, but if you have a roommate, noise does. Whenever you want to print, so does the rest of the campus. Always keep an extra ream of paper and extra ink on hand.
You probably have a really fast network connection or a cheap network card. BT makea a lot of connections, and some cards/drivers tie up the kernel for a fraction of a second for each connection. It's not noticeable when you make a new connection every minute or two, but extremely noticeable when you're running something that's making dozens of connections every second.
The BT FAQ addresses the issue with windows cards, but I saw the same problem on my cheapo NE2000-compatable ISA ethernet card. I replaced it with a PCI card, and the problems went away.
You're right, but a bit of pedantry: The original Airport cards were based on the Lucent Wavelan II cards (and even had the same Lucent chips inside).
The original Lucent Wavelan cards (at least, the ones that I'm familiar with) were pre-Wi-Fi 900MHz and had a big deck-of-cards sized external antenna that you velcro-ed onto the top back of your laptop's screen.
An important question: Why set up wireless AP to wireless AP links for the apartment complex? That seems prone to failure.
Most installations that I've seen use wires to the APs (often with Power-Over-Ethernet, to reduce wiring to a single cable). That way, you have fewer channel collisions, less overall wireless traffic, and significantly lower latency. You can put all of your wires in hallways, basements, and other "public" areas where you can work without going into people's apartments. Too ugly? Use drop ceilings and ceiling-mounted cable runs. Not to mention that it's far harder for a repirman to mess up a cheap 10/100 switch than an elaborate antenna array.
Also, think hard about setting up some sort of monitoring system. You want to have someone on the way to fix a failing AP before the angry calls come in.
I've ranted about this before, but the short answer is that 802.11g isn't an actual standard yet. It's still under development, and you have absolutely no guarantee that the "802.11g*" (where the * takes you to a note -- in 4 point text -- saying that it's not really 802.11g) card that you buy today will work when the standard comes out.
One word of caution: DON'T get 802.11g stuff yet unless you have money to burn and want it now. Stick with 802.11b. A lot of companies (Apple and Linksys are the biggies) are selling stuff that's "based on the emerging 802.11g standard". Nobody knows yet whether that hardware will be compatable with the standard when it comes out.
Anyone who bought an X2 or K56Flex modem will be able to tell you about what "software upgradable" really means in the face of unknown standards. It means (if you're lucky) "the manufacturer will reluctantly exchange your product if you're willing to ship it to us and wait two months for us to send a different reconditioned unit with a new chip."
You can't afford $1 for a song. You have to use the money for food, instead.
Meanwhile, your.sig and home page refer to Transgaming, a place where you can pay $5 a month to get access to software that'll let you play $50 dollar Windows computer games. (or do you steal those, too?)
And you "might need an ibook". Like someone who can barely afford food could come up with two grand for a new computer.
Sorry, man. You had a good troll going, but the bullshitometer just went off the scale on that one.
Ah, nothing like a mature discussion. At least, you won't find anything like one around here.
I suggest that you go back and re-read the article that you're referring to and check out the license links in it. The person who buys your computers DOES get the software, including any upgrades that were made available under the support contract before the sale. They just don't get any MORE upgrades. In other words, the software IS transferrable. The support and upgrade contract isn't.
You'd be correct if there weren't multiple bugs in both the WEP protocol and in most common implementations. As it stands, most implementations (Cisco's is a notable exception) use the same keys for all traffic. That means that you can park your wireless notebook near an AP and deduce the keys in a few hours. If your wired office is next to mine, there's no way for me to sniff your traffic. If your wireless office is next to mine, I can crack your keys and see your traffic.
There are also fundamental key distribution problems in large organizations. I work at a university. We don't use WEP because we can't come up with a way to distribute the keys that doesn't also make it trivial for non-authorized users to use the network. We can use physical security to keep our wiring closets and network jacks safe. We can't put faraday cages around our buildings to keep the wireless network safe.
Essentially, Microsoft are saying "We're going to sell you a three-year non-transferrable support and upgrade contract." Nothing wrong with that. They also let you make payments instead of paying for the full three years up front. Again, nothing bad about that. They DO specify that you have to pay off the balance on any machines that you sell, though. How's that any different than, say, requiring you to pay off your bank loan before you sell your car?
It seems to be that nobody'd be complaining if they just required the entire payment for three years up-front and said"It's non-transferrable. Cope." People are pissed because Microsoft offers a payment plan, but they won't automatically transfer that plan.
I encourage people to put the redhat.isos onto other p2p networks (gnutella, etc.) as well as BT. Sure, BT's nifty, but it's good to get the data out as widely as possible. Not everyone can run BT.
After I've finished my download, I'll put it up on gnutella. My copy shouldn't be the only one.
I'm not sure that the argument is "It's hard" or "It can't be done." I don't use Gentoo because I don't want to lose use of my computer for the day or two that it can take to compile everything. Gaining 10% more speed isn't worth losing the use of my computer for a day every time a major update appears. With a slow computer, the compile time's a killer. With a fast machine, the extra efficiency is even less worthwhile.
Slashdot Mirror
I have no idea how FEAD works, but here's how I'd do something similar:
A large portion of shipped executables are blocks of standard code from the compiler. If you're using a Microsoft compiler, you can strip out the standard chunks and pull those chunks in from the binaries that are already in Windows.
If you're using another compiler, you can still probably do the same kind of thing: some intelligent block compression with the included code that'll do better than the "dumb" compression from "zip" or other algorithms.
You could combine that with compiler space optimization tricks, too: loop RE-rolling, for example. A lot of compilers do tricks to make code faster. Not many do things to explicitly make code smaller. A "shriking" compiler (or a disassembler/reassembler) that produced small code combined with an "expander" app that made the code bigger but faster could make very small apps.
It's a question of manpower and applications.
If you're running Big Expensive Software (Oracle or DB2, for example), you want to run a configuration that your Big Expensive Software vendor supports, and you don't want to upgrade any more than necessary. Same thing's true for in-house apps. Upgrades can take a lot of manpower (thus money). In cases like that, plunking down a few hundred (or even thousand) for a support contract and/or a configuration that's guaranteed to be supported for five years is a great deal.
On the other hand, if you're running "commodity" services and servers (i.e. the kind of stuff that's installed by default when you click the "server" box during a RH install), you probably have failover anyway, so there's no harm in running the cheap/free version of RH with some sort of internal RPM update service (AutoRPM, rpm-update, or the like) and upgrading all of those machines every year. On the flip side, it might still be cheaper for your organization to buy Enterprise or Workstation and put your "upgrade time" to use somewhere else.
The big case for the free version, IMO, is software development and desktop clients. If you're WRITING software for RH, your developers should be using the platform that they're developing for. Also, if you have a whole lot of desktop linux users, the per-seat price for the Workstation version can add up quickly.
"precious processing power"?
I've not encountered a machine in ages where the extra percentage of CPU or disk usage involved in journalling a normally read-only partition mattered much. Or, more accurately, when load matters, I either make the disk read-only (if things don't change) or journalling is even more important than usual (if things do change).
If your disks are 99% full and journalling overhead matters, the solution isn't "turn off journalling", it's "buy a bigger disk". Same with CPU. Sure, there are quick tricks like "noatime" for heavily-used filesystems, but most of the time, spending money on people who will optimize the system is more expensive than just buying more hardware and making a bigger system.
My question'd be "Why bother installing any FS that's not the default for your OS?" Unless the answer's "I'm doing something so specialized that the default FS won't work", you're probably doing something wrong.
As far as "well built" goes, I have one of the wobbly ones, too.
This is a two week old laptop, BTW.
If you argue "that's too complicated!", you probably don't want an iso image of a compile-from-source linux distro in the first place.
Ah, screw it. Don't use mine. Use the one at rit.edu that's already been modded to 5.
The md5 checks out, but I haven't yet burned it or tested it.
Ah, that explains it. I wondered why the hell anyone would drink a vodka martini. Gin and vermouth go together nicely (esp. with the tiny bit of olive flavor). Vodka and vermouth just tastes like really crappy vodka (Gilby's, anyone?)
I noticed one of the recent Bond movies finally stopped pretending that they were martinis. You just see Bond drinking vodka shots. Now if only he'd drink good vodka...
The problem with qmail is that the author is a screaming loony (albeit a very smart one).
We all know how support works for open source products: you use the mailing lists, IRC channels, mail to the author, etc. Woe be unto you if you ask a qmail question that was answered 3 years ago (and is thus archived somewhere), or worse yet, you ask for functionality that qmail doesn't have and DJB deems "inessential".
I've not even asked questions, but I've read the archives. A surlier, more user-hostile community would be hard to find.
If you're going to a private "geek" university, pick up a decent laptop (with wireless if your school supports it) and a pile of extra batteries (and an external charger if possible). Laptops are damned near fashion accessories at many private universities. Bonus style points for a Powerbook. Go ahead and use it to take notes: everyone else will, and the professors (in the tech classes, at least) will often be wearing microphones to be heard over the typing. Really.
If you're going to a public university, the rules are different. Spent the money on a good monitor and an okay PC. Got an awesome custom case? Leave it at home. See if you can ram your PC into a POS Packard-Bell case scavanged from the neighbor's trash, and throw a brick in the bottom for good measure. Cover the monitor (not the vents!) and PC case with stickers to crap it up. You want this thing to scream "too crappy to steal!" Use a composition book (the kind with stitched pages) for each class. No laptops: the clicking will piss off your classmates. No spiral-bound notebooks: you will crush the spiral wire and end up switching notebooks mid-semester.
Regardless of university: get a small, cheap PDA. Don't use it for notes. Use it as an alarm/calendar and carry it everywhere. Set an alarm for 15 minutes before each class and recitation, and an alarm for two hours before each homework assignment is due: you want to allow time for the "Oh, shit!" followed by an emergency trip to your dorm/apartment to pick it up or (if you completely forgot) time to put something quick and shoddy together so that you can get partial credit. For major projects, put a reminder in for a week before it's due so that you can ask questions, get missing materials, etc. from the professor. Oh, yeah: sync religiously. Since you'll be packing this thing everywhere, you're going to lose or destroy at least one of them (beer and PDAs are mortal enemies).
Get a printer. Speed doesn't matter, but if you have a roommate, noise does. Whenever you want to print, so does the rest of the campus. Always keep an extra ream of paper and extra ink on hand.
The BT FAQ addresses the issue with windows cards, but I saw the same problem on my cheapo NE2000-compatable ISA ethernet card. I replaced it with a PCI card, and the problems went away.
I'm a moron. This discussion happened before the story I mentioned was posted.
Guess I need to read slashdot more often...
Maybe you could convince them to post the above link as a story. They'd never notice.
You're right, but a bit of pedantry: The original Airport cards were based on the Lucent Wavelan II cards (and even had the same Lucent chips inside).
The original Lucent Wavelan cards (at least, the ones that I'm familiar with) were pre-Wi-Fi 900MHz and had a big deck-of-cards sized external antenna that you velcro-ed onto the top back of your laptop's screen.
Most installations that I've seen use wires to the APs (often with Power-Over-Ethernet, to reduce wiring to a single cable). That way, you have fewer channel collisions, less overall wireless traffic, and significantly lower latency. You can put all of your wires in hallways, basements, and other "public" areas where you can work without going into people's apartments. Too ugly? Use drop ceilings and ceiling-mounted cable runs. Not to mention that it's far harder for a repirman to mess up a cheap 10/100 switch than an elaborate antenna array.
Also, think hard about setting up some sort of monitoring system. You want to have someone on the way to fix a failing AP before the angry calls come in.
Or am I feeding the trolls again?
The speed is OK with the latest generation
One word of caution: DON'T get 802.11g stuff yet unless you have money to burn and want it now. Stick with 802.11b. A lot of companies (Apple and Linksys are the biggies) are selling stuff that's "based on the emerging 802.11g standard". Nobody knows yet whether that hardware will be compatable with the standard when it comes out.
Anyone who bought an X2 or K56Flex modem will be able to tell you about what "software upgradable" really means in the face of unknown standards. It means (if you're lucky) "the manufacturer will reluctantly exchange your product if you're willing to ship it to us and wait two months for us to send a different reconditioned unit with a new chip."
You can't afford $1 for a song. You have to use the money for food, instead.
.sig and home page refer to Transgaming, a place where you can pay $5 a month to get access to software that'll let you play $50 dollar Windows computer games. (or do you steal those, too?)
Meanwhile, your
And you "might need an ibook". Like someone who can barely afford food could come up with two grand for a new computer.
Sorry, man. You had a good troll going, but the bullshitometer just went off the scale on that one.
Do you know how to read?
Ah, nothing like a mature discussion. At least, you won't find anything like one around here.
I suggest that you go back and re-read the article that you're referring to and check out the license links in it. The person who buys your computers DOES get the software, including any upgrades that were made available under the support contract before the sale. They just don't get any MORE upgrades. In other words, the software IS transferrable. The support and upgrade contract isn't.
You'd be correct if there weren't multiple bugs in both the WEP protocol and in most common implementations. As it stands, most implementations (Cisco's is a notable exception) use the same keys for all traffic. That means that you can park your wireless notebook near an AP and deduce the keys in a few hours. If your wired office is next to mine, there's no way for me to sniff your traffic. If your wireless office is next to mine, I can crack your keys and see your traffic.
There are also fundamental key distribution problems in large organizations. I work at a university. We don't use WEP because we can't come up with a way to distribute the keys that doesn't also make it trivial for non-authorized users to use the network. We can use physical security to keep our wiring closets and network jacks safe. We can't put faraday cages around our buildings to keep the wireless network safe.
Essentially, Microsoft are saying "We're going to sell you a three-year non-transferrable support and upgrade contract." Nothing wrong with that. They also let you make payments instead of paying for the full three years up front. Again, nothing bad about that. They DO specify that you have to pay off the balance on any machines that you sell, though. How's that any different than, say, requiring you to pay off your bank loan before you sell your car?
It seems to be that nobody'd be complaining if they just required the entire payment for three years up-front and said"It's non-transferrable. Cope." People are pissed because Microsoft offers a payment plan, but they won't automatically transfer that plan.
I encourage people to put the redhat .isos onto other p2p networks (gnutella, etc.) as well as BT. Sure, BT's nifty, but it's good to get the data out as widely as possible. Not everyone can run BT.
After I've finished my download, I'll put it up on gnutella. My copy shouldn't be the only one.
Sounds like Debian, except for the "documented" part. ;)
I'm not sure that the argument is "It's hard" or "It can't be done." I don't use Gentoo because I don't want to lose use of my computer for the day or two that it can take to compile everything. Gaining 10% more speed isn't worth losing the use of my computer for a day every time a major update appears. With a slow computer, the compile time's a killer. With a fast machine, the extra efficiency is even less worthwhile.