1. MSN
2. MS Applications
3. MS Operating System 1
4. MS Operating System 2
5. MS Operating System 3
From then on, the three OS groups, which get identical snapshots of the code base, have to compete with each other, and deal with the software world, including the MS Apps group, as foreign entities. First one to go open-source and get compatible with Linux wins a lot of interesting markets (not to mention undercutting the closed-sourcedness of the other two).
The Apps group has to treat the browser as a software product, rather than an integral part of the revenue generator for its networking and OS divisions.
And MSN has to compete with the rest of the internet for the Apps and OS groups' support.
This of course would never happen, especially not under a Bush DoJ, but not even under Clinton's DoJ, because it makes perfect sense, and there's no way that something that makes perfect sense can be allowed to happen so long as Bill and Ballmer have money to spend.
These are the wanks who own the government contract to extort hundreds of dollars from anyone wishing a copy of a standard developed by the government.
They get rich selling the public its own property.
They can go pound sand.
--Blair
"NIST, on the other hand, is a national treasure."
Intel got out of the MHz business a long time ago when it introduced the iCOMP to show its chips growing performance linearly even though clock speeds weren't keeping pace (we're talking way back in the 50-MHz pentium days when AMD wasn't an issue and Intel was just trying to keep the press from spanking it).
iCOMP is a combination of several benchmarks and some hincky math (I've seen the formula--eeugh). Its saving grace is that it's applied uniformly to all parts. Marketing weasels can't get around standardization.
The fact is, one benchmark, be it "how fast does it tick?" to "how long does it take to decode the human genome?" is a worthless way to pick a computer. Look at all the benchmarks you can find. Understand the systemic issues in each test setup. Evaluate the use cases you will put to the computer. And if someone waves a benchmark at you and says "bias! this has bias!" say, "of course, it's biased towards the processor that runs that software faster."
--Blair
"If you don't run the software faster, you lose."
The top-end, newest-model units have always been several hundred dollars more than the next step down. And the 1.4 GHz Athlon is at least three steps down, being comparable to a 1.7 GHz P4, which is now behind the 1.8, 1.9, and 2.0.
I can see a lot of people finding value in that. Personally (yes) I could do with 4 GHz now, and would gladly pay $1k just for the CPU.
Porsche stays in business by not worrying what BMW is doing.
--Blair
P.S. I think you have it backwards. The current P4 is the Willamette design. The new one is the Northwood. And it's not a phase-out; it's a shrink and a bus upgrade. The Willamette price curve will continue like all of them. Northwood will scale up to 6 GHz, and semi-official hype (it was an Intel guy, in an interview) says 10 GHz. Brookdale is the i845 chipset, which will allow the Northwood to interface to SDRAM and DDR-SDRAM.
(Go to TomsHardware.com and search on "intel roadmap"; I'd post a link, but the net is totally packed up right now...)
Okay, my bad, fair enough, looked like you were being bench-selective, I'll take your word that you weren't.
Is it worth the $400?
Not if you're an Unreal nut, no.
But over on Tom's Hardware, almost all the benchmarks other than UT go to the P4.
There's one from SiS about memory bandwidth that I don't trust that shows every P4 with nearly a 2X advantage on any Athlon, but there it is. Maybe it isolates the CPU and just demonstrates the point behind RDRAM (which is also getting cheaper).
Is it still worth $400?
I have been first-day-of-issue adopter of a CPU or two, when I saw the same system two months later on the second tier and for $400 less, I knew that I'd had the nuts for those two months, and still owned a computer that would be nonobsolete for a year, maybe two.
I won't upgrade my two-month-old 1.8GHz platform to 2.0GHz, especially when clock increases are not 1-for-1 with performance increases.
But the upgrade sweet spot is an 18-30 month cycle.
Pro: I have 2 y.o. 400MHz iron running on my desk at home. I can upgrade without shame.
Con: If I buy now, how do I choose between 1.8 or 2.0?* The difference in system price is a few hundred dollars.
Pro: I'm rich, and have a big ego.
You do the math.
--Blair
* - actually, for other reasons, I have no reason to UG that DT until xMas or so. Rumor is we may have 4GHz by then. Crazy rumor, yes, but most promising for the 2.4-2.5GHz probabilities. We'll also know if DDR on the i845 chipset is faster or slower than RDRAM on the i850. Those aren't my reasons for delaying, but they're predictable benefits.
Right. The client initiates the transaction, using the bank's "well-known" public key. The man in the middle can't decrypt that to get the random cookie to send back to the client, so he can't mimic the server.
The only spoof possible now is if the card is issued by the man in the middle. It's still vulnerable to physical compromise and cracking the private keys.
(BTW, when I called the card a crypto engine, I meant that the private key never leaves the card; it wouldn't be secure if it could; how you keep someone from reverse-engineering it off the chip is another story. I like the thing about the PIN, though; it improves over simple physical security. The on-card key generator will work iff the old key is used to perform the new-key-registration transaction.)
--Blair
"I can't tell what's better: pegging my karma at 50, or making people knock it down so I can peg it again..."
You can always find a particular benchmark that makes your desired result occur.
There are benchmarks where 1.2GHz Athlons and P4's beat 1.4GHz Athlons and 1.7GHz P4's.
A benchmark can't be biased. Either you run the piece of software faster or you don't. But selection of benchmarks can be biased. And other value-determining factors can get pulled into the evaluations that are supposed to be made solely on benchmarks.
If all you care about is Unreal Tournament, then you've found your answer. But using that to make an overgeneralized statement like "AMD has the better chip" means you're probably lying to everyone else.
--Blair
Price/Performanc vs. Raw Performance
on
Pentium IV Hits 2 Ghz
·
· Score: 3, Insightful
Yes. $ for $ the AMD chips win. But you need a computer engineering degree to understand why. Consumers still measure Sony TV's horizontally to determine if they're 27 or 35 inches (try it! Sony makes them that way so they don't have to educate the public).
However, the 1.4 GHz Athlon with DDR SDRAM was about par on the benchmarks with the 1.7 GHz P4 with RDRAM.
1. You can't get 1.5 GHz Athlons yet, and the P4 has gone on to 1.8, 1.9, and 2.0 GHz.
2. Intel and VIA are releasing motherboards that will run DDR SDRAM, reducing memory cost significantly with an unknown but predicted to be very small performance hit vs. RDRAM.
Ergo, if you want the fastest commercial desktop, you buy the newest P4 platform. And the early adopters, speed queens, and obsolescence anxiety victims have always justified exhorbitant price differentials.
Businesswise, Intel made a bad, bad mistake putting all its chips in the Rambus basket. AMD was also able to leverage some serendipity when Digital went belly-up, leaving a lot of Alpha engineers with nowhere else they could stomach to go. But Intel has been through this before (remember the PowerPC? Apple, Motorola, and IBM combined are about 40x the size of AMD, and they couldn't take Intel...) and has already reposition itself.
Intel can be bloodied, but it's never been knocked down, much less knocked out.
Am I cheerleading? Maybe a little. I own a ton of INTC. But I have always known they make inferior products. 6502, m68k, Alpha, PowerPC, even Intel's own i960 line are superior products to any chip that eats x86 assembly. But if you get prejudiced on the characteristics of a product you will totally fail to understand the value of the company.
Intel will rule in the end. Start from that premise, and then try to prove otherwise to yourself.
--Blair
"It's not an 800 lb gorilla. It's an 800 lb gorilla with a PhD in process technology and 30 Superbowl rings."
Actually, I was clarifying a glaring misapprehension in the media.
The fact that I almost got First Post is because: (1) it's exceptionally glaring, so it didn't take long to devise or express; and, (2) Slashdot was hiccupping that day.
The story had been online a good minute and a half before I sent my response. Time for 12-15 actual Firts Psots on a normal day.
I didn't even expect I'd be first-in on the subject of crystals as big, repetitive molecules...
--Blair
"You're faceless and your mother karmas you funny."
Desktop operating systems are designed for configurability, not zippy boots for HA. It has reams of code that's only there to make reconfiguration easier. You can pre-set a desktop kernel with its fixed peripheral configurations, but there are embedded OSes designed for that.
Linux would only be repurposed to do this sort of thing because it's open source, and because hackers understand its operation. That doesn't make it the right tool for the job.
If the circuit on the smart card can be used as a public-key crypto engine, then you could use it to secure any interaction with the card issuer's database.
Nobody could get your private key unless they stole your physical card, since there's no need to have the key printed anywhere except in the card's circuit.
Here's the loop: Client (cardholder) sends server (issuer) a cookie encoded with Server's public key. Server decrypts it with its private key and sends it back along with its own cookie, encrypted with Client's public key. Client decrypts, compares the Client cookie it sent with its copy of it, thus validating Server's authority. Client then encrypts Server's cookie and sends it back. Server decrypts, compares with its copy, thus validating the client's authority. This is basic RSA/PGP stuff.
One simple handshake--it's about as complicated as the TCP/IP connection that was made to transport it--and your SmartCard is money.
This gets rid of the current problem of credit-card numbers being stolen ex proprio that arises because you have to copy the number itself off the card in order to use it.
--Blair
"I was speculating about the meaning of ex proprio, too. So sue me."
Diamond is not an ionic compound. It is composed of carbon atoms in the optimal arrangement to form covalent bonds. Tell me diamond isn't a crystal.
Bonds are rarely 100% covalent or 100% ionic. A crystal is a single molecule (but not all single molecules are crystals). These facts are old. Very old. Older than your textbooks. Shame on your school.
I won't get into the semantic argument about solids, liquids, gases, and how any of them can be said to form or be formed by crystals, because that would only confuse you (and because in the more intricate cases I'm bound to forget the details and my book is in another state). Just trust that the definition of "crystal" that you are using here is very inadequate.
Go search on a few things:
Ionic Bond
Covalent Bond
Ionic Character
_General Chemistry_, by Linus Pauling*
--Blair
"We teach chemistry like it's either a foreign language about a dead religion or a way to make the neighborhood kids think we're cool."
P.S. I'd like to thank the Academy for down-modding my original post. It's always nice to see that the forces of intolerant ignorance continue to crawl the planet. It keeps an exterminator of such things in poker money.
Meanwhile, your name insults you, you have no karma, and your posts are moderated down by default.
You do the math.
--Blair
"Who loves ya baby."
* - They used to be first, but people stopped buying cell-phones, and the whole transistorized-crap market went south from there. Now the money's in gasoline and baby clothes.
Moore's law will hold for quite a long time, inasmuch as it's already been crocked by adapting it to apply to microprocessor computing power, when it was originally developed to describe memory-chip bit capacity.
Once it starts to break down for silicon-transistor circuits, the "capacity" metrics will be transferred to whatever follows.
The interesting thing about Moore's law is that it may be unprovably vague.
Einstein posed a theorem he said he never could prove:
If you travel from point A to point B at an average speed of v miles per hour (where B and A are more than v miles apart), there will always be an interval exactly v miles long that you will transit in exactly one hour.
How this relates to Moore's law is if you replace distance with transistor count, then along the way you will find intervals where you have doubled the transistor count in 18-24 months.
This feature allows hypesters every once in a while to prove to themselves that "it still works" to whatever precision they desire.
But they're not entirely dishonest, since this only works because Moore's law has a long-term stable average.
Companies like General Motors or Boeing must abide by safety and quality standards, while a Microsoft doesn't, even though it's products may or may not have more of an impact on daily lives and safety than cars by GM or planes from Boeing.
Much as I hate defending the nine-headed demon, that's not Bill Gates' fault, and it's not his problem.
Anyone who relies on uncertified software in a safety-critical situation is running with scissors. It's their own fault if something goes wrong.
In fact, if you read the documentation of the certification process, you realize that it's not directly about making the software safer, it's about indemnifying the government (and the company somewhat) against liability for known and unknown bugs. (Known ones are documented and the conditions for their occurrence are defeatured. It's then the product operator's fault if that condition occurs. Unknown ones are mitigated by the "well, we tried" defense. Your lawyer's mileage may vary.)
Bill's warranty tells you outright that he's not responsible for your bad decisions, and that part is enforceable, because he really doesn't have to say it at all.
Slashdot Mirror
So what you're saying is,
There's nothing more expensive than free software.
--Blair
"Except maybe an equity position in a can't-miss internet opportunity."
Which is why you break it into FIVE companies:
1. MSN
2. MS Applications
3. MS Operating System 1
4. MS Operating System 2
5. MS Operating System 3
From then on, the three OS groups, which get identical snapshots of the code base, have to compete with each other, and deal with the software world, including the MS Apps group, as foreign entities. First one to go open-source and get compatible with Linux wins a lot of interesting markets (not to mention undercutting the closed-sourcedness of the other two).
The Apps group has to treat the browser as a software product, rather than an integral part of the revenue generator for its networking and OS divisions.
And MSN has to compete with the rest of the internet for the Apps and OS groups' support.
This of course would never happen, especially not under a Bush DoJ, but not even under Clinton's DoJ, because it makes perfect sense, and there's no way that something that makes perfect sense can be allowed to happen so long as Bill and Ballmer have money to spend.
--Blair
These are the wanks who own the government contract to extort hundreds of dollars from anyone wishing a copy of a standard developed by the government.
They get rich selling the public its own property.
They can go pound sand.
--Blair
"NIST, on the other hand, is a national treasure."
Intel got out of the MHz business a long time ago when it introduced the iCOMP to show its chips growing performance linearly even though clock speeds weren't keeping pace (we're talking way back in the 50-MHz pentium days when AMD wasn't an issue and Intel was just trying to keep the press from spanking it).
iCOMP is a combination of several benchmarks and some hincky math (I've seen the formula--eeugh). Its saving grace is that it's applied uniformly to all parts. Marketing weasels can't get around standardization.
The fact is, one benchmark, be it "how fast does it tick?" to "how long does it take to decode the human genome?" is a worthless way to pick a computer. Look at all the benchmarks you can find. Understand the systemic issues in each test setup. Evaluate the use cases you will put to the computer. And if someone waves a benchmark at you and says "bias! this has bias!" say, "of course, it's biased towards the processor that runs that software faster."
--Blair
"If you don't run the software faster, you lose."
Man.
If that's not enough of a reason to love these guys, I don't know what.
The Logitech CyberMan II was the king-hell perfecto par excellence of game controllers. And it was gorgeous and sexy.
But of course, nobody bought it so nobody developed for it so no body bought it...
It will come back. The world continues to spin, and lost clues are merely buried to be discovered anew when we rebuild our civilization.
--Blair
What exact amplitude do they call "imperceptible"?
A few mm? A few um? A few nm?
--Blair
"Gotta love those irreproducible results."
The top-end, newest-model units have always been several hundred dollars more than the next step down. And the 1.4 GHz Athlon is at least three steps down, being comparable to a 1.7 GHz P4, which is now behind the 1.8, 1.9, and 2.0.
I can see a lot of people finding value in that. Personally (yes) I could do with 4 GHz now, and would gladly pay $1k just for the CPU.
Porsche stays in business by not worrying what BMW is doing.
--Blair
P.S. I think you have it backwards. The current P4 is the Willamette design. The new one is the Northwood. And it's not a phase-out; it's a shrink and a bus upgrade. The Willamette price curve will continue like all of them. Northwood will scale up to 6 GHz, and semi-official hype (it was an Intel guy, in an interview) says 10 GHz. Brookdale is the i845 chipset, which will allow the Northwood to interface to SDRAM and DDR-SDRAM.
(Go to TomsHardware.com and search on "intel roadmap"; I'd post a link, but the net is totally packed up right now...)
How's that Java thing going for you?
--Blair
Okay, my bad, fair enough, looked like you were being bench-selective, I'll take your word that you weren't.
Is it worth the $400?
Not if you're an Unreal nut, no.
But over on Tom's Hardware, almost all the benchmarks other than UT go to the P4.
There's one from SiS about memory bandwidth that I don't trust that shows every P4 with nearly a 2X advantage on any Athlon, but there it is. Maybe it isolates the CPU and just demonstrates the point behind RDRAM (which is also getting cheaper).
Is it still worth $400?
I have been first-day-of-issue adopter of a CPU or two, when I saw the same system two months later on the second tier and for $400 less, I knew that I'd had the nuts for those two months, and still owned a computer that would be nonobsolete for a year, maybe two.
$400 ain't that much for that kind of egoboo.
--Blair
Incremental improvement is the name of the game.
I won't upgrade my two-month-old 1.8GHz platform to 2.0GHz, especially when clock increases are not 1-for-1 with performance increases.
But the upgrade sweet spot is an 18-30 month cycle.
Pro: I have 2 y.o. 400MHz iron running on my desk at home. I can upgrade without shame.
Con: If I buy now, how do I choose between 1.8 or 2.0?* The difference in system price is a few hundred dollars.
Pro: I'm rich, and have a big ego.
You do the math.
--Blair
* - actually, for other reasons, I have no reason to UG that DT until xMas or so. Rumor is we may have 4GHz by then. Crazy rumor, yes, but most promising for the 2.4-2.5GHz probabilities. We'll also know if DDR on the i845 chipset is faster or slower than RDRAM on the i850. Those aren't my reasons for delaying, but they're predictable benefits.
Right. The client initiates the transaction, using the bank's "well-known" public key. The man in the middle can't decrypt that to get the random cookie to send back to the client, so he can't mimic the server.
The only spoof possible now is if the card is issued by the man in the middle. It's still vulnerable to physical compromise and cracking the private keys.
(BTW, when I called the card a crypto engine, I meant that the private key never leaves the card; it wouldn't be secure if it could; how you keep someone from reverse-engineering it off the chip is another story. I like the thing about the PIN, though; it improves over simple physical security. The on-card key generator will work iff the old key is used to perform the new-key-registration transaction.)
--Blair
"I can't tell what's better: pegging my karma at 50, or making people knock it down so I can peg it again..."
That you would ask that question in this context makes it pretty clear that you don't.
IOW, don't mistake my populist renaming of the ports as inexperience.
--Blair
This sort of joke is so old it could only be funny to adolescents who hadn't heard one like it before.
Slashdot is showing its demographic.
--Blair
"No wonder I got so much karma..."
You can always find a particular benchmark that makes your desired result occur.
There are benchmarks where 1.2GHz Athlons and P4's beat 1.4GHz Athlons and 1.7GHz P4's.
A benchmark can't be biased. Either you run the piece of software faster or you don't. But selection of benchmarks can be biased. And other value-determining factors can get pulled into the evaluations that are supposed to be made solely on benchmarks.
If all you care about is Unreal Tournament, then you've found your answer. But using that to make an overgeneralized statement like "AMD has the better chip" means you're probably lying to everyone else.
--Blair
Yes. $ for $ the AMD chips win. But you need a computer engineering degree to understand why. Consumers still measure Sony TV's horizontally to determine if they're 27 or 35 inches (try it! Sony makes them that way so they don't have to educate the public).
However, the 1.4 GHz Athlon with DDR SDRAM was about par on the benchmarks with the 1.7 GHz P4 with RDRAM.
1. You can't get 1.5 GHz Athlons yet, and the P4 has gone on to 1.8, 1.9, and 2.0 GHz.
2. Intel and VIA are releasing motherboards that will run DDR SDRAM, reducing memory cost significantly with an unknown but predicted to be very small performance hit vs. RDRAM.
Ergo, if you want the fastest commercial desktop, you buy the newest P4 platform. And the early adopters, speed queens, and obsolescence anxiety victims have always justified exhorbitant price differentials.
Businesswise, Intel made a bad, bad mistake putting all its chips in the Rambus basket. AMD was also able to leverage some serendipity when Digital went belly-up, leaving a lot of Alpha engineers with nowhere else they could stomach to go. But Intel has been through this before (remember the PowerPC? Apple, Motorola, and IBM combined are about 40x the size of AMD, and they couldn't take Intel...) and has already reposition itself.
Intel can be bloodied, but it's never been knocked down, much less knocked out.
Am I cheerleading? Maybe a little. I own a ton of INTC. But I have always known they make inferior products. 6502, m68k, Alpha, PowerPC, even Intel's own i960 line are superior products to any chip that eats x86 assembly. But if you get prejudiced on the characteristics of a product you will totally fail to understand the value of the company.
Intel will rule in the end. Start from that premise, and then try to prove otherwise to yourself.
--Blair
"It's not an 800 lb gorilla. It's an 800 lb gorilla with a PhD in process technology and 30 Superbowl rings."
Yeah.
It only works if you color the rim with green magic marker.
--Blair
"Peace. Out."
Actually, I was clarifying a glaring misapprehension in the media.
The fact that I almost got First Post is because: (1) it's exceptionally glaring, so it didn't take long to devise or express; and, (2) Slashdot was hiccupping that day.
The story had been online a good minute and a half before I sent my response. Time for 12-15 actual Firts Psots on a normal day.
I didn't even expect I'd be first-in on the subject of crystals as big, repetitive molecules...
--Blair
"You're faceless and your mother karmas you funny."
Desktop operating systems are designed for configurability, not zippy boots for HA. It has reams of code that's only there to make reconfiguration easier. You can pre-set a desktop kernel with its fixed peripheral configurations, but there are embedded OSes designed for that.
Linux would only be repurposed to do this sort of thing because it's open source, and because hackers understand its operation. That doesn't make it the right tool for the job.
--Blair
If the circuit on the smart card can be used as a public-key crypto engine, then you could use it to secure any interaction with the card issuer's database.
Nobody could get your private key unless they stole your physical card, since there's no need to have the key printed anywhere except in the card's circuit.
Here's the loop: Client (cardholder) sends server (issuer) a cookie encoded with Server's public key. Server decrypts it with its private key and sends it back along with its own cookie, encrypted with Client's public key. Client decrypts, compares the Client cookie it sent with its copy of it, thus validating Server's authority. Client then encrypts Server's cookie and sends it back. Server decrypts, compares with its copy, thus validating the client's authority. This is basic RSA/PGP stuff.
One simple handshake--it's about as complicated as the TCP/IP connection that was made to transport it--and your SmartCard is money.
This gets rid of the current problem of credit-card numbers being stolen ex proprio that arises because you have to copy the number itself off the card in order to use it.
--Blair
"I was speculating about the meaning of ex proprio, too. So sue me."
They don't know what they mean.
A transistor either switches a current path off (output_on = power_on AND NOT control_on) or it switches it on (output_on = power_on AND control_on).
Those are the real building blocks. Larger structures like gates and flip-flops are combinations of those two facts.
Some circuits use multi-leveled logic, but those have to be converted to boolean logic* before they can get anywhere near your computer.
--Blair
* - or whatever passes for it at the NY Times...
Your education is shallow and misled.
Diamond is not an ionic compound. It is composed of carbon atoms in the optimal arrangement to form covalent bonds. Tell me diamond isn't a crystal.
Bonds are rarely 100% covalent or 100% ionic. A crystal is a single molecule (but not all single molecules are crystals). These facts are old. Very old. Older than your textbooks. Shame on your school.
I won't get into the semantic argument about solids, liquids, gases, and how any of them can be said to form or be formed by crystals, because that would only confuse you (and because in the more intricate cases I'm bound to forget the details and my book is in another state). Just trust that the definition of "crystal" that you are using here is very inadequate.
Go search on a few things:
Ionic Bond
Covalent Bond
Ionic Character
_General Chemistry_, by Linus Pauling*
--Blair
"We teach chemistry like it's either a foreign language about a dead religion or a way to make the neighborhood kids think we're cool."
P.S. I'd like to thank the Academy for down-modding my original post. It's always nice to see that the forces of intolerant ignorance continue to crawl the planet. It keeps an exterminator of such things in poker money.
* - the Dover 1989 reprint of the 1954 edition only costs like $14. How much did you pay for a semester of your college's misapprehensions?
Gordon Moore has several billion dollars and is Chairman Emeritus of one of the fifth or sixth biggest corporation in the world*.
Meanwhile, your name insults you, you have no karma, and your posts are moderated down by default.
You do the math.
--Blair
"Who loves ya baby."
* - They used to be first, but people stopped buying cell-phones, and the whole transistorized-crap market went south from there. Now the money's in gasoline and baby clothes.
Once it starts to break down for silicon-transistor circuits, the "capacity" metrics will be transferred to whatever follows.
The interesting thing about Moore's law is that it may be unprovably vague.
Einstein posed a theorem he said he never could prove:
How this relates to Moore's law is if you replace distance with transistor count, then along the way you will find intervals where you have doubled the transistor count in 18-24 months.
This feature allows hypesters every once in a while to prove to themselves that "it still works" to whatever precision they desire.
But they're not entirely dishonest, since this only works because Moore's law has a long-term stable average.
--Blair
Geez. Didn't you guys learn anything in chemistry?
A crystal is a single molecule. A transistor is a single molecular structure. It won't work any other way.
--Blair
Companies like General Motors or Boeing must abide by safety and quality standards, while a Microsoft doesn't, even though it's products may or may not have more of an impact on daily lives and safety than cars by GM or planes from Boeing.
Much as I hate defending the nine-headed demon, that's not Bill Gates' fault, and it's not his problem.
Anyone who relies on uncertified software in a safety-critical situation is running with scissors. It's their own fault if something goes wrong.
In fact, if you read the documentation of the certification process, you realize that it's not directly about making the software safer, it's about indemnifying the government (and the company somewhat) against liability for known and unknown bugs. (Known ones are documented and the conditions for their occurrence are defeatured. It's then the product operator's fault if that condition occurs. Unknown ones are mitigated by the "well, we tried" defense. Your lawyer's mileage may vary.)
Bill's warranty tells you outright that he's not responsible for your bad decisions, and that part is enforceable, because he really doesn't have to say it at all.
--Blair
"Great. And I have to fly next week..."