I doubt any Grand Marquis versions had the Taxi package. What you want is the 2.73:1 axle gears out of the taxi version of the Crown Vic. The easiest way is to swap the entire axle (unless you have rear air suspension, in which case, yeah, you're going to have to either take the axle apart and switch gears, or chop the air suspension out of the Marquis de Sade and switch to springs. And, if you have air suspension, you're going to have to replace or repair it eventually and it isn't cheap. Anyway, the 2.73:1 gears will turn in about 30 mpg on the highway and the car will run quieter to boot.
Fastest way to spot a Taxi version is to look for "P71" in the VIN on the dash through the windshield. When you find one, then you can crawl under it and read the numbers off of the sheet metal tag attached to one of the differential cover bolts. You're looking for "73" or "273", depending on the year. P71s are taxis, but quite a few of them got used as police cars by more cost-conscious departments.
It turns out you actually can use several smaller towers instead of one big one for broadcasting. In digital TV, it's trivially easy - several fully independent stations pump out their signal on various UHF frequencies, but they all use the same PSIP - the "virtual" channel number encoded into the stream so the blissfully ignorant can "keep on watching channel three like I always did, bohygawd." Consumer decoder boxes are clever enough to only offer the one with the lowest bit error rate (=best signal). I have no idea how you license this, and if I did have some idea under 47CFR, it wouldn't apply in Tokyo, so whatever...
Analog audio broadcasting is slightly more complicated, and this is where it gets good. With AM, you can phase lock the carriers (and as a result, the modulation) to GPS. Sub-nanosecond phase accuracy is more than good enough at 1700 kilocycles. Vaisala will sell you such a rig off-the-shelf. It's kind of a neat way to get around licensing, too: each station can be configured as a legal, license-free Part 15 device, and yet with a handful of them you can cover a decent-sized town. FM is a lot more expensive, but it works the same way. You just write a bigger check.
Bonus points for running Bluetooth USB adapters without actually having a computer connected (at least after startup). I experimented some with this sort of scheme and my one warning is this: it takes 10.24 seconds to check all possible hopping patterns for all possible Bluetooth devices, so if it needs to respond quickly, you're hosed. But it's good enough for showing off, or updating a "Boss Proximity Detector", or that sort of thing.
The FCC licenses US Citizens and Foreign Nationals living in the US not under the employment of a foreign government. If you're, say, a foreign embassy employee, then you need to be licensed by your home government's agency.
There are an awful lot of PCI slots in all sorts of embedded systems out there, and some of them may be looking for a graphics upgrade. For that matter, I still have instruments that have EISA slots in them. I suspect I'll be running them for another decade. 10Base2 is getting to be a pain to deal with, though.
If your code is pure MPI C or Fortran, then the BG is a decent idea. Remember, the original name of the machine was "QCDOC", or "QCD On a Chip" - if you're running QCD, it rocks. Other things, not so good. Let's say you have a big code in Java and you want to run it on your Blue Gene. Well, you're screwed - there's no JVM for the worker nodes. Let's say you have a big code in Perl (and don't laugh - Perl is what about half of computational biology gets done in). That's a problem, because there's no OS on the nodes, so there's no way to run Perl. Couple that with the bugginess of the software, the brittleness of the hardware, and desktop-class I/O and you have a machine that basically is just good for QCD and linpack. So, yeah, running a real OS on the nodes isn't all that bad an idea. Which is probably why slashdot reported on the port of Plan 9 for Blue Gene back in 2007. Links to "official" IBM site down in there, which is now throwing Lotus Notes' version of a 404 - and did we expect anything else from IBM?
The Rocks approach is nice for quickly regenerating a failed node. And it's Centos under the covers, as noted, so it's RHEL in disguise. If you're running 16 boxes with dual quad-cores, you'll lose the occasional disk drive. If you run 64 cheap desktops with single-socket dual-cores, you'll lose a disk drive every week or two.
I have some experience flying 15 foot diameter balloons for research. If you use a regulator to control the flow of He then the regulator can get frosty. If you skip the regulator and just use a CGA580-to-hose-barb adapter, then it will ice over pretty quickly and you'll be glad you have work gloves. Our experience was that as long as there was still humidity in the air to freeze, it wouldn't get colder than 0 Celsius. Which was exactly what we expected. Oh, and skipping the regulator sounds like a good idea - faster flow, faster fills. Except it isn't - the rate limiting step is how fast the balloon will accept gas before it just tears itself up.
And no, you don't need a coat unless it's just generally cold outside anyway.
Commercial outfits still offer it but it's a pain to license outside of the polar regions. Hams (Amateur Radio) have picked it up to great effect: http://www.physics.princeton.edu/pulsar/K1JT/
All you need to do is define logicals in your login.com to refer to the disks - USERNAME$THEDISK for instance. Then always use those names. As far as set def silliness, you're on your own.
I can't believe I remember this crap. It's been a shade over twenty years. So, that gets me wondering - does the NT kernel have logicals? It'd be nice to substitute parts of the filesystem from time to time...
Unmanned vehicles are, by regulation, not feasible for commercial use in the US. Having been through this with sciences and engineering departments at two universities (jointly operated institute), I can tell you with certainty that you wouldn't believe the hassle. To fly a UAV, you need a type certified UAV for US airspace, and there aren't any, and it will be decades before there are. So you have to do an individual, case-by-case type approval. It takes months, it takes engineers, and it takes lawyers. Then you have to have a pilot with an IFR rating file the flight plan. Then you have to have positive surveillance of your Area of Operations (AoO) - you're allowed to use either radar or a chase plane. The (manned) chase plane is cheaper, by the way. Oh, and you have to show that no manned aircraft is suitable for your mission (and if your mission is "training undergraduate aerospace engineers how to design, build, and operate UAVs", then it's easier).
Far, far easier is to work with the National Guard to use the no-fly zone over Ft. Bragg or some other similar federal installations.
Compare this with the fact that you can rent a small plane, a pilot, and a camera for less than $200/hr *right this minute* and you'll see why no one even bothers with UAVs, except as research projects.
You can only operate under the model aircraft rules if you aren't being paid, and no, you can't just claim you're taking a vacation day. The university would have to testify that you stole the UAV if anything happened.
Let's see - SAS from NC State, Linux from U. Helsinki, X11 from MIT, kerberos from MIT, BSD from Berkeley, Maple from Waterloo (?). Matlab from U. of New Mexico. Firefox from Mozilla from Netscape from Mosaic from UIUC. I'd say pretty much any interesting software I can think of came from a university one way or another.
Amidst the sea of negativity, I feel obligated to point out that this has been done. Much to my surprise, it worked. Course development was through The Shodor Foundation and a faculty member from the NC School of Science and Mathematics, Bob Gotwols. The course is aimed at advanced students but who haven't had diff eqns yet. They use WebMO as a front end to GAMES, GAUSIAN, and all the other usual suspects. Hardware was fairly modest - seemed like maybe two or four linux boxes.
The former will be really useful if you decide to work on database-related things. Not just "using SQL to get my work done" but actually crafting the internals of a database. Similarly useful for compilers - both of them have optimizers, and that's just one big graph traversal. Too big to do in a useful amount of time, so all kinds of heuristics are used for graph pruning. Go for it.
The latter is a good foundation for numerical analysis, a field occupied by a lot of engineers and fairly few CS type people. It's a nice differentiator (pardon the absurdly weak pun). You'd much, much rather try to find simulation work than, say, writing yet another inventory management application these days.
But if they didn't start two years ago, their stuff won't be publishable by the time it's finished. 48 cores isn't really jack these days. To get on the Top 500 list, you need, very roughly, >2K cores. It's still a big, expensive game to play in, and if your institution doesn't have a pretty big machine, you need to look at the solicitations from the national labs and get some processor-hours there.
I've been using a TS-7800 from Technologic Systems for a few months, running it off of solar panels. It draws 4 watts and has half a gig of flash on board and an SD card socket. It runs cool without even so much as a heatsink, let alone a fan. Gig-E, 10 serial ports, 6 A/D, more digital I/O than I could even use, and USB. Runs Debian. Buy the development kit - the slight extra cost is worth it.
You know, a company only pays taxes on the money they make. Businesses that are losing money don't pay taxes. (OK, this might shield them from property taxes on the buildings and equipment, depending on state and municipal laws). It's a nice gesture, but it will have absolutely no tangible effect.
Slashdot Mirror
I doubt any Grand Marquis versions had the Taxi package. What you want is the 2.73:1 axle gears out of the taxi version of the Crown Vic. The easiest way is to swap the entire axle (unless you have rear air suspension, in which case, yeah, you're going to have to either take the axle apart and switch gears, or chop the air suspension out of the Marquis de Sade and switch to springs. And, if you have air suspension, you're going to have to replace or repair it eventually and it isn't cheap. Anyway, the 2.73:1 gears will turn in about 30 mpg on the highway and the car will run quieter to boot.
Fastest way to spot a Taxi version is to look for "P71" in the VIN on the dash through the windshield. When you find one, then you can crawl under it and read the numbers off of the sheet metal tag attached to one of the differential cover bolts. You're looking for "73" or "273", depending on the year. P71s are taxis, but quite a few of them got used as police cars by more cost-conscious departments.
No, a fork(2).
It turns out you actually can use several smaller towers instead of one big one for broadcasting. In digital TV, it's trivially easy - several fully independent stations pump out their signal on various UHF frequencies, but they all use the same PSIP - the "virtual" channel number encoded into the stream so the blissfully ignorant can "keep on watching channel three like I always did, bohygawd." Consumer decoder boxes are clever enough to only offer the one with the lowest bit error rate (=best signal). I have no idea how you license this, and if I did have some idea under 47CFR, it wouldn't apply in Tokyo, so whatever...
Analog audio broadcasting is slightly more complicated, and this is where it gets good. With AM, you can phase lock the carriers (and as a result, the modulation) to GPS. Sub-nanosecond phase accuracy is more than good enough at 1700 kilocycles. Vaisala will sell you such a rig off-the-shelf. It's kind of a neat way to get around licensing, too: each station can be configured as a legal, license-free Part 15 device, and yet with a handful of them you can cover a decent-sized town. FM is a lot more expensive, but it works the same way. You just write a bigger check.
http://dspace.mit.edu/bitstream/handle/1721.1/7431/CS021.pdf
Bonus points for running Bluetooth USB adapters without actually having a computer connected (at least after startup). I experimented some with this sort of scheme and my one warning is this: it takes 10.24 seconds to check all possible hopping patterns for all possible Bluetooth devices, so if it needs to respond quickly, you're hosed. But it's good enough for showing off, or updating a "Boss Proximity Detector", or that sort of thing.
The FCC licenses US Citizens and Foreign Nationals living in the US not under the employment of a foreign government. If you're, say, a foreign embassy employee, then you need to be licensed by your home government's agency.
Nit picky, huh? :-)
There are an awful lot of PCI slots in all sorts of embedded systems out there, and some of them may be looking for a graphics upgrade. For that matter, I still have instruments that have EISA slots in them. I suspect I'll be running them for another decade. 10Base2 is getting to be a pain to deal with, though.
If your code is pure MPI C or Fortran, then the BG is a decent idea. Remember, the original name of the machine was "QCDOC", or "QCD On a Chip" - if you're running QCD, it rocks. Other things, not so good. Let's say you have a big code in Java and you want to run it on your Blue Gene. Well, you're screwed - there's no JVM for the worker nodes. Let's say you have a big code in Perl (and don't laugh - Perl is what about half of computational biology gets done in). That's a problem, because there's no OS on the nodes, so there's no way to run Perl. Couple that with the bugginess of the software, the brittleness of the hardware, and desktop-class I/O and you have a machine that basically is just good for QCD and linpack. So, yeah, running a real OS on the nodes isn't all that bad an idea. Which is probably why slashdot reported on the port of Plan 9 for Blue Gene back in 2007. Links to "official" IBM site down in there, which is now throwing Lotus Notes' version of a 404 - and did we expect anything else from IBM?
The Rocks approach is nice for quickly regenerating a failed node. And it's Centos under the covers, as noted, so it's RHEL in disguise. If you're running 16 boxes with dual quad-cores, you'll lose the occasional disk drive. If you run 64 cheap desktops with single-socket dual-cores, you'll lose a disk drive every week or two.
I have some experience flying 15 foot diameter balloons for research. If you use a regulator to control the flow of He then the regulator can get frosty. If you skip the regulator and just use a CGA580-to-hose-barb adapter, then it will ice over pretty quickly and you'll be glad you have work gloves. Our experience was that as long as there was still humidity in the air to freeze, it wouldn't get colder than 0 Celsius. Which was exactly what we expected. Oh, and skipping the regulator sounds like a good idea - faster flow, faster fills. Except it isn't - the rate limiting step is how fast the balloon will accept gas before it just tears itself up. And no, you don't need a coat unless it's just generally cold outside anyway.
Commercial outfits still offer it but it's a pain to license outside of the polar regions. Hams (Amateur Radio) have picked it up to great effect: http://www.physics.princeton.edu/pulsar/K1JT/
I can't believe I remember this crap. It's been a shade over twenty years. So, that gets me wondering - does the NT kernel have logicals? It'd be nice to substitute parts of the filesystem from time to time...
Far, far easier is to work with the National Guard to use the no-fly zone over Ft. Bragg or some other similar federal installations.
Compare this with the fact that you can rent a small plane, a pilot, and a camera for less than $200/hr *right this minute* and you'll see why no one even bothers with UAVs, except as research projects.
You can only operate under the model aircraft rules if you aren't being paid, and no, you can't just claim you're taking a vacation day. The university would have to testify that you stole the UAV if anything happened.
Let's see - SAS from NC State, Linux from U. Helsinki, X11 from MIT, kerberos from MIT, BSD from Berkeley, Maple from Waterloo (?). Matlab from U. of New Mexico. Firefox from Mozilla from Netscape from Mosaic from UIUC. I'd say pretty much any interesting software I can think of came from a university one way or another.
Amidst the sea of negativity, I feel obligated to point out that this has been done. Much to my surprise, it worked. Course development was through The Shodor Foundation and a faculty member from the NC School of Science and Mathematics, Bob Gotwols. The course is aimed at advanced students but who haven't had diff eqns yet. They use WebMO as a front end to GAMES, GAUSIAN, and all the other usual suspects. Hardware was fairly modest - seemed like maybe two or four linux boxes.
Ah, someone hasn't been eating in the south lately... yum.
The former will be really useful if you decide to work on database-related things. Not just "using SQL to get my work done" but actually crafting the internals of a database. Similarly useful for compilers - both of them have optimizers, and that's just one big graph traversal. Too big to do in a useful amount of time, so all kinds of heuristics are used for graph pruning. Go for it.
The latter is a good foundation for numerical analysis, a field occupied by a lot of engineers and fairly few CS type people. It's a nice differentiator (pardon the absurdly weak pun). You'd much, much rather try to find simulation work than, say, writing yet another inventory management application these days.
But if they didn't start two years ago, their stuff won't be publishable by the time it's finished. 48 cores isn't really jack these days. To get on the Top 500 list, you need, very roughly, >2K cores. It's still a big, expensive game to play in, and if your institution doesn't have a pretty big machine, you need to look at the solicitations from the national labs and get some processor-hours there.
I've been using a TS-7800 from Technologic Systems for a few months, running it off of solar panels. It draws 4 watts and has half a gig of flash on board and an SD card socket. It runs cool without even so much as a heatsink, let alone a fan. Gig-E, 10 serial ports, 6 A/D, more digital I/O than I could even use, and USB. Runs Debian. Buy the development kit - the slight extra cost is worth it.
Lost to the ages is a couple of Goodyear supercomputers, remembered really only as a footnote in computer architecture textbooks...
You know, a company only pays taxes on the money they make. Businesses that are losing money don't pay taxes. (OK, this might shield them from property taxes on the buildings and equipment, depending on state and municipal laws). It's a nice gesture, but it will have absolutely no tangible effect.