First, make sure your case has positive pressure with respect to the outside. All your filtered fans should face in, such that any leakage is in the out direction. This is very important because otherwise, your removable media drives will act as filters, trapping the dust from the air that gets sucked in through them.
In my case, the power supply fan is the only exhaust fan. All the other fans in the case face in, and all are equipped with filters. They flow less because of the filters, but they keep up with the power supply fan, so the case is ever so slightly pressurized.
Secondly, hardware stores abound with cheap filter material. There's this great open-cell foam filter that you're supposed to wrap around your air conditioner's condenser coil to keep cottonwood fluff out. I usually just blast the thing out with the garden hose... but the filter medium is ideal for trapping large dust, and since it's intended to wrap an entire condenser, you get about 20 square feet for just a couple bucks.
There are also these little 4"x12" filters designed to be slipped behind your HVAC system's registers. Don't do that -- your forced air furnace was designed for a particular flow rate, and impeding it will cause problems. The filters, on the other hand, are just what your computer needs. They're denser material, and they work well as a second stage filter, behind the foam described above.
Ironically, about the only material at the hardware store that doesn't work very well as a computer air filter, is furnace air filter media. It's too thick for convenient mounting, and it's hard to work with.
Thirdly, the case you pick is important. Try a large tower "server" case, they usually have plenty of room for mounting extra fans, and they're better designed with respect to airflow. Filters will drastically reduce the effectiveness of each fan, so plan for at least double your usual number of fans. (Invest in some quiet ones with the fluid bearings.)
A friend recently gave me an unused computer that just happened to be built in the world's coolest case. The little fan mounting trays have plenty of space to tuck filter material inside, and everything just clips into place, no tools needed. It's a SupermicrO tower. I've been inside a lot of computer cases, and this one is by far the best-designed I've ever seen. (I have no relationship to Supermicro, I'm just impressed by their product.)
I'm going to echo the sentiment of another poster who said to elevate the machine. Get it off the floor, although I don't see a problem with setting it on a table. Just make double-sure that it's not sucking air in through openings near the bottom. I've seen lots of office desktops turn into little stationary vacuum cleaners, neatly inhaling every shred of dust that falls near them.
You're assuming that just because something operates at 2.4GHz, it must be 802.11b, and you're quite wrong.
The FCC regulates who can use what bits of the radio spectrum. Most pieces are allocated for things like radio broadcast, millitary use, amateur radio, GPS, etc.. There are a couple chunks, known as "part 15" bands, which are outlined in the rules for anyone to use, with a minimum of restrictions.
There's a band in the 900MHz range, and a band in 2.4GHz, and they're both called ISM, which stands for Industrial, Scientific, and Medical. I don't know why the words Consumer Putz didn't make it in there, because that's mostly what transmits there.
Cordless phones, wireless computer networking, certain video cameras, baby monitors, wireless microphones, you name it, all sorts of devices have been designed to operate in the 2.4GHz ISM band. It's handy! These new TVs work just like the X10 wireless video cameras, they use part of the spectrum to send video. They'll probably stomp all over an 802.11b LAN if you have one set up.
The world does not run on packets. There are things on the air, and things on wires, which are not Ethernet. This is my new mantra.
Remember those ads when the Zip came out, that featured a stack of seventysome floppies, towering and teetering? "You can store ALL THIS on one Zip disk!", they screamed.
Just recently I was looking at 120GB hard drives and thinking, shit, that would be one huge pile of floppies. Nevermind how long it would take to read them all!
Then I said, wait, what if each one were in a drive? Figure you can read/write an entire floppy in about a minute or two. The maximum sustained transfer rate on a hard drive isn't that great. Given a hundred thousand floppy drives and controllers for them all, you could beat the fastest hard drives with ease. Just imagine the noise and heat generation, though.
Let's hear it for stupid applications of obsolete technology! (next week's installment: dot matrix printers feeding sheetfed scanners as a backup medium / high-latency network interface.)
Unrelated to, and years earlier than, LS120. The floptical uses a floppy-like head mechanism, except it packs the tracks much closer together. It does this by using an optical pickup to align the head, rather than the dumb stepping used by regular floppy designs.
Does anyone know where these folks got the idea for this page? I could swear they got their inspiration from FreeSky Wireless, who made their first public announcement at Rubi-Con on Saturday.
It's obvious that both groups are very popular, since FreeSky also uses the same SSID, and all these folks must be members of both projects. It's great to see such a community-minded spirit across America!
This is an amazing kludge! They might as well have runners with CD-R's of the data, jogging up to the car as it passes by the station. Ugh. Cellular data is too slow, and wireless LANs are just too short-range. These folks would do better with a real wireless WAN system.
That's exactly the niche where Ricochet fit: High speed data, anywhere in a metropolitan area. While stationary, Ricochet speeds routinely exceeded 128kbps, and while moving at up to 60mph, the acquisition of new nodes pushed the throughput down, though not much. It would pick back up as soon as you stayed within range of one microcell for more than just a moment. (Software tweaks could improve mobile performance too.)
Imagine the applications of an ISDN-speed link in a moving patrol car: Realtime video from the dashboard camera. Instant images from headquarters displayed in the cruiser.
It's unfortunate that the present owners don't seem interested in reactivating the network in most of the old service areas (only Denver and San Fran are up so far), and they're certainly not planning to deploy equipment in any new cities. The LAPD would've loved Ricochet.
Ever since I heard about Digita, the OS for digital cameras, I wanted to program something like this...
When the prosecutor subpoena'dphotos from reporters, to identify people involved in the MSU Riots, the reporters obviously cried foul: Their independence is critical to the accurate gathering of news. If people see reporters as an arm of the law, coverage becomes skewed.
What if the camera could be loaded with a public key, and encrypt photos as they're taken, so that only the private key (back at the Editor's desk) could decrypt them? Perhaps keep a plaintext thumbnail in RAM for convenience, but make sure it vanishes if the card is removed or the camera's powered off.
It still doesn't prevent the judge from throwing the editor in jail for failing to turn over the private keys, but it adds a layer of complication, where the editor could simply lose the keys. Would that be destroying evidence? The encrypted photos still exist -- they're just unreadable.
Have you looked into the Terapin Mine? It sounds like exactly what you're looking for:
USB master, USB slave, PCMCIA slot, 10baseT, and it's small! Replace the rotating-platters laptop drive with a solid state version if you're really worried about vibration.
No shit! The original fit on four 1.2 meg floppies. The source to the PC version was lost, however. What's up on SF now is based on the 3DO version, which apparently had speech. (I'm unwilling to download 140 meg over dialup to find out.)
Odd tidbit: The.MOD files from the original version are still on my drive and I listen to them regularly.. They've been shuffled around as I've upgraded hard drives over the years, but the original file creation dates in 1994 remain in the directory. I think at some point I put them all in a.zip which would've preserved the file dates.
I'm pretty excited about this, I'd love to see a network-enabled version of super-Melee. Anyone with the requisite skill wanna try it? (I once tried to fix gorilla.bas to play across a lan. Since then I've left such things to professionals.)
Funny, I have a bad connector on my keyboard.. it's an IBM M-series (clicky monster), the type with the removable cord. Whenever I set the keyboard down sharply, it must jiggle the power pin or something, and it doesn't work until I reseat the connector.
Of course this causes the keyboard to reinitialize, which hoses up my typematic rate, so I have a hotkey (Ctrl-Alt-X) which pops up the Keyboard control panel. I just change a setting then change it back, then hit OK. Windows commits the "changes" by writing them to the keyboard, and my repeat rate is back to normal.
I've done this hundreds of times now, it happens once or twice a day and I've been using this keyboard at least since '99. Nothing's fried yet. You'd think I'd have the common sense to open the thing and resolder the connector.
Y'know what DOES cook motherboards? Remeber when cases had keyboard locks on them? The two-wire lead from the keyboard lock was installed backwards, so the metal body of the lock wasn't connected to ground, but to the keyboard lock sense lead. One dry winter day, I touched the lock and drew a large spark. The machine locked up and by the time I reached for the power button, I was smelling smoke.
Opening the machine revealed that the keyboard BIOS EPROM had gotten quite hot, enough to wrinkle and discolor the shiny sticker placed over the erase window. Just for kicks, we plugged the machine back in and powered it up to see what would happen.
Much to our surprise, the machine booted normally! The keyboard worked and everything, and that system remained in service for another 2 or 3 years before bloatware forced a motherboard upgrade. During the upgrade, we got a good chuckle out of the memory of the melted sticker.
The FBI could quite easily sneak a wireless AP onto the library's intranet, place the AP into bridging or sniffing mode (see the OpenAP project for ideas), and watch database lookups from their van parked at the coffee shop down the street.
That's beside the point anyway. Patrons frequently use the library terminals for things they can't find in the dead-tree collection. This does require an internet connection.
Here's an idea: Many libraries are on reasonably fast connections. Someone needs to create a small ram-based Linux distribution in a DOS executable. Something like Loadlin packed with a kernel image and filesystem image.
Then place this distro on a server with some carefully-scripted web pages which exploit flaws in IE to run the package. This is because the library probably runs Windows security software that prevents the user from easily downloading and running arbitrary programs.
The library patron walks up to the computer, and punches in the URL of this distro. The thing downloads, IE is tricked into running it, and the library's computer is now running a real OS from RAM.
Build in whatever freedom and paranoia you want: Perhaps the kernel shuffles things around in RAM every few seconds to prevent memory cell persistence. Perhaps the browsing is done via a secure tunnel to prevent the feds from simply Carnivoring the library's pipe. Perhaps this system meets other similar systems (gnutella-style?) on the net and they proxy each other's data (Freedom.net-style) to make things hard to trace. Perhaps the screen font is Tempest-annoying.
When the patron is finished, the distro wipes memory and reboots, back into the library's Windows for the next person to come along.
Of course such a distro would also be used by lame crackers with nothing better to do than attack websites from the library. *sigh*
The hard drive is but a small chunk of a computer. I don't mind if the drive is sealed and the firmware proprietary, as long as it conforms to a well-documented standard interface, and doesn't require any proprietary tweaks to perform well.
Likewise, I don't care what shape my pistons are ground in, or how the valvetrain actually works. Overhead or pushrod, seal it up so I never see it, I don't care.
However, my computer as a whole had better not be welded shut! I want to be able to inspect its operation, make sure I'm the user and it's the tool, not the other way around. I want to be able to control its operation, write my own programs, and tweak the misfeatures of others' programs. I'll never need to manually adjust the flying gap on my hard drive, but I definitely do want to customize my boot sequence, and scrutinize data leaving over my net connection.
Likewise, my car should be open enough that I can change the parts I want to. I'm happy with the engine, but why can't I turn off my headlights when I pull into my friend's driveway, so as not to wake his neighbors? I want wiring diagrams, so I can make the lighter socket behave more sensibly. I don't care if certain parts are sealed for practical reasons, but the rest of it should be open for me, the owner, to customize.
I think your analoy is flawed because you compare the hard drive to a car. The drive can be a black-box with no ill effect. It's the computer system as a whole that must be open to the user. The car can contain several black-box components, but I should be free to inspect and reconfigure the connections between them.
If you think this tendency in cars is scary, go read this article on Microsoft's Palladium plan. Talk about welded shut! At least your car doesn't refuse to carry passengers and cargo that haven't been approved by the manufacturer.
I'm aware that a good reference has multiple inputs, I'm simply saying that there isn't a good reference at most offices.
The CO's I've been in have a Telecom Solutions (by Symmetricom) DCD-LPR with GPS GTI cards feeding a DCD-ST2 with Stratum-2 oscillators, which drives the TOxA cards to feed the BITS-clocked network elements in the office.
In such a situation, if the GTI boards lose lock, the ST2 shelf goes into holdover, where it should be good for a few days. (I don't have the specs in front of me.) Equipment still has timing, it's just not locked to anything in particular. The switch and stuff will continue to run, but interoffice links will suffer as slip increases.
I'm sure all the major carriers had a Cesium reference in an office at one time, but nowadays I don't think that's used for anything. It's simply too awkward to push that signal out to each office. The GPS constellation is considered the primary reference.
The phone system won't go down completely, but it will break up into islands until a terrestrial sync distribution system can be established, or GPS can be restored.
Sure, the GPS satellites could be taken out by a rogue nation with too much laser power on their hands. The orbit data are public, after all. It's not a direct military strike, just a nasty thing to do, with repercussions that wouldn't be realized until after the fact.
The entire infrastructure is carried on SONET equipment. (That's Synchronous Optical Network, and if you didn't know that, you should read up on it, it's neat stuff.) Being synchronous, this stuff royally shits if there's something wrong with the timing.
Way back when T-carrier was first deployed, Bell realized this and set up a nationwide synchronization distribution. I think the master clock was in Kansas City. Anyway, the sync signal was distributed over wireline circuits to every central office in the country. Maybe Canada too?
However, most interoffice links are fiber now, the same SONET rings that depend on such precise synchronization. Ring-timing is awkward, and without very careful planning, sync loops can form. (Long story, look it up. The short version is that when a SONET system loses sync, it doesn't carry traffic.)
The modern concept is called BITS, or Building Integrated Timing Supply. Each office has a sync signal source, driven by an LPR (local primary reference) oscillator, which is in turn frequency-locked to a reference signal derived from GPS satellite signals.
Yes, that's right, the whole telephone network will fall apart if the Global Positioning System stops transmitting. Depending on the stratum class of the LPR, it might be able to "hold over" for a couple days, maintaining an accurate timing signal in the absence of an upstream reference. They will eventually drift, and most offices only have stratum-3 units anyway.
The network is so poorly planned in the first place, most transport engineers haven't got a clue about ring timing and such. They just hook each terminal to the BITS clock and hope it works, which it does, until something happens to the BITS clock. If all the BITSes in the network started drifting from one another, the system would slowly fail over a few days, as timing slips exceeded the tolerances of the various systems.
If such a thing were to happen, don't bet on the ability to patch things up quickly. Recordkeeping is horrible, and even if it weren't, it would be a daunting task to spontaneously set up a new sync distribution network independent of GPS.
I've heard on good authority that you wouldn't even need to take out the satellites themselves. A couple properly placed nuclear detinations could screw up the somethingsphere such that GPS signal propagation would suffer. Any physicists care to clarify?
Point 1: When a telco person says "switch", it means something totally different than what a data person means when they say "switch". This is a persistent annoyance.
You can't simply packet an ESS out of existence, because it doesn't know what a packet is. It's not connected to the internet. There are SS7 signaling links and X.25 control links, and maybe a few IP control links if you're lucky. None of them are connected to the internet. Your phone line is payload, not control.
Exactly how do you propose to access the switch in order to DoS it? There are switch dialins, but most are pretty secure, and good luck finding them. You're planning to do a lot of wardialing first?
Point 2: Telcos lie about bandwidth. When someone says they have a 10 Gigabit backbone, it means they own a couple OC192 circuits. Most of the channels in those circuits are probably not filled.
That's like saying I can move a thousand shipping containers a day, because there's a large river between me and my destination, and seaports at each end. Nevermind that I don't own any ships!
An OC192 circuit, for instance, can carry four OC48 signals, or 16 OC12 signals, or a mix thereof. Anything that adds up to 192 STS-1 payload envelopes, or equivalent concatenated payloads. You get the idea. Chances are, they're carrying one or two OC48s on the thing, and the rest is for future expansion. Each of those OC48s in turn is probably only 70% full.
Yeah, but it's not in a ghetto hotel this year! Nothing beats roof access that nobody gives a shit about. Watching the drizzle fall on the city from up there was a hell of a way to relax in the middle of a stressful con.
I don't think the Marriott will be so amused. On the other hand, there's a Starbuck's within easy range for any clever wireless user.
They still haven't gotten to the fun part: Transparent circuit boards (copper-clad glass substrate, brittle but beautiful) and transparent chip packaging.
Think about it! Most QFP and PGA chips have boring black plastic bodies. How hard would it be to replace them with clear plastic? Ceramic packages could probably be made at least translucent.
Then you embed light-emitting junctions at important areas of the chip, so you can watch the whole thing brighten, dim, and change color as the computational load changes. NOPs would be faint blue, cache misses would make the prefetch unit flash red. Floating-point would cause the FPU to glow green. Imagine it! You could tell what was eating most of your timeslices just by looking at the chip. Nevermind how you'd see through the heatsink to perceive all this.;) Oh wait, if the board's clear just watch all the action from underneath.
Seriously though, if the whole mobo chipset were clear-encased too, you could tell the difference between RAM accesses, drive activity, interrupts, DMA storms... Ooooh.
We already have SCSI terminators with activity indicators, am I really asking for too much?
(Now why didn't I patent this 5 years ago when I came up with it?)
Except I think Roadie had all his stuff screwed to the wall. The whole computer took up about the space of a poster, with power and ethernet leaving the area.
In hopes of avoiding the financial trouble that ruined Metricom, Aerie Networks will limit the Ricochet service to markets where the company can sign up enough customers to cover costs, executives said. The service proved particularly popular in the tech-driven Bay Area, where Ricochet had 25,000 subscribers.
This means, my friends, that we will NEVER see the service revived in Detroit. I wonder if it'd be possible for a community group to acquire the remaining poletops and infrastructure, and activate a grassroots wireless network?
At the very least, I'd like to put the poletops into a friendly mode where they'll pass packets for any customer modem that asks. Even if there's no route out to the internet, some wide-area data service would be great.
Slashdot Mirror
First, make sure your case has positive pressure with respect to the outside. All your filtered fans should face in, such that any leakage is in the out direction. This is very important because otherwise, your removable media drives will act as filters, trapping the dust from the air that gets sucked in through them.
In my case, the power supply fan is the only exhaust fan. All the other fans in the case face in, and all are equipped with filters. They flow less because of the filters, but they keep up with the power supply fan, so the case is ever so slightly pressurized.
Secondly, hardware stores abound with cheap filter material. There's this great open-cell foam filter that you're supposed to wrap around your air conditioner's condenser coil to keep cottonwood fluff out. I usually just blast the thing out with the garden hose... but the filter medium is ideal for trapping large dust, and since it's intended to wrap an entire condenser, you get about 20 square feet for just a couple bucks.
There are also these little 4"x12" filters designed to be slipped behind your HVAC system's registers. Don't do that -- your forced air furnace was designed for a particular flow rate, and impeding it will cause problems. The filters, on the other hand, are just what your computer needs. They're denser material, and they work well as a second stage filter, behind the foam described above.
Ironically, about the only material at the hardware store that doesn't work very well as a computer air filter, is furnace air filter media. It's too thick for convenient mounting, and it's hard to work with.
Thirdly, the case you pick is important. Try a large tower "server" case, they usually have plenty of room for mounting extra fans, and they're better designed with respect to airflow. Filters will drastically reduce the effectiveness of each fan, so plan for at least double your usual number of fans. (Invest in some quiet ones with the fluid bearings.)
A friend recently gave me an unused computer that just happened to be built in the world's coolest case. The little fan mounting trays have plenty of space to tuck filter material inside, and everything just clips into place, no tools needed. It's a SupermicrO tower. I've been inside a lot of computer cases, and this one is by far the best-designed I've ever seen. (I have no relationship to Supermicro, I'm just impressed by their product.)
I'm going to echo the sentiment of another poster who said to elevate the machine. Get it off the floor, although I don't see a problem with setting it on a table. Just make double-sure that it's not sucking air in through openings near the bottom. I've seen lots of office desktops turn into little stationary vacuum cleaners, neatly inhaling every shred of dust that falls near them.
Good luck!
Unless the cable company approves it. This would retransmit telecommunications service, and thus run afoul of Michigan's new law.
Actually it would be cool if your porn-loving neighbor got one of these, you could probably watch for free.
You're assuming that just because something operates at 2.4GHz, it must be 802.11b, and you're quite wrong.
The FCC regulates who can use what bits of the radio spectrum. Most pieces are allocated for things like radio broadcast, millitary use, amateur radio, GPS, etc.. There are a couple chunks, known as "part 15" bands, which are outlined in the rules for anyone to use, with a minimum of restrictions.
There's a band in the 900MHz range, and a band in 2.4GHz, and they're both called ISM, which stands for Industrial, Scientific, and Medical. I don't know why the words Consumer Putz didn't make it in there, because that's mostly what transmits there.
Cordless phones, wireless computer networking, certain video cameras, baby monitors, wireless microphones, you name it, all sorts of devices have been designed to operate in the 2.4GHz ISM band. It's handy! These new TVs work just like the X10 wireless video cameras, they use part of the spectrum to send video. They'll probably stomp all over an 802.11b LAN if you have one set up.
The world does not run on packets. There are things on the air, and things on wires, which are not Ethernet. This is my new mantra.
Remember those ads when the Zip came out, that featured a stack of seventysome floppies, towering and teetering? "You can store ALL THIS on one Zip disk!", they screamed.
Just recently I was looking at 120GB hard drives and thinking, shit, that would be one huge pile of floppies. Nevermind how long it would take to read them all!
Then I said, wait, what if each one were in a drive? Figure you can read/write an entire floppy in about a minute or two. The maximum sustained transfer rate on a hard drive isn't that great. Given a hundred thousand floppy drives and controllers for them all, you could beat the fastest hard drives with ease. Just imagine the noise and heat generation, though.
Let's hear it for stupid applications of obsolete technology! (next week's installment: dot matrix printers feeding sheetfed scanners as a backup medium / high-latency network interface.)
Unrelated to, and years earlier than, LS120. The floptical uses a floppy-like head mechanism, except it packs the tracks much closer together. It does this by using an optical pickup to align the head, rather than the dumb stepping used by regular floppy designs.
Does anyone know where these folks got the idea for this page? I could swear they got their inspiration from FreeSky Wireless, who made their first public announcement at Rubi-Con on Saturday.
It's obvious that both groups are very popular, since FreeSky also uses the same SSID, and all these folks must be members of both projects. It's great to see such a community-minded spirit across America!
This is an amazing kludge! They might as well have runners with CD-R's of the data, jogging up to the car as it passes by the station. Ugh. Cellular data is too slow, and wireless LANs are just too short-range. These folks would do better with a real wireless WAN system.
That's exactly the niche where Ricochet fit: High speed data, anywhere in a metropolitan area. While stationary, Ricochet speeds routinely exceeded 128kbps, and while moving at up to 60mph, the acquisition of new nodes pushed the throughput down, though not much. It would pick back up as soon as you stayed within range of one microcell for more than just a moment. (Software tweaks could improve mobile performance too.)
Imagine the applications of an ISDN-speed link in a moving patrol car: Realtime video from the dashboard camera. Instant images from headquarters displayed in the cruiser.
It's unfortunate that the present owners don't seem interested in reactivating the network in most of the old service areas (only Denver and San Fran are up so far), and they're certainly not planning to deploy equipment in any new cities. The LAPD would've loved Ricochet.
Ever since I heard about Digita, the OS for digital cameras, I wanted to program something like this...
When the prosecutor subpoena'd photos from reporters, to identify people involved in the MSU Riots, the reporters obviously cried foul: Their independence is critical to the accurate gathering of news. If people see reporters as an arm of the law, coverage becomes skewed.
What if the camera could be loaded with a public key, and encrypt photos as they're taken, so that only the private key (back at the Editor's desk) could decrypt them? Perhaps keep a plaintext thumbnail in RAM for convenience, but make sure it vanishes if the card is removed or the camera's powered off.
It still doesn't prevent the judge from throwing the editor in jail for failing to turn over the private keys, but it adds a layer of complication, where the editor could simply lose the keys. Would that be destroying evidence? The encrypted photos still exist -- they're just unreadable.
Have you looked into the Terapin Mine? It sounds like exactly what you're looking for:
USB master, USB slave, PCMCIA slot, 10baseT, and it's small! Replace the rotating-platters laptop drive with a solid state version if you're really worried about vibration.
Why not make it render something different but valid in as many different browsers as possible?
My sympathy goes out to the judges of this contest.
No shit! The original fit on four 1.2 meg floppies. The source to the PC version was lost, however. What's up on SF now is based on the 3DO version, which apparently had speech. (I'm unwilling to download 140 meg over dialup to find out.)
.MOD files from the original version are still on my drive and I listen to them regularly.. They've been shuffled around as I've upgraded hard drives over the years, but the original file creation dates in 1994 remain in the directory. I think at some point I put them all in a .zip which would've preserved the file dates.
Odd tidbit: The
I'm pretty excited about this, I'd love to see a network-enabled version of super-Melee. Anyone with the requisite skill wanna try it? (I once tried to fix gorilla.bas to play across a lan. Since then I've left such things to professionals.)
Funny, I have a bad connector on my keyboard.. it's an IBM M-series (clicky monster), the type with the removable cord. Whenever I set the keyboard down sharply, it must jiggle the power pin or something, and it doesn't work until I reseat the connector.
Of course this causes the keyboard to reinitialize, which hoses up my typematic rate, so I have a hotkey (Ctrl-Alt-X) which pops up the Keyboard control panel. I just change a setting then change it back, then hit OK. Windows commits the "changes" by writing them to the keyboard, and my repeat rate is back to normal.
I've done this hundreds of times now, it happens once or twice a day and I've been using this keyboard at least since '99. Nothing's fried yet. You'd think I'd have the common sense to open the thing and resolder the connector.
Y'know what DOES cook motherboards? Remeber when cases had keyboard locks on them? The two-wire lead from the keyboard lock was installed backwards, so the metal body of the lock wasn't connected to ground, but to the keyboard lock sense lead. One dry winter day, I touched the lock and drew a large spark. The machine locked up and by the time I reached for the power button, I was smelling smoke.
Opening the machine revealed that the keyboard BIOS EPROM had gotten quite hot, enough to wrinkle and discolor the shiny sticker placed over the erase window. Just for kicks, we plugged the machine back in and powered it up to see what would happen.
Much to our surprise, the machine booted normally! The keyboard worked and everything, and that system remained in service for another 2 or 3 years before bloatware forced a motherboard upgrade. During the upgrade, we got a good chuckle out of the memory of the melted sticker.
I got "You do not exist. Go away." while trying to ctrl-alt-del a Slackware installation many years ago.
I giggled for a moment, then got really worried and jabbed for the power switch.
The FBI could quite easily sneak a wireless AP onto the library's intranet, place the AP into bridging or sniffing mode (see the OpenAP project for ideas), and watch database lookups from their van parked at the coffee shop down the street.
That's beside the point anyway. Patrons frequently use the library terminals for things they can't find in the dead-tree collection. This does require an internet connection.
Here's an idea: Many libraries are on reasonably fast connections. Someone needs to create a small ram-based Linux distribution in a DOS executable. Something like Loadlin packed with a kernel image and filesystem image.
Then place this distro on a server with some carefully-scripted web pages which exploit flaws in IE to run the package. This is because the library probably runs Windows security software that prevents the user from easily downloading and running arbitrary programs.
The library patron walks up to the computer, and punches in the URL of this distro. The thing downloads, IE is tricked into running it, and the library's computer is now running a real OS from RAM.
Build in whatever freedom and paranoia you want: Perhaps the kernel shuffles things around in RAM every few seconds to prevent memory cell persistence. Perhaps the browsing is done via a secure tunnel to prevent the feds from simply Carnivoring the library's pipe. Perhaps this system meets other similar systems (gnutella-style?) on the net and they proxy each other's data (Freedom.net-style) to make things hard to trace. Perhaps the screen font is Tempest-annoying.
When the patron is finished, the distro wipes memory and reboots, back into the library's Windows for the next person to come along.
Of course such a distro would also be used by lame crackers with nothing better to do than attack websites from the library. *sigh*
The hard drive is but a small chunk of a computer. I don't mind if the drive is sealed and the firmware proprietary, as long as it conforms to a well-documented standard interface, and doesn't require any proprietary tweaks to perform well.
Likewise, I don't care what shape my pistons are ground in, or how the valvetrain actually works. Overhead or pushrod, seal it up so I never see it, I don't care.
However, my computer as a whole had better not be welded shut! I want to be able to inspect its operation, make sure I'm the user and it's the tool, not the other way around. I want to be able to control its operation, write my own programs, and tweak the misfeatures of others' programs. I'll never need to manually adjust the flying gap on my hard drive, but I definitely do want to customize my boot sequence, and scrutinize data leaving over my net connection.
Likewise, my car should be open enough that I can change the parts I want to. I'm happy with the engine, but why can't I turn off my headlights when I pull into my friend's driveway, so as not to wake his neighbors? I want wiring diagrams, so I can make the lighter socket behave more sensibly. I don't care if certain parts are sealed for practical reasons, but the rest of it should be open for me, the owner, to customize.
I think your analoy is flawed because you compare the hard drive to a car. The drive can be a black-box with no ill effect. It's the computer system as a whole that must be open to the user. The car can contain several black-box components, but I should be free to inspect and reconfigure the connections between them.
If you think this tendency in cars is scary, go read this article on Microsoft's Palladium plan. Talk about welded shut! At least your car doesn't refuse to carry passengers and cargo that haven't been approved by the manufacturer.
Like the things your buttons go into, that have the stitched reinforcements on the sides?
Still, a hilarious typo.
I'm aware that a good reference has multiple inputs, I'm simply saying that there isn't a good reference at most offices.
The CO's I've been in have a Telecom Solutions (by Symmetricom) DCD-LPR with GPS GTI cards feeding a DCD-ST2 with Stratum-2 oscillators, which drives the TOxA cards to feed the BITS-clocked network elements in the office.
In such a situation, if the GTI boards lose lock, the ST2 shelf goes into holdover, where it should be good for a few days. (I don't have the specs in front of me.) Equipment still has timing, it's just not locked to anything in particular. The switch and stuff will continue to run, but interoffice links will suffer as slip increases.
I'm sure all the major carriers had a Cesium reference in an office at one time, but nowadays I don't think that's used for anything. It's simply too awkward to push that signal out to each office. The GPS constellation is considered the primary reference.
The phone system won't go down completely, but it will break up into islands until a terrestrial sync distribution system can be established, or GPS can be restored.
Sure, the GPS satellites could be taken out by a rogue nation with too much laser power on their hands. The orbit data are public, after all. It's not a direct military strike, just a nasty thing to do, with repercussions that wouldn't be realized until after the fact.
The entire infrastructure is carried on SONET equipment. (That's Synchronous Optical Network, and if you didn't know that, you should read up on it, it's neat stuff.) Being synchronous, this stuff royally shits if there's something wrong with the timing.
Way back when T-carrier was first deployed, Bell realized this and set up a nationwide synchronization distribution. I think the master clock was in Kansas City. Anyway, the sync signal was distributed over wireline circuits to every central office in the country. Maybe Canada too?
However, most interoffice links are fiber now, the same SONET rings that depend on such precise synchronization. Ring-timing is awkward, and without very careful planning, sync loops can form. (Long story, look it up. The short version is that when a SONET system loses sync, it doesn't carry traffic.)
The modern concept is called BITS, or Building Integrated Timing Supply. Each office has a sync signal source, driven by an LPR (local primary reference) oscillator, which is in turn frequency-locked to a reference signal derived from GPS satellite signals.
Yes, that's right, the whole telephone network will fall apart if the Global Positioning System stops transmitting. Depending on the stratum class of the LPR, it might be able to "hold over" for a couple days, maintaining an accurate timing signal in the absence of an upstream reference. They will eventually drift, and most offices only have stratum-3 units anyway.
The network is so poorly planned in the first place, most transport engineers haven't got a clue about ring timing and such. They just hook each terminal to the BITS clock and hope it works, which it does, until something happens to the BITS clock. If all the BITSes in the network started drifting from one another, the system would slowly fail over a few days, as timing slips exceeded the tolerances of the various systems.
If such a thing were to happen, don't bet on the ability to patch things up quickly. Recordkeeping is horrible, and even if it weren't, it would be a daunting task to spontaneously set up a new sync distribution network independent of GPS.
I've heard on good authority that you wouldn't even need to take out the satellites themselves. A couple properly placed nuclear detinations could screw up the somethingsphere such that GPS signal propagation would suffer. Any physicists care to clarify?
Point 1: When a telco person says "switch", it means something totally different than what a data person means when they say "switch". This is a persistent annoyance.
You can't simply packet an ESS out of existence, because it doesn't know what a packet is. It's not connected to the internet. There are SS7 signaling links and X.25 control links, and maybe a few IP control links if you're lucky. None of them are connected to the internet. Your phone line is payload, not control.
Exactly how do you propose to access the switch in order to DoS it? There are switch dialins, but most are pretty secure, and good luck finding them. You're planning to do a lot of wardialing first?
Point 2: Telcos lie about bandwidth. When someone says they have a 10 Gigabit backbone, it means they own a couple OC192 circuits. Most of the channels in those circuits are probably not filled.
That's like saying I can move a thousand shipping containers a day, because there's a large river between me and my destination, and seaports at each end. Nevermind that I don't own any ships!
An OC192 circuit, for instance, can carry four OC48 signals, or 16 OC12 signals, or a mix thereof. Anything that adds up to 192 STS-1 payload envelopes, or equivalent concatenated payloads. You get the idea. Chances are, they're carrying one or two OC48s on the thing, and the rest is for future expansion. Each of those OC48s in turn is probably only 70% full.
Yeah, but it's not in a ghetto hotel this year! Nothing beats roof access that nobody gives a shit about. Watching the drizzle fall on the city from up there was a hell of a way to relax in the middle of a stressful con.
I don't think the Marriott will be so amused. On the other hand, there's a Starbuck's within easy range for any clever wireless user.
They still haven't gotten to the fun part: Transparent circuit boards (copper-clad glass substrate, brittle but beautiful) and transparent chip packaging.
;) Oh wait, if the board's clear just watch all the action from underneath.
Think about it! Most QFP and PGA chips have boring black plastic bodies. How hard would it be to replace them with clear plastic? Ceramic packages could probably be made at least translucent.
Then you embed light-emitting junctions at important areas of the chip, so you can watch the whole thing brighten, dim, and change color as the computational load changes. NOPs would be faint blue, cache misses would make the prefetch unit flash red. Floating-point would cause the FPU to glow green. Imagine it! You could tell what was eating most of your timeslices just by looking at the chip. Nevermind how you'd see through the heatsink to perceive all this.
Seriously though, if the whole mobo chipset were clear-encased too, you could tell the difference between RAM accesses, drive activity, interrupts, DMA storms... Ooooh.
We already have SCSI terminators with activity indicators, am I really asking for too much?
(Now why didn't I patent this 5 years ago when I came up with it?)
Except I think Roadie had all his stuff screwed to the wall. The whole computer took up about the space of a poster, with power and ethernet leaving the area.
At the very least, I'd like to put the poletops into a friendly mode where they'll pass packets for any customer modem that asks. Even if there's no route out to the internet, some wide-area data service would be great.
Didn't they originally host jegelhof's AOLsucks page?
Damn, one of our own.
is finished. :)
LOAD "*",8
(retrieve munchies from fridge)
(complain that the 1541 drive is a slow P.O.S.)
(fall asleep)
READY.