Even better, Pierre-Phillipe Coupard of Lineo Inc. has written a driver for the CueCat that captures the escape sequences, decodes the data, and echoes it in human-readable format to/dev/cuecat (which you obviously have to create)
This is a good thing(tm) because the escape sequence is alt-F10, which makes it a real pain to try and use the cuecat from a linux console (unless you have a useable vt on tty9)
It's early code, so there's obviously work to be done, but I've tried it and it does work pretty well for v.0.0.1
Consider, 18 gig drives are on the small end of what companies want to build a serious storage array out of these days. And nobody builds a serious storage array out of UW drives anymore anyway. These days a storage array is going to be using large Ultra2 or Ultra160 drives.
At any rate, I see 18 gig 7200rpm UW scsi drives being liquidated for less than $200 all the time.
I don't know what the performance of the 3Ware cards is compared to SCSI. If it can overcome the general boneheadedness of the ATA interface that generally bogs down your entire system, it might be worth it. But they're still so expensive that it's probably not worth it.
Then you learn the nitty gritty. Like price. And capabilities.
The basic model has two channels. These cards only let you put one drive on each channel. So you've paid $200 for two drives. A two drive raid. whoop-de-effin-do.
The prices ramp up very quickly from there. Close to $400 for a four-drive controller. You're not saving a red cent vs. comparible scsi. You'd be better off with a couple Promise controllers and software raid.
UW scsi is comparable to UDMA66 in terms of throughput, even with multiple drives on the channel. UW scsi is considered old-crappy-junk by the RAID array set. Thus, UW scsi based raid controllers are selling quite cheap. I have personally purchased (for my employer) several AMI Megaraid 2 channel UW scsi raid controllers for $115 each. It's quite common to find them for less than $200.
Don't bother with this IDE raid malarky. It's pointless and silly. If you insist on RAIDing IDE drives, just use software raid, linux is quite good at it.
Do you know how many 30 gig drives you can buy for the price of a reliable scsi tape system and tapes?
Around a bakers dozen, at least compared to the last time i priced DLT drives.
Now, to head off the inevitable "What about (insert cheezy tape system here)?" - I've used OnStream, etc. They fail to meet the designations of either fast or reliable. Not that I'm a big fan of DLT either, but DLT is a lot better.
Whoever they are, they don't know how to spell Timpanogos, that alone is pretty embarrasing.
Maybe it's not their fault. See, it's the name of the tallest mountain around these parts. There are two trails to the peak, and the easiest one is the one maintained by the girl scouts of america. And since they're ignorant, they spelled the sign at the trail head Timpanogas. And since they're cheap, they never fixed the sign. But it's still embarrasing. It's spelled Timpanogos. Anyone who had to study it in elementary school knows that.
So, aparantly, these guys have never hiked it (and it's almost a requirement to hike it at least once if you live here), or they've only hiked it the easy way.
All that being said, there are facts to be pointed out.
1: They've been mentioned here before. At that time i think it was netware support for linux or somesuch.
2: Former Novell Execs are a dime a dozen. Honestly, being a Vice President at Novell is like being a middle-managment pencil pusher anywhere else. They have so many of them that the term no longer has any meaning.
3: If they can't spell the name of the valley's most prominant landmark correctly I've absolutly no faith in them.
Why in the hell would you want a router and MP3 player on the same system? Sorry the network is so slow today Boss, it must be that damm Seti @ Home screensaver thing on the print server and that stack of CDs that the router is ripping!
Ahh, earth to anonymous coward. We have one ethernet and one (soft) 56k modem in the box. Clearly, this isn't a business solution in the first place.
What's more, a 266mhz Cyrix MediaGX kicks the pants off the Mips cpu in, say, a Cisco 7206/VXR. And that's a $20,000 router.
I look at this like it's hobby junk. It's got to be lower power consumption than the DEC Multia currently driving my network, and faster too.
I get the impression you've never even attempted to build a router from scratch. You really don't need a lot of cpu horsepower. Most of the stuff from Cisco, etc, tops out at about 200mhz, and those things are running BGP. If you're just doing simple static stuff, 99% of the time, your cpu is going to be idle. Why waste the cycles on a non-critical system?
What i really wonder about where this box is concerned is if the bios can be convinced to boot the 4 megs of flash.
I mean, since it includes netscape and realplayer, it's probably Cyrix MediaGX or Geode based. So it's pretty plain jane stuff.
The "4 megs EEPROM" is probably some sort of linear disk-on-chip. I have access to programmers for those sorts of things at work, and 4 megs is plenty to boot a minimal (say, router & mp3) system.
I understand that sometimes stuff is gonna show up twice, but this is silly.
for the record, the foot pedals were their own article HERE so we're what, four out of 10 confirmed already posted, and a fifth that may have been?
You may call a technicality on the symphony. Personally, I think it's stupid to link to the knowledge base article. It's not a windows bug. It's not a bug. The hardware does it, it'd do the same thing if it overheated under Linux or BSD. I used to work for a shop years ago that had a Netware box that would play Fur Elise when it got too hot. It's a function of the hardware monitor on some motherboards.
I can believe ATi might getkicked out of the show for that, but the rest is a real stretch of the imagination.
Apple is selling a "Supercomputer" here with 64 megs of ram. Their inclusion of last year's most so-so 3d graphics contender hurts Apple more than anyone else. I mean, come on, even S3 had better 3d performance.
Furthermore, I've not seen the inside guts of this cube, but I would be very surprised if the video is on a PCI card. In keeping with Apple tradition it's probably soldered right to the mainboard.
And, re-design the mainboard of your flagship product mere weeks before the launch just so that you can use the crappier version of an offending vendors product? It just doesn't make sense. Not even if you're insane.
The programming interface used by the Mach64 through the Rage128 is so similar that I'm pretty certian the only reason they put the Rage128 in there is simplicity of driver support. ie, they were too lazy to switch to 3dfx or nVidia.
There are plenty of good, solid reasons why the tops of most computers are very featureless.
First, because folks like to put stuff on top of them. the big vent in the middle of the top of this thing will probably prevent people from resting their coffee mug on it, but I'm sure they'll have overheating problems from folks putting papers on top of it.
Second, the Crud Factor. That big pretty vent probably makes nice pretty ringing noises when you push coins through it. Hopefully they've thought to put some mesh behind it. I grew up in a big family and have many nieces and nephews, don't think your kids won't try this, I've seen coins (and worse, like green jello) done to this sort of box in the past.
And that CD-ROM drive. Eesh. It even *LOOKS* like it's designed to store paperclips. I don't even want to think about that.
And, finally, is it just me or does the whole affair look like a 1950's impression of a futuristic deep fat fryer? All we need is a wire handle attached to those two bolts on the top so that we can pull the basket out to drain . . .
So, would that mean that Mulder could retire from the FBI, allowing him to persue his desire to dress in women's clothing while he does freelance private investigator work?
The problem with antisniff tho is that it's really, really easy to tell when someone is running antisniff on your segment. Anyone who's paying attention *Will* know you're looking for a sniffer.
But that's beside the point. Most switches (and I've worked with everything from linksucks to 3com to smc to hp to cisco to foundry to extreme, and most inbetween too) don't give a rat's hind quarters if you're in promiscuous mode. I can't think of a recent switch that does. You can look at all the ARP broadcasts you like but they won't just start funneling the whole backplane to your port. Not unless you're doing something really evil to shut down the filter.
What you generally need, and I've set up security sniffers for large, flat networks, is what they call a monitoring port. A monitoring port is just a port that essentially gets cc:'d all the traffic going through one other port on the switch.
Now, most low-end managed switches, like 3Com (ugh, what cruft), support one monitoring port at a time. In this sort of situation, you need a topology where you're funneling all your data through a particular port, or you need many, many sniffers, because switching loops are bad juju. There are ways to set this up that don't suck very much, but they all go to crap when your utilization creeps past 40% or so.
Mid-range managed switches, like Cisco switches, generally support multiple monitoring ports. This makes it a lot easier on your overall network topology, but you need many sniffers, or many ports on your sniffer.
Of course, ALL of this presumes that your link is ethernet. 100mbps ethernet isn't a particularly fat pipe for the internal backbone of even a mid-size isp. ethernet isn't what you'd call an adaptive technology, it starts to suck when you're using only 1/3rd of it's capacity. Which quickly means that you end up buying big core routers, and having several separate ethernet segments. You start to have a topology that just doesn't lend itself to off the shelf sniffing hardware.
Yeah, there's gigabit ethernet. But in my network admin days, had a spook shown up and told me that he wanted me to dedicate a gbps port as a monitoring port for my whole pipe, I'd have told him that either he can show me a court order or warrant or he can cram his sniffer where the sun doesn't shine. Those ports are *Expensive*.
Other technologies used for high speed backbone links - fiberchannel, sonnet, etc, really aren't all that easy to sniff with off the shelf hardware.
What I'm betting is the fbi said "We have a consumer-grade ethernet port on our sniffer and it has to be able to see allll the traffic on your isp, so you have to funnel every last link on your whole network onto a wire that acheives 14 megabytes per second on paper but rarely in reality more than maybe half that, so that we can protect you from crime"
And earthlink probably put forth their best effort to implement it merely so that they could document how bad the idea is.
This is a good example of a coherent, interesting/. bost that's also complete bullshit. Moderators seem to like these.
This isn't a troll. Matts obviously hasn't bothered to read any of articles referenced above. If he had, he would have noticed that TUX is designed *Specifically* to be integrated into Apache.
Forgot to mention - *Could* also be that your system is a Micron or something, and the PS/2 ports are not PC98 compliant - that is, they don't support hot-re-plugging of mice.
Try just hot-re-plugging the mouse (pull the cable, wait a minute, plug it back in) without the KVM in series. If that doesn't work either, the problem is both that your motherboard isn't very smart and that your kvm is pretty braindead.
I wrote the original testing specs used by KeyLabs (the linux-tested.com people) last spring. I can attest to the fact that at least when i was in charge, this was indeed a meaningful set of tests. I handed the project off to another tech (a rabid linux user as well) in june of 99 and left it in good hands.
KeyLabs definately isn't the only shop offering testing. Linuxcare used to, maybe they still do. I don't work for KeyLabs anymore, and I don't know if James is still the project lead for the linux project. What I do know about KeyLabs is that the culture among the techs is that it is more fulfilling to *fail* a device than to go through days of boring testing and just say it passed.
I took *Pride* in being a major speedbump in marketing initiatives when something wasn't up to par. When you have to spend five hours pressing every key on a keyboard over and over with different modifiers, in a SPECIFIC SEQUENCE, it's the only joy possible!
All that being said - the issue here is probably that the mouse you're using recieves some initialization from the driver (to set up a wheel, or similar) and that the KVM is inadvertently causing the mouse to be reset when you switch between systems.
Belkin used to haven an issue like this with XWindows, there was aparantly a workaround until they fixed the hardware, for the life of me I don't remember what it was, but it's probably on belkin's support site.
Better yet you can just install a software shim between the player and your sound card (ie, a software-only audio "driver" that pretends to be a sound card, using your real sound card as the output device) and copy the decrypted data to disk.
It's a *simple* concept, and i keep wondering why people have such difficulty with it: If the media can be used at all, it's content can be duplicated.
*All* floating point units have bugs. Floating point math isn't exactly easy to implement on silicon at high speeds. That's why there's such bragging rights on being able to do it quickly.
Usually these bugs are simply rounding errors, simple things to work around. Look in the code of any compiler, you'll see plenty of workarounds.
Intel's major error wasn't in having an FDIV bug in their chip. Intel's first error was in marketing their processor as though it were a consumer device.
Intel's second error was in initially pretending there was no bug. The appropriate response would have been to rush the workaround code to every compiler vendor in existance and make sure the engineering and mathematical communities that really needed it could get it.
Intel's third error was in treating every customer like a dumb consumer and telling them they weren't nearly cool enough to be affected by such an obscure little bug, and to go back to playing tetris and forget about it.
THAT's how you turn a minor logic error into a major recall.
but what would stop a company from doing the same thing to SDRAM or whatever else RAMBUS is liscencing that tons of IBM clones did to IBM?
What most people are unaware of is that while IBM appeared to be getting "slaughtered" by clone makers, IBM was reaping billions of dollars in license fees for their patented technologies. Most of this was intentionally kept very hush-hush, the license included an NDA. You could not even admit that you had spoken with IBM regarding their patents.
Yes, the ISA bus came from an Intel periodical. Yes, the CGA graphics adapter came from a Motorola periodical. Yes, the trick of using the keyboard controller as an MMU was also public knowledge, but there were *Many* aspects of the original PC, XT, and AT, that were patented IBM technology.
People get the impression that IBM somehow got screwed. They made out like bandits, and enjoyed every minute of it.
There are kernel patches already available for both usb and suspend-to-disk, get with it man.
Of course, it would be nice if the power managment support got finer grained. The screen on my Portege turns off after 2 minutes of keyboard inactivity and there's nothing I can do about it. (believe me, I've tried).
First, i imagine the original design stopped at 2Mbps because it wasn't really feasable to go over that speed using the parts on hand.
Second, the statement about "not much operating at 2Mbps" as a reason for trying for 10Mbps makes it sound like he really doesn't understand the way asynchronous networks function.
At least with tcp, you simply transmit more and more (use larger and larger windows) until you start to experience transmission errors, or it just isn't getting you any better throughput, and then step one back from the last increase (well, you're supposed to) and call that your transmission rate.
Basically, the device on the other end ACK's every once in a while to let you know your packet got through. If the device on the other end didn't understand the data, or wasn't ready to recieve it, it can't ACK it. If you don't get your ACK in time, you retransmit. If you're having to retransmit too often, your transmission rate is obviously too high, so you back off your window size and slow down.
It's a cheezy form of bandwidth controll, but it works, sortof. It's the electronic equivalent of not reading a signifigant portion of your email in hopes that people will stop sending so much of it. Except that the protocol is designed such that a correctly written application *will* stop sending so much.
Yes, speed is nice. but relatively low-speed long haul links are the way the world goes round at this point. "T1" is only a 1.4Mbps link and it's more than enough for most corporate internet connections, even if the internal lan is 100Mbps.
The trick is, you have to remember that long haul links, no matter how fast, make really crappy bridges between homogenous networks. The segment length limitations are there for a reason. The speed of light is not just a plot complication in a larry niven novel.
The best way to handle a long haul link is to put a router on each end and let the routers deal with the general bullshit inherent in a really distant bridge.
Better yet, employ some form of traffic shaping (The buzzword is "QOS" but most people don't actually need or want guaranteed quality of service) in order to make sure that the pipe gets used to it's fullest capability rather than let applications choose their own window size and thus often cause the network to behave in a "bursty" manner, where a rude application or server occasionally kicks all other connections off the link for a few miliseconds so it can send it's data in first-class, and then let all the polite apps try and figure out if they can use a large window again.
The traffic controller code in recent 2.2 kernels is great for that kind of thing, btw.
Of course, as other people have pointed out, the legality of a 2Mbps HAM link, let alone 10Mbps, is quite suspect. Someone with access to recent FCC rulings should comment on it.
A licensed amateur may design and build their own device to transmit legally in the HAM bands, and it seems from the rules most people know about that the 10ghz HAM band is limited to 56kbps.
Consumer devices can use unlicensed frequencies such as the 900mhz range used by cordless phones and cellular devices, and the 2.4ghz range that's starting to be used by cordless and cellular devices, but this requires that the design of the device be FCC approved, not an easy thing for joe solderingiron to acheive.
It's posible to use a licensed frequency for high speed data transmission of just about anything you like, but this also requires an FCC approved design.
Is there anyone here who's recently researched FCC rules regarding packet data rates?
Slashdot Mirror
Even better, Pierre-Phillipe Coupard of Lineo Inc. has written a driver for the CueCat that captures the escape sequences, decodes the data, and echoes it in human-readable format to /dev/cuecat (which you obviously have to create)
This is a good thing(tm) because the escape sequence is alt-F10, which makes it a real pain to try and use the cuecat from a linux console (unless you have a useable vt on tty9)
It's early code, so there's obviously work to be done, but I've tried it and it does work pretty well for v.0.0.1
See the freshmeat appindex record at http://freshmeat.net/projects/cuecat/
Actually, prices are coming down quite a bit.
Consider, 18 gig drives are on the small end of what companies want to build a serious storage array out of these days. And nobody builds a serious storage array out of UW drives anymore anyway. These days a storage array is going to be using large Ultra2 or Ultra160 drives.
At any rate, I see 18 gig 7200rpm UW scsi drives being liquidated for less than $200 all the time.
I don't know what the performance of the 3Ware cards is compared to SCSI. If it can overcome the general boneheadedness of the ATA interface that generally bogs down your entire system, it might be worth it. But they're still so expensive that it's probably not worth it.
They look like a real neat idea from the outset.
Then you learn the nitty gritty. Like price. And capabilities.
The basic model has two channels. These cards only let you put one drive on each channel. So you've paid $200 for two drives. A two drive raid. whoop-de-effin-do.
The prices ramp up very quickly from there. Close to $400 for a four-drive controller. You're not saving a red cent vs. comparible scsi. You'd be better off with a couple Promise controllers and software raid.
UW scsi is comparable to UDMA66 in terms of throughput, even with multiple drives on the channel. UW scsi is considered old-crappy-junk by the RAID array set. Thus, UW scsi based raid controllers are selling quite cheap. I have personally purchased (for my employer) several AMI Megaraid 2 channel UW scsi raid controllers for $115 each. It's quite common to find them for less than $200.
Don't bother with this IDE raid malarky. It's pointless and silly. If you insist on RAIDing IDE drives, just use software raid, linux is quite good at it.
Do you know how many 30 gig drives you can buy for the price of a reliable scsi tape system and tapes?
Around a bakers dozen, at least compared to the last time i priced DLT drives.
Now, to head off the inevitable "What about (insert cheezy tape system here)?" - I've used OnStream, etc. They fail to meet the designations of either fast or reliable. Not that I'm a big fan of DLT either, but DLT is a lot better.
- Eric
I agree completely. A cheezy hack that's held on far too long.
Unfortunately, really big scsi drives are more expensive than really big ide drives.
If i wanted 30 gigs for mp3 storage or something similarly trivial, i'm afraid it'd be IDE in a heartbeat.
keep in mind, my workstation at home has five (5) discrete scsi busses, and no ide drives. But I've got less than 20 gigs of scsi storage total.
At least, with storage arrays getting bigger and bigger, the old UW stuff is hitting the market pretty cheap occasionally.
"Jeff V. Merkey" wrote: > > Actually, > > > > You're wrong. > > > > There are four spellings used for timp: > > > >
Fine. But the one you picked still sounds stupid.
And fsck in the kernel is a lousy idea that shows way too much microsoft mentality.
Whoever they are, they don't know how to spell Timpanogos, that alone is pretty embarrasing.
Maybe it's not their fault. See, it's the name of the tallest mountain around these parts. There are two trails to the peak, and the easiest one is the one maintained by the girl scouts of america. And since they're ignorant, they spelled the sign at the trail head Timpanogas. And since they're cheap, they never fixed the sign. But it's still embarrasing. It's spelled Timpanogos. Anyone who had to study it in elementary school knows that.
So, aparantly, these guys have never hiked it (and it's almost a requirement to hike it at least once if you live here), or they've only hiked it the easy way.
All that being said, there are facts to be pointed out.
1: They've been mentioned here before. At that time i think it was netware support for linux or somesuch.
2: Former Novell Execs are a dime a dozen. Honestly, being a Vice President at Novell is like being a middle-managment pencil pusher anywhere else. They have so many of them that the term no longer has any meaning.
3: If they can't spell the name of the valley's most prominant landmark correctly I've absolutly no faith in them.
Why in the hell would you want a router and MP3 player on the same system? Sorry the network is so slow today Boss, it must be that damm Seti @ Home screensaver thing on the print server and that stack of CDs that the router is ripping!
Ahh, earth to anonymous coward. We have one ethernet and one (soft) 56k modem in the box. Clearly, this isn't a business solution in the first place.
What's more, a 266mhz Cyrix MediaGX kicks the pants off the Mips cpu in, say, a Cisco 7206/VXR. And that's a $20,000 router.
I look at this like it's hobby junk. It's got to be lower power consumption than the DEC Multia currently driving my network, and faster too.
I get the impression you've never even attempted to build a router from scratch. You really don't need a lot of cpu horsepower. Most of the stuff from Cisco, etc, tops out at about 200mhz, and those things are running BGP. If you're just doing simple static stuff, 99% of the time, your cpu is going to be idle. Why waste the cycles on a non-critical system?
What i really wonder about where this box is concerned is if the bios can be convinced to boot the 4 megs of flash.
I mean, since it includes netscape and realplayer, it's probably Cyrix MediaGX or Geode based. So it's pretty plain jane stuff.
The "4 megs EEPROM" is probably some sort of linear disk-on-chip. I have access to programmers for those sorts of things at work, and 4 megs is plenty to boot a minimal (say, router & mp3) system.
Does anybody have info on the guts and the bios?
Sure, occasionally stuff gets re-posted, but this is absurd. You missed a few.
The x86 Still was a quickie not quite a month back
RSA implemented in javascript was a quickie just a few weeks before the still
I understand that sometimes stuff is gonna show up twice, but this is silly.
for the record, the foot pedals were their own article HERE so we're what, four out of 10 confirmed already posted, and a fifth that may have been?
You may call a technicality on the symphony. Personally, I think it's stupid to link to the knowledge base article. It's not a windows bug. It's not a bug. The hardware does it, it'd do the same thing if it overheated under Linux or BSD. I used to work for a shop years ago that had a Netware box that would play Fur Elise when it got too hot. It's a function of the hardware monitor on some motherboards.
MOOO!
if you don't know what a tgz file is, where do you get off talking about internet standards?
I can believe ATi might getkicked out of the show for that, but the rest is a real stretch of the imagination.
Apple is selling a "Supercomputer" here with 64 megs of ram. Their inclusion of last year's most so-so 3d graphics contender hurts Apple more than anyone else. I mean, come on, even S3 had better 3d performance.
Furthermore, I've not seen the inside guts of this cube, but I would be very surprised if the video is on a PCI card. In keeping with Apple tradition it's probably soldered right to the mainboard.
And, re-design the mainboard of your flagship product mere weeks before the launch just so that you can use the crappier version of an offending vendors product? It just doesn't make sense. Not even if you're insane.
The programming interface used by the Mach64 through the Rage128 is so similar that I'm pretty certian the only reason they put the Rage128 in there is simplicity of driver support. ie, they were too lazy to switch to 3dfx or nVidia.
There are plenty of good, solid reasons why the tops of most computers are very featureless.
First, because folks like to put stuff on top of them. the big vent in the middle of the top of this thing will probably prevent people from resting their coffee mug on it, but I'm sure they'll have overheating problems from folks putting papers on top of it.
Second, the Crud Factor. That big pretty vent probably makes nice pretty ringing noises when you push coins through it. Hopefully they've thought to put some mesh behind it. I grew up in a big family and have many nieces and nephews, don't think your kids won't try this, I've seen coins (and worse, like green jello) done to this sort of box in the past.
And that CD-ROM drive. Eesh. It even *LOOKS* like it's designed to store paperclips. I don't even want to think about that.
And, finally, is it just me or does the whole affair look like a 1950's impression of a futuristic deep fat fryer? All we need is a wire handle attached to those two bolts on the top so that we can pull the basket out to drain . . .
So, would that mean that Mulder could retire from the FBI, allowing him to persue his desire to dress in women's clothing while he does freelance private investigator work?
The problem with antisniff tho is that it's really, really easy to tell when someone is running antisniff on your segment. Anyone who's paying attention *Will* know you're looking for a sniffer.
But that's beside the point. Most switches (and I've worked with everything from linksucks to 3com to smc to hp to cisco to foundry to extreme, and most inbetween too) don't give a rat's hind quarters if you're in promiscuous mode. I can't think of a recent switch that does. You can look at all the ARP broadcasts you like but they won't just start funneling the whole backplane to your port. Not unless you're doing something really evil to shut down the filter.
What you generally need, and I've set up security sniffers for large, flat networks, is what they call a monitoring port. A monitoring port is just a port that essentially gets cc:'d all the traffic going through one other port on the switch.
Now, most low-end managed switches, like 3Com (ugh, what cruft), support one monitoring port at a time. In this sort of situation, you need a topology where you're funneling all your data through a particular port, or you need many, many sniffers, because switching loops are bad juju. There are ways to set this up that don't suck very much, but they all go to crap when your utilization creeps past 40% or so.
Mid-range managed switches, like Cisco switches, generally support multiple monitoring ports. This makes it a lot easier on your overall network topology, but you need many sniffers, or many ports on your sniffer.
Of course, ALL of this presumes that your link is ethernet. 100mbps ethernet isn't a particularly fat pipe for the internal backbone of even a mid-size isp. ethernet isn't what you'd call an adaptive technology, it starts to suck when you're using only 1/3rd of it's capacity. Which quickly means that you end up buying big core routers, and having several separate ethernet segments. You start to have a topology that just doesn't lend itself to off the shelf sniffing hardware.
Yeah, there's gigabit ethernet. But in my network admin days, had a spook shown up and told me that he wanted me to dedicate a gbps port as a monitoring port for my whole pipe, I'd have told him that either he can show me a court order or warrant or he can cram his sniffer where the sun doesn't shine. Those ports are *Expensive*.
Other technologies used for high speed backbone links - fiberchannel, sonnet, etc, really aren't all that easy to sniff with off the shelf hardware.
What I'm betting is the fbi said "We have a consumer-grade ethernet port on our sniffer and it has to be able to see allll the traffic on your isp, so you have to funnel every last link on your whole network onto a wire that acheives 14 megabytes per second on paper but rarely in reality more than maybe half that, so that we can protect you from crime"
And earthlink probably put forth their best effort to implement it merely so that they could document how bad the idea is.
Don't feed the trolls!
This is a good example of a coherent, interesting /. bost that's also complete bullshit. Moderators seem to like these.
This isn't a troll. Matts obviously hasn't bothered to read any of articles referenced above. If he had, he would have noticed that TUX is designed *Specifically* to be integrated into Apache.
Forgot to mention - *Could* also be that your system is a Micron or something, and the PS/2 ports are not PC98 compliant - that is, they don't support hot-re-plugging of mice.
Try just hot-re-plugging the mouse (pull the cable, wait a minute, plug it back in) without the KVM in series. If that doesn't work either, the problem is both that your motherboard isn't very smart and that your kvm is pretty braindead.
I wrote the original testing specs used by KeyLabs (the linux-tested.com people) last spring. I can attest to the fact that at least when i was in charge, this was indeed a meaningful set of tests. I handed the project off to another tech (a rabid linux user as well) in june of 99 and left it in good hands.
KeyLabs definately isn't the only shop offering testing. Linuxcare used to, maybe they still do. I don't work for KeyLabs anymore, and I don't know if James is still the project lead for the linux project. What I do know about KeyLabs is that the culture among the techs is that it is more fulfilling to *fail* a device than to go through days of boring testing and just say it passed.
I took *Pride* in being a major speedbump in marketing initiatives when something wasn't up to par. When you have to spend five hours pressing every key on a keyboard over and over with different modifiers, in a SPECIFIC SEQUENCE, it's the only joy possible!
All that being said - the issue here is probably that the mouse you're using recieves some initialization from the driver (to set up a wheel, or similar) and that the KVM is inadvertently causing the mouse to be reset when you switch between systems.
Belkin used to haven an issue like this with XWindows, there was aparantly a workaround until they fixed the hardware, for the life of me I don't remember what it was, but it's probably on belkin's support site.
Better yet you can just install a software shim between the player and your sound card (ie, a software-only audio "driver" that pretends to be a sound card, using your real sound card as the output device) and copy the decrypted data to disk.
It's a *simple* concept, and i keep wondering why people have such difficulty with it: If the media can be used at all, it's content can be duplicated.
*All* floating point units have bugs. Floating point math isn't exactly easy to implement on silicon at high speeds. That's why there's such bragging rights on being able to do it quickly.
Usually these bugs are simply rounding errors, simple things to work around. Look in the code of any compiler, you'll see plenty of workarounds.
Intel's major error wasn't in having an FDIV bug in their chip. Intel's first error was in marketing their processor as though it were a consumer device.
Intel's second error was in initially pretending there was no bug. The appropriate response would have been to rush the workaround code to every compiler vendor in existance and make sure the engineering and mathematical communities that really needed it could get it.
Intel's third error was in treating every customer like a dumb consumer and telling them they weren't nearly cool enough to be affected by such an obscure little bug, and to go back to playing tetris and forget about it.
THAT's how you turn a minor logic error into a major recall.
but what would stop a company from doing the same thing to SDRAM or whatever else RAMBUS is liscencing that tons of IBM clones did to IBM?
What most people are unaware of is that while IBM appeared to be getting "slaughtered" by clone makers, IBM was reaping billions of dollars in license fees for their patented technologies. Most of this was intentionally kept very hush-hush, the license included an NDA. You could not even admit that you had spoken with IBM regarding their patents.
Yes, the ISA bus came from an Intel periodical. Yes, the CGA graphics adapter came from a Motorola periodical. Yes, the trick of using the keyboard controller as an MMU was also public knowledge, but there were *Many* aspects of the original PC, XT, and AT, that were patented IBM technology.
People get the impression that IBM somehow got screwed. They made out like bandits, and enjoyed every minute of it.
There are kernel patches already available for both usb and suspend-to-disk, get with it man.
Of course, it would be nice if the power managment support got finer grained. The screen on my Portege turns off after 2 minutes of keyboard inactivity and there's nothing I can do about it. (believe me, I've tried).
First, i imagine the original design stopped at 2Mbps because it wasn't really feasable to go over that speed using the parts on hand.
Second, the statement about "not much operating at 2Mbps" as a reason for trying for 10Mbps makes it sound like he really doesn't understand the way asynchronous networks function.
At least with tcp, you simply transmit more and more (use larger and larger windows) until you start to experience transmission errors, or it just isn't getting you any better throughput, and then step one back from the last increase (well, you're supposed to) and call that your transmission rate.
Basically, the device on the other end ACK's every once in a while to let you know your packet got through. If the device on the other end didn't understand the data, or wasn't ready to recieve it, it can't ACK it. If you don't get your ACK in time, you retransmit. If you're having to retransmit too often, your transmission rate is obviously too high, so you back off your window size and slow down.
It's a cheezy form of bandwidth controll, but it works, sortof. It's the electronic equivalent of not reading a signifigant portion of your email in hopes that people will stop sending so much of it. Except that the protocol is designed such that a correctly written application *will* stop sending so much.
Yes, speed is nice. but relatively low-speed long haul links are the way the world goes round at this point. "T1" is only a 1.4Mbps link and it's more than enough for most corporate internet connections, even if the internal lan is 100Mbps.
The trick is, you have to remember that long haul links, no matter how fast, make really crappy bridges between homogenous networks. The segment length limitations are there for a reason. The speed of light is not just a plot complication in a larry niven novel.
The best way to handle a long haul link is to put a router on each end and let the routers deal with the general bullshit inherent in a really distant bridge.
Better yet, employ some form of traffic shaping (The buzzword is "QOS" but most people don't actually need or want guaranteed quality of service) in order to make sure that the pipe gets used to it's fullest capability rather than let applications choose their own window size and thus often cause the network to behave in a "bursty" manner, where a rude application or server occasionally kicks all other connections off the link for a few miliseconds so it can send it's data in first-class, and then let all the polite apps try and figure out if they can use a large window again.
The traffic controller code in recent 2.2 kernels is great for that kind of thing, btw.
Of course, as other people have pointed out, the legality of a 2Mbps HAM link, let alone 10Mbps, is quite suspect. Someone with access to recent FCC rulings should comment on it.
A licensed amateur may design and build their own device to transmit legally in the HAM bands, and it seems from the rules most people know about that the 10ghz HAM band is limited to 56kbps.
Consumer devices can use unlicensed frequencies such as the 900mhz range used by cordless phones and cellular devices, and the 2.4ghz range that's starting to be used by cordless and cellular devices, but this requires that the design of the device be FCC approved, not an easy thing for joe solderingiron to acheive.
It's posible to use a licensed frequency for high speed data transmission of just about anything you like, but this also requires an FCC approved design.
Is there anyone here who's recently researched FCC rules regarding packet data rates?
Transmeta's answered that question several times.
The thing is, the current breed of chips from transmeta are designed, that is, the hardware is designed, to be able to emulate x86.
Not to say they couldn't come up with a new one that can pretend to be PPC as well, but it sounds like an awful lot of work for such a small market.