Yes, because GPS receivers are tiny and don't need a few minutes with a clear view of the sky to get a fix. And cell phone modules are tiny and work amazingly well without any kind of antenna. And all of this fits inside a tiny, functional USB drive...
The beacon is a far more plausible scenario, but now you've got to try and find it somewhere in the world.:-)
The only problem I've ever had with Slackware is that it didn't always do as good a job at recognizing USB devices as some others.
My only real problem with Slackware is that Pat won't modify the base vim install to
set nocompatible
in the global vimrc. Damn that bugs the shit out of me. I've emailed him and he was very nice about it, he simply said that he does the bare minimum mangling of any package configuration and if I could convince the vim maintainer to make that change that Slackware'd have it. Now the ball's back in my court. Damnit.:-)
Offhand, why is this not a default config option for vim? The cursor keys are totally useless without it.
that's a bad idea swaret and slapt-get can bork your install very easily unless you know exactly what slapt-get does
I hear this again and again. While I can see it being dangerous for an upgrade, I have also heard the general "swaret will eat your system" under normal use. Do you have any proof to this end?
If it really bothers you then DON'T enter your login name and password,
just click on the login button. You will be redirected to a secure page
asking you to please enter you login infomation.
Actually that's exactly what I do -- If I don't see the browser bar go yellow I don't put any information in, regardless of whether it's through a secure form or not. I understand that the story's exploit would turn my browser bar yellow so I'd be lulled into a false sense of security to begin with, but my whole point was that having the login credentials on that page, secure form or not, is a poor idea.
If they want to save their CPU power for casual visitors, they should just have a "Log in now" link, like my bank does. It just seems to be a good idea, and follows the path of least surprise which is what every UI should strive for.
Paypal's main site (http://www.paypal.com) does *NOT* do a permanent redirect to https://www.paypal.com, so if you hit www.paypal.com you give your paypal login and password in the clear. I've emailed them several times on this and have finally given up, as they don't bother to respond.
So if you can get inbetween Paypal and your target, you don't even need to fool anybody.
Forget speakeasy bonded T1, you can bond your standard DSL lines through an OpenBSD firewall using CARP. Read also about VRRP and (HSRP and GLBP) for cisco solutions. They add not only redundancy but also load balancing, and recovery is real fast as opposed to something like RIP2. You can also use OSPF but careful, OpenOSPFD and zebra dont provide load balancing and redundancy of default routes. IOS does.
I recently did something similar with Linux.
Business wanted bigger bandwidth and higher availability than was available in town without going to expensive bandwidth. We took ADSL from two different providers (could have also used wireless, cable, etc.) and a machine in Toronto in a data centre on a multihomed network and created a pair of GRE tunnels between them.
The multihomed server extrudes an IP on its network for the business, and the ip-up scripts for the PPPoE connections fiddle with the tunnels. Now they have an IP address which is located in multihomed space in this little old town of 5300 people, and through creative use of iproute2 the only multihomed bandwidth they use is what they actually really need to be redundant. Everything else (basic web browsing and outgoing email being the biggest ones) is just coarsely load-balanced over the basic PPPoE connections, without travelling over the tunnels. I've got route caching turned off so it works pretty well.
The next step is to encrypt the tunnels using OpenS/WAN -- that was the original idea, but unfortunately it's not possible to do this without the GRE tunnels, since OpenS/WAN would get confused about which traffic was for which connection since they both have the same endpoints in the configuration file. I'm sure I could use another encryption package but I'd rather stick to something standard.
I had to get T.H.E. to replace my N64 controllers 3 times until i learnt to dismantle the joystick and fill it with vaseline to stop it wearing out after a month of use.
You filled your N64 joystick (I'm assuming the analog joystick, not the entire controller) with vaseline?
you do realize that the analog joysticks on those things are optical, right?
'm not faulting the engineers. I'm just saying that if these predicted problems were really as some people painted them (inevitable), then engineers would surely share some of the blame for not communicating properly if it was not fixed.
Engineer: "Don't point the gun at that clown and pull the trigger, because he will die."
Manager: "Don't be absurd, there is a chance I may not even hit him."
Engineer: "Don't point the gun at that clown and pull the trigger, because he will die."
Manager: "Thank you for your analysis of the situation. I've determined that the risk is worth the reward."
The engineer is to blame, how exactly? Managers go against the judgment of Those In The Know all the time, even effectively communicated. If the engineer says "We can't launch, the rocket will explode because of A and B and C, not to mention Z," and the manager decides to push ahead with the launch anyway, how did the engineer not effectively communicate this? You cannot lay blame on someone for someone else's stupidity or willful ignorance.
Sometimes, sure, an engineer may downplay a problem. THAT is something he can surely be blamed for. I was taking exception to your blanket statement where you said that if a project has inevitable problems and is pushed ahead anyway, that the engineers must share the blame (supposedly because Managers never go against advice given to them).
Something else to keep in mind : if I spotted a problem that would surely doom my project, and can't get engineering leadership/management to agree with me, I should share some of the blame.
I beg your pardon?
If I bring an issue up and explain its ramifications and make sure everyone is very much aware of what is at stake and my manager decides that the chance of failure is too small to worry about fixing the issue, how exactly should I be sharing the blame? I did my job; I documented that I did my job. I am clearly not in the wrong here.
If every ISP blocked outgoing SMTP messages from their users and either 1) forced them to relay mail through their servers or 2) ensured that any user-run mail servers were properly configured with SPF, etc. before allowing them to access outgoing port 25 traffic, it would allow allow much better assurance that the sender was who they said they were.
Nonsense. I'd say almost ALL of my spam these days comes from zombies who ARE using their ISP's SMTP server. Neither #1 or #2 fix that.
It worst habbit is turning off the phone for no apparent reason.
No, it's rebooting and not turning the phone back on. Get something like RadioState (it's free) to fix this. I have to admit though my 650 RARELY reboots now. It's still a pain when it does, but it's very, very, very infrequent.
Many publishing houses produce all their output with LaTeX, so saying it's useful just for math doesn't reflect its actual usage.
I'm trying to find some kind of latex front end that I can give to the people who make the manuals here... We've gone from Word to Ventura and dabbled a little in FrameMaker... nothing seems to do quite what we want, so I'm looking at latex. Lyx is a little weak here, and the other products I've found are either quite old (TrueTex) or geared heavily toward mathematical publication (Scientific Word). PCTex seems to be about the closest (write in word, mangle in PCTex, export how you want) but that seems to keep some of the biggest encumbrances I've got (people trying to do proper layout in Word).
Any suggestions? Our manuals are about 100 to 250 pages long, some diagrams, but mostly "Parameter Foo: values 1,2,3, meanings x,y,z, caution, warning, application hint, see Parameter Bar (page 50) and Baz (Page 90)" type of things. Some charts, and a display shot for each parameter... Nothing fancy, but lots of cross referencing.
Now if you guys can just make Thunderbird stop sucking, I'd be much happier.
Agreed. I can't compose messages in plain text? I can't have signature lines automatically removed when replying and quoting? I can't change the name of my outgoing account when composing? Crazy. Gimme kmail on Win32 and I'll be much happier.
It is time for a second Internet to come into action -- one that is voluntarily connected, one that is run over cabling (or satellite) connections that are not subsidized by any government regime.
L0pht Heavy Industries had quite a bit of information about this. They called it guerilla.net. It seems to be gone now, though. All 9600 baud ham, wifi and optical links... Very cool idea but the problem with it is the same problem that faces amateur radio these days. Lack of general interest.
It's sad. There is a lot of good that can come from this kind of thing, but people don't give a shit unless it has something to do with the latest fads on gossip on TV.
Exchange4Linux is what I've chosen, after having evaluated a dozen others, including SLOX, OpenOffice's equivalent, a few web-based ones, Hitachi's and even the one the Oracle bought out and closed down.
So no, there are plenty of Exchange replacements out there. For me, Exchange4Linux is the only one which actually replaces Exchange properly, and the server is 100% open source.
The treo650 I have allows you to have different rings for calls or SMS messages, and even different rings per caller-id (based on the number in the address book). I'm fairly certain the 600 does this too.
The other solution is to put in a very high-efficiency switching power supply, but those are very expensive.
If you're an EE, you're not a very good one.
There are switchers in TVs (most modern televisions run practically everything off of a small switcher and a larger switcher running all secondary supplies off of the flyback circuit) -- There hasn't ever been a linear supply in a PC power supply (even the old XTs ran switchers) and your DVD player, receiver, UPS... all switchers. Your basic switcher efficiency these days starts at about 80% under moderate load, with most achieving upward of 95% without even doing careful design. Now yes, while idle, they often drop to ~70% even when pulse-dropping but let's be honest here... 70% efficiency on pulling a hundred miliamps is not exactly killing the grid, nor is it adding up to any appreciable amount when multiplied by a million units, especially not compared to the HVAC and lighting requirements of any urban business centre, or the electrical requirements of even a couple of industrial processes.
To get back on topic, switchmode supplies are dirt cheap. The magnetics are smaller and lighter and cheaper than 60Hz transformers, the controllers are oftentimes cheaper than the ancient 78xx series regulators, the run cooler which means you save money on heat dissipation, you save weight because you don't need the same heat sinks... They are far cheaper than any old linear circuit for any appreciable load. Now you do run into higher output harmonic content with a switchmode power supply but simple filters and efficient switch design (already built in to most switchmode controller circuitry) minimize these effects, and linear supplies generate much different, and much more difficult to deal with input harmonic distortion.
In fact the benefits of switchers over linear supplies is so great that it's actually illegal to power any commercial equipment with linear supplies in the EU. You simply cannot achieve CE approval with a linear power supply due to the input current waveform distortion. Switchers spread out the input current wave over more of the voltage waveform, saving the power company from having to supply 3x-4x the current over a very narrow part of the voltage waveform.
Switchers are cheap. Hell even expensive switchmode supplies are cheap compared to the magnetics, semiconductors and heatsinking required for the equivalent power level linear supply, and that's not including weight and shipping considerations. I can't even think of a wall wart I have for a doo-dad in my house that doesn't use a switcher over a linear supply. Honestly, this is a non-argument, and yet I've managed to babble on about it for what, four paragraphs?:-)
You do realize that almost every single electronic device that consumes any appreciable amount of energy these days does not use a 60Hz transformer, right? Practically everything uses high-frequency magnetics, a very small input capacitor and some form of switchmode power supply.
Honestly you will find the largest savings with fine-tuning HVAC, office/home lighting and industrial process control. All this little penny-ass stuff you're complaining about is going after the 5%, when you could attack the 80% far easier and achieve much greater savings with much less effort. The other 15% would be the next step and then looking to save a couple hundred miliwatts here and there.
I don't even want a kernel to have module loading enabled; a traditional full-custom kernel is faster and leaner.
You were doing alright until you threw that in there; it would have fooled the majority of the/. posters. Throw up some benchmarks to show everyone how much you're saving in ram and how much more efficient your kernel is. I'm afraid it's quite clearly you who are clueless. I'm going to hazard a guess that you run Gentoo and have at least 240 characters in your USE flags, claiming that it makes your system "teh most efficient system evar, lolz".
Timing issues are very nontrivial in telephony. Throw in a nondeterministic PCI bus and it gets even harier. Just because you believe something to be true does not mean that it is. Ask the professional audio guys how difficult it is to maintain timing on standard PC hardware. While the zaptel timing can be reworked and eventually eliminated for most cases, it was selected as a least path of resistance method of moving ahead.
Now please run along, I think your system's idle task could use some optimization.
Pile on the routine annoyances of Linux (the handful of commands necessary to connect to any AP) and you'll get frustrated quickly. No sugar coating; WLAN on Linux sucks.
I think you're on crack. I mean I use slackware and I don't have to type a damn thing. Plug in the card, DHCP takes over and I'm on the first AP I find. No typing whatsoever. Yes there are problems with Linux on the desktop but if your card is supported (and really if you buy a win32-only piece of shit that needs ndiswrapper that isn't Linux's fault) it is pretty damn seamless.
The 80386 was never a BPGA footprint. It was a regular 0.100" PGA, which no, is not impossible but still unbelievably difficult on a multilayer board where you have heavy copper pour for the ground and power planes.
Fast-forward to today, where the pin spacing has shrunk to sub-millimeter pitch and the number of pins has jumped by a factor of 7x. I have personally desoldered 0.5mm pitch TQFP devices and hand-soldered the same and it's not difficult with "prosumer" equipment (Weller iron with a superfine tip) but BGA/uBGA and the pinned versions are no, not realistically desolderable/resolderable.
I's possible to upgrade a soldered chip...just takes a soldering iron, a little skill, and a lot of paitence.
As someone who does this as part of his job -- you're full of shit. I would love to see you desolder a 900-and-some-odd pin chip (or are these uBGA?) without causing damage to the board, even with professional equipment. It's simply not worth the time or energy. If this thing is an actual pinned chip it's even more difficult than with a uBGA. And let's not forget the x-ray equipment required to verify that your soldering was done properly.
Not to mention that once you've got the chip off, you now have to solder another one on. High-density SMT rework is possible, but it's prohibitively expensive and done only out of necessity.
Slashdot Mirror
Yes, because GPS receivers are tiny and don't need a few minutes with a clear view of the sky to get a fix. And cell phone modules are tiny and work amazingly well without any kind of antenna. And all of this fits inside a tiny, functional USB drive...
The beacon is a far more plausible scenario, but now you've got to try and find it somewhere in the world. :-)
how to bluff a Texan out of a pot.
Huh?
The only problem I've ever had with Slackware is that it didn't always do as good a job at recognizing USB devices as some others.
My only real problem with Slackware is that Pat won't modify the base vim install to in the global vimrc. Damn that bugs the shit out of me. I've emailed him and he was very nice about it, he simply said that he does the bare minimum mangling of any package configuration and if I could convince the vim maintainer to make that change that Slackware'd have it. Now the ball's back in my court. Damnit.Offhand, why is this not a default config option for vim? The cursor keys are totally useless without it.
that's a bad idea swaret and slapt-get can bork your install very easily unless you know exactly what slapt-get does
I hear this again and again. While I can see it being dangerous for an upgrade, I have also heard the general "swaret will eat your system" under normal use. Do you have any proof to this end?
If it really bothers you then DON'T enter your login name and password, just click on the login button. You will be redirected to a secure page asking you to please enter you login infomation.
Actually that's exactly what I do -- If I don't see the browser bar go yellow I don't put any information in, regardless of whether it's through a secure form or not. I understand that the story's exploit would turn my browser bar yellow so I'd be lulled into a false sense of security to begin with, but my whole point was that having the login credentials on that page, secure form or not, is a poor idea.
If they want to save their CPU power for casual visitors, they should just have a "Log in now" link, like my bank does. It just seems to be a good idea, and follows the path of least surprise which is what every UI should strive for.
Paypal's main site (http://www.paypal.com) does *NOT* do a permanent redirect to https://www.paypal.com, so if you hit www.paypal.com you give your paypal login and password in the clear. I've emailed them several times on this and have finally given up, as they don't bother to respond.
So if you can get inbetween Paypal and your target, you don't even need to fool anybody.
Forget speakeasy bonded T1, you can bond your standard DSL lines through an OpenBSD firewall using CARP. Read also about VRRP and (HSRP and GLBP) for cisco solutions. They add not only redundancy but also load balancing, and recovery is real fast as opposed to something like RIP2. You can also use OSPF but careful, OpenOSPFD and zebra dont provide load balancing and redundancy of default routes. IOS does.
I recently did something similar with Linux.
Business wanted bigger bandwidth and higher availability than was available in town without going to expensive bandwidth. We took ADSL from two different providers (could have also used wireless, cable, etc.) and a machine in Toronto in a data centre on a multihomed network and created a pair of GRE tunnels between them.
The multihomed server extrudes an IP on its network for the business, and the ip-up scripts for the PPPoE connections fiddle with the tunnels. Now they have an IP address which is located in multihomed space in this little old town of 5300 people, and through creative use of iproute2 the only multihomed bandwidth they use is what they actually really need to be redundant. Everything else (basic web browsing and outgoing email being the biggest ones) is just coarsely load-balanced over the basic PPPoE connections, without travelling over the tunnels. I've got route caching turned off so it works pretty well.
The next step is to encrypt the tunnels using OpenS/WAN -- that was the original idea, but unfortunately it's not possible to do this without the GRE tunnels, since OpenS/WAN would get confused about which traffic was for which connection since they both have the same endpoints in the configuration file. I'm sure I could use another encryption package but I'd rather stick to something standard.
Me.
Although Unitrode will probably come in and squash that theory. :-)
Then, after you've tried everything, you discover that there's a special patched hacked up driver just for that card. What a pain in the ass.
Do you have any more info? I'm afraid my google-fu isn't quite as good as I'd hoped.
I had to get T.H.E. to replace my N64 controllers 3 times until i learnt to dismantle the joystick and fill it with vaseline to stop it wearing out after a month of use.
You filled your N64 joystick (I'm assuming the analog joystick, not the entire controller) with vaseline?
you do realize that the analog joysticks on those things are optical, right?
'm not faulting the engineers. I'm just saying that if these predicted problems were really as some people painted them (inevitable), then engineers would surely share some of the blame for not communicating properly if it was not fixed.
Engineer: "Don't point the gun at that clown and pull the trigger, because he will die."
Manager: "Don't be absurd, there is a chance I may not even hit him."
Engineer: "Don't point the gun at that clown and pull the trigger, because he will die."
Manager: "Thank you for your analysis of the situation. I've determined that the risk is worth the reward."
The engineer is to blame, how exactly? Managers go against the judgment of Those In The Know all the time, even effectively communicated. If the engineer says "We can't launch, the rocket will explode because of A and B and C, not to mention Z," and the manager decides to push ahead with the launch anyway, how did the engineer not effectively communicate this? You cannot lay blame on someone for someone else's stupidity or willful ignorance.
Sometimes, sure, an engineer may downplay a problem. THAT is something he can surely be blamed for. I was taking exception to your blanket statement where you said that if a project has inevitable problems and is pushed ahead anyway, that the engineers must share the blame (supposedly because Managers never go against advice given to them).
Something else to keep in mind : if I spotted a problem that would surely doom my project, and can't get engineering leadership/management to agree with me, I should share some of the blame.
I beg your pardon?
If I bring an issue up and explain its ramifications and make sure everyone is very much aware of what is at stake and my manager decides that the chance of failure is too small to worry about fixing the issue, how exactly should I be sharing the blame? I did my job; I documented that I did my job. I am clearly not in the wrong here.
If every ISP blocked outgoing SMTP messages from their users and either 1) forced them to relay mail through their servers or 2) ensured that any user-run mail servers were properly configured with SPF, etc. before allowing them to access outgoing port 25 traffic, it would allow allow much better assurance that the sender was who they said they were.
Nonsense. I'd say almost ALL of my spam these days comes from zombies who ARE using their ISP's SMTP server. Neither #1 or #2 fix that.
It worst habbit is turning off the phone for no apparent reason.
No, it's rebooting and not turning the phone back on. Get something like RadioState (it's free) to fix this. I have to admit though my 650 RARELY reboots now. It's still a pain when it does, but it's very, very, very infrequent.
Many publishing houses produce all their output with LaTeX, so saying it's useful just for math doesn't reflect its actual usage.
I'm trying to find some kind of latex front end that I can give to the people who make the manuals here... We've gone from Word to Ventura and dabbled a little in FrameMaker... nothing seems to do quite what we want, so I'm looking at latex. Lyx is a little weak here, and the other products I've found are either quite old (TrueTex) or geared heavily toward mathematical publication (Scientific Word). PCTex seems to be about the closest (write in word, mangle in PCTex, export how you want) but that seems to keep some of the biggest encumbrances I've got (people trying to do proper layout in Word).
Any suggestions? Our manuals are about 100 to 250 pages long, some diagrams, but mostly "Parameter Foo: values 1,2,3, meanings x,y,z, caution, warning, application hint, see Parameter Bar (page 50) and Baz (Page 90)" type of things. Some charts, and a display shot for each parameter... Nothing fancy, but lots of cross referencing.
Now if you guys can just make Thunderbird stop sucking, I'd be much happier.
Agreed. I can't compose messages in plain text? I can't have signature lines automatically removed when replying and quoting? I can't change the name of my outgoing account when composing? Crazy. Gimme kmail on Win32 and I'll be much happier.
It is time for a second Internet to come into action -- one that is voluntarily connected, one that is run over cabling (or satellite) connections that are not subsidized by any government regime.
L0pht Heavy Industries had quite a bit of information about this. They called it guerilla.net. It seems to be gone now, though. All 9600 baud ham, wifi and optical links... Very cool idea but the problem with it is the same problem that faces amateur radio these days. Lack of general interest.
It's sad. There is a lot of good that can come from this kind of thing, but people don't give a shit unless it has something to do with the latest fads on gossip on TV.
Exchange4Linux is what I've chosen, after having evaluated a dozen others, including SLOX, OpenOffice's equivalent, a few web-based ones, Hitachi's and even the one the Oracle bought out and closed down.
So no, there are plenty of Exchange replacements out there. For me, Exchange4Linux is the only one which actually replaces Exchange properly, and the server is 100% open source.
The treo650 I have allows you to have different rings for calls or SMS messages, and even different rings per caller-id (based on the number in the address book). I'm fairly certain the 600 does this too.
The other solution is to put in a very high-efficiency switching power supply, but those are very expensive.
If you're an EE, you're not a very good one.
There are switchers in TVs (most modern televisions run practically everything off of a small switcher and a larger switcher running all secondary supplies off of the flyback circuit) -- There hasn't ever been a linear supply in a PC power supply (even the old XTs ran switchers) and your DVD player, receiver, UPS... all switchers. Your basic switcher efficiency these days starts at about 80% under moderate load, with most achieving upward of 95% without even doing careful design. Now yes, while idle, they often drop to ~70% even when pulse-dropping but let's be honest here... 70% efficiency on pulling a hundred miliamps is not exactly killing the grid, nor is it adding up to any appreciable amount when multiplied by a million units, especially not compared to the HVAC and lighting requirements of any urban business centre, or the electrical requirements of even a couple of industrial processes.
To get back on topic, switchmode supplies are dirt cheap. The magnetics are smaller and lighter and cheaper than 60Hz transformers, the controllers are oftentimes cheaper than the ancient 78xx series regulators, the run cooler which means you save money on heat dissipation, you save weight because you don't need the same heat sinks... They are far cheaper than any old linear circuit for any appreciable load. Now you do run into higher output harmonic content with a switchmode power supply but simple filters and efficient switch design (already built in to most switchmode controller circuitry) minimize these effects, and linear supplies generate much different, and much more difficult to deal with input harmonic distortion.
In fact the benefits of switchers over linear supplies is so great that it's actually illegal to power any commercial equipment with linear supplies in the EU. You simply cannot achieve CE approval with a linear power supply due to the input current waveform distortion. Switchers spread out the input current wave over more of the voltage waveform, saving the power company from having to supply 3x-4x the current over a very narrow part of the voltage waveform.
Switchers are cheap. Hell even expensive switchmode supplies are cheap compared to the magnetics, semiconductors and heatsinking required for the equivalent power level linear supply, and that's not including weight and shipping considerations. I can't even think of a wall wart I have for a doo-dad in my house that doesn't use a switcher over a linear supply. Honestly, this is a non-argument, and yet I've managed to babble on about it for what, four paragraphs? :-)
You do realize that almost every single electronic device that consumes any appreciable amount of energy these days does not use a 60Hz transformer, right? Practically everything uses high-frequency magnetics, a very small input capacitor and some form of switchmode power supply.
Honestly you will find the largest savings with fine-tuning HVAC, office/home lighting and industrial process control. All this little penny-ass stuff you're complaining about is going after the 5%, when you could attack the 80% far easier and achieve much greater savings with much less effort. The other 15% would be the next step and then looking to save a couple hundred miliwatts here and there.
I don't even want a kernel to have module loading enabled; a traditional full-custom kernel is faster and leaner.
You were doing alright until you threw that in there; it would have fooled the majority of the /. posters. Throw up some benchmarks to show everyone how much you're saving in ram and how much more efficient your kernel is. I'm afraid it's quite clearly you who are clueless. I'm going to hazard a guess that you run Gentoo and have at least 240 characters in your USE flags, claiming that it makes your system "teh most efficient system evar, lolz".
Timing issues are very nontrivial in telephony. Throw in a nondeterministic PCI bus and it gets even harier. Just because you believe something to be true does not mean that it is. Ask the professional audio guys how difficult it is to maintain timing on standard PC hardware. While the zaptel timing can be reworked and eventually eliminated for most cases, it was selected as a least path of resistance method of moving ahead.
Now please run along, I think your system's idle task could use some optimization.
Pile on the routine annoyances of Linux (the handful of commands necessary to connect to any AP) and you'll get frustrated quickly. No sugar coating; WLAN on Linux sucks.
I think you're on crack. I mean I use slackware and I don't have to type a damn thing. Plug in the card, DHCP takes over and I'm on the first AP I find. No typing whatsoever. Yes there are problems with Linux on the desktop but if your card is supported (and really if you buy a win32-only piece of shit that needs ndiswrapper that isn't Linux's fault) it is pretty damn seamless.
The 80386 was never a BPGA footprint. It was a regular 0.100" PGA, which no, is not impossible but still unbelievably difficult on a multilayer board where you have heavy copper pour for the ground and power planes.
Fast-forward to today, where the pin spacing has shrunk to sub-millimeter pitch and the number of pins has jumped by a factor of 7x. I have personally desoldered 0.5mm pitch TQFP devices and hand-soldered the same and it's not difficult with "prosumer" equipment (Weller iron with a superfine tip) but BGA/uBGA and the pinned versions are no, not realistically desolderable/resolderable.
I's possible to upgrade a soldered chip...just takes a soldering iron, a little skill, and a lot of paitence.
As someone who does this as part of his job -- you're full of shit. I would love to see you desolder a 900-and-some-odd pin chip (or are these uBGA?) without causing damage to the board, even with professional equipment. It's simply not worth the time or energy. If this thing is an actual pinned chip it's even more difficult than with a uBGA. And let's not forget the x-ray equipment required to verify that your soldering was done properly.
Not to mention that once you've got the chip off, you now have to solder another one on. High-density SMT rework is possible, but it's prohibitively expensive and done only out of necessity.