In order to stop spam, we need to stop using SMTP and switch to a protocol that rejects mail by default.
I wasn't aware that SMTP was incompatible with whitelisting. In fact, I'm pretty sure I've heard of setups that do just that, result in an SMTP server that rejects connections from people it doesn't know.
There's a trivial technological means to fight spam. It just requires abandoning SMTP and moving to a new protocol with the following requirements.
All compliant mail transport daemons must require all connections from client computers to be authenticated.
All compliant mail transport daemons must sign all messages as they pass them along.
All compliant mail transport daemons must have a service record in DNS for their host name that provides a public key for verification of the signature.
All compliant mail transport daemons must refuse to accept any email if the signature cannot be verified immediately (even if this is due to load), forcing the sending end to retry.
All compliant mail transport daemons must refuse to accept any email if the host name does not resolve to the IP number from which the inbound message was received.
You forgot one:
All relevant DNS servers must implement DNSSEC.
With that, spam is basically dead. As soon as you require those restrictions, suddenly spammers have to actually own a domain name and provide a working DNS server in order to deliver spam, and that DNS server must contain up-to-date mappings for those hosts to IP numbers. That pretty much obliterates the use of zombies for delivering mail.
Unless they can 0wn a DNS server, or have the zombies send through the owner's legitimate outbound email accounts, or can get a steady supply of disposable domains somewhere (zombie-XXXXXX.disposable-20081217.com, etc).
It also means that there is now a domain name, which by ICANN policy, is required to have a valid postal address, phone number, and other contact information associated with it.
And when the spammers don't follow the policy? Sure the domains might get shut down after someone realized (and got the registrar to verify) that the contact info was bogus, but that's a bit too late.
If Google were more libertarian than liberal, I would expect them to be proposing a referendum in California to sweep away all of the franchising laws so that there are no local or state limits on who can enter what Internet or TV market.
Part of the logic behind franchising laws is that they give more revenue to local governments, but so what?
Might be kinda difficult right now, isn't California (among many others) having some budged difficulties?
Just look at what great shape our economy is in...
That's partly a case of people (well, banks) not understanding what they were buying, and I think partly a case of the nation approaching its collective credit limit. Different (not more, different) regulations would have helped with the former, and the government not price-fixing the credit markets would probably have helped with the latter.
It's good that you bring that on, because it also illustrates a security problem: You can not wipe your drive anymore... All your data will be readable by anyone.
Better to just encrypt everything from the start, wiping it only works if you know in advance that your drive is about to change hands. The key can be a boot password or a hardware dongle (USB stick, SD card, etc) or something.
What would bother me more than the high price is the limited number of writes.
Sure, there are ways to limit writes to the disk, like disable swapping and delaying writes whenever possible, but I would still rather go with a reliable HDD over a SDD.
I never dropped one of my notebooks until today, but then again I never had one that looked like a toy...
Based on what I've heard an SSD is actually much more reliable, even if the lifetime is a bit shorter than the average lifetime for an HDD (which I don't think it necessarily is, especially if you get one of the SLC (faster write) drives).
The question should be "is this the year that SSDs will be price competitive with hard drives?" Until that day comes, SSDs will only sell in small quantities.
Some of them are (might be just the Intel ones?). People just don't know this, because "I/Os per second" isn't one of the commonly given numbers like read/write speeds and capacity are. (Probably because most people don't know why it matters.)
4) Energy security (not sure what this means, sounds like nationalism)
5) Water supply (not sure what this is about, obviously, if water is scarce, it would quickly become very expensive, leading to the power plant using it to close down, assuming a free-market environment)
I do know that there were a couple cases where nuclear plants had to shut down due to drought conditions during summer when people really need electricity for air conditioning (because the rivers they were taking cooling water from started to run low). Not sure if this is what that's about or not, it should really apply to pretty much any highly concentrated power plant.
I"m wondering if their evaluation of nukes..was based on the current 'laws and regulations' in the US (encacted by Carter I think?), that pretty much prohibit things like breeder reactors, that 'can' be used to manufacture weapons grade stuff, but, also can allow the fuel to be used much more efficiently, leaving much less waste than the first run we currently do?
From my limited understanding, if we repealed those laws...we could really stretch the nuclear fuel in a massive way, and have much, much less radioactive waste to have to manage, that has a much lower half life, etc.
THey do assume that. (that's the "HTML article" in the second story link, in case they don't like direct links like this).
They also try to calculate how much use of nuclear electric plants would increase the chance of a nuclear war (by giving more groups access to various nuclear technologies), and the environmental impact such a war might have.
This turns out to be rather insignificant, at least as far a carbon emissions are concerned (table 3). "Lifecycle" emissions for nuclear are "9-70" which is about equivalent to Solar PV or Geothermal, and somewhat worse than the 10-20-ish range of most others. Wind is significantly better at <10. They also calculate an "Opportunity cost", which they have much higher for nuclear because it takes so long to build (partly regulatory issues again).
I think centralization/concentration also works out as a negative, even if not counted directly.
Hydro plants scored about the same as nuclear, only wind/sun/ocean powered systems came out ahead.
I thought it was the parents who were "violated", by not getting the required assistance in keeping track of what their children are doing online (because putting the computer where they can see it is too hard)?
Now, in BOTH cases, you lose your data. But in one case you lose anyone else's data AND the OS, and in the other one, only your data.
This only matters on a shared system, where the different users actually use different accounts and set nontrivial passwords. Any envioronment with users clueful enough to do this, or administrators clueful enough to enforce this, will also have users clueful enough to avoid trojans or administrators clueful enough to secure their systems and make regular backups.
Plus I think the default umask is world-readable, so any one user running a spyware will in fact compromise everyone's data. Malware these days isn't so intent on causing corruption, because that isn't profitable.
But the one where you lose ALL security is the better one????
WTF are you talking about? The only "better" security model I mentioned is per-application access limitations, which I do not believe any mainstream system offers in a usable manner, and which will still be circumvented by a sufficiently motivated (chance of free porn or cute puppies) idiot.
Microsoft just spent $9 billion and many years to create Vista, so it does not sound reasonable that some new alternative could just snap into existence overnight like that. IBM tried, and spent a huge amount of money developing OS/2 but could never keep up with Windows.
And since the malware can't infect at the system level it is then a simple matter for an administrator to nix the offending files?
Is "administrator" a fancy term for "geeky neighbor kid"?
The only files that matter are the user's files, everything else can be fixed with apt-get and a livecd. If those files are messed up, it does not matter that the stock OS files are still intact. The *nix security model is good for protecting users from eachother, while malware protection requires protecting users from themselves. Probably the only ways to get the latter are some unmaintainable SELinux config or a highly inconvenient browser-in-a-VM and email-in-a-different-VM setup, and even those can't ever be idiotproof.
But to be quite frank, I lost all faith in any open source business application after I tried using Zen Cart. Those guys fucking deleted the *god damn row from the database* when you removed an order!!! Now I ain't no lawyer or professional accountant, but I do know you *never remove anything from the database unless you are planing some scheme to commit fraud!*. Hide it, yes, but you don't delete the fucking thing forever!
Materials? We're talking about web development here, which should be a pure SERVICE. If you bill for any kind of materials, you are considered a retailer/reseller and have to deal with specific sales tax, B&O tax, recycling & disposal fees, storage and transport laws etc related to physical goods in cities and states that have them. Also, very few people live in neighborhoods zoned for business so you can only do limited services when working out of your home. You generally can't warehouse materials without running afoul of local zoning ordinances.
Make the customer buy their own materials or have them pay you for them separately as an nontaxable reimbursement. If someone needs to buy software or hardware to do the job, YOU buy it and bill them for using it by the hour - that's "time and equipment" and it's how you bill when you are a pure service contractor. You get to amortize the price of the equipment over one or more years and you don't have to deal with most of the stuff that physical contractors have to put up with.
That might not work, at least if the hosting provider is in the same state as you...
Generally, all activity performed on computer software is subject to [Tennessee] state and local sales tax. These activities may include creation, sale, lease, rental, license to use, installation, repair, programming, modification, upgrading, or other activities. Installation of tangible personal property, even in labor-only contracts when the customer provides the materials, machinery, and equipment that is being installed, is subject to sales tax.
Additionally, web site development and design are taxable under state law. If a web site is created in Tennessee for hosting outside of Tennessee, the sale is in interstate commerce and not subject to sales tax. However, if a web site is hosted in Tennessee, the sale is creation of tangible personal property and subject to sales tax.
While I'm aware that machine code can be optimized slightly better than C code, no one programs in it. (Maybe only electrical engineers when working on very specific types of embedded systems.)
I've seen multiple x86 assembler versions of SHA-1 in the same program, optimized for different processor families (old Intel vs old AMD vs things with SSE2). There's also a tiny notepad replacement that some crazy person wrote entirely in assembly ("grown up notepad" or something, I think). I also know that a few small parts of the Linux kernel must be done is assembly, but AIUI this is more often due to "cannot be expressed in cross-platform languages" rather than performance.
Most software engineering techniques rely heavily upon testing to provide quality assurances.
I'm not sure such a lack of proper design really counts an "engineering".
With parallel algorithms, testing provides almost not guarantees because of the combinatorial explosion in the test cases that you will need to test all the timing paths.
That means that you're doing it wrong. It should at all times be clear what data is owned by what thread. Race conditions should be designed out and the code reviewed for adherence to the design, or you could use one of the languages that don't have a way to express race conditions.
And no, engineers don't deal with stuff like that all the time because if they did they simple would fail get the job done correctly.
And because they're legally required to know what they're doing.
So a quick question before I go and master FP...does the compiler automatically compile the code that can be done in parallel in the proper "way", or do I have to specify something?
I would imagine that this depends on the language and compiler you use.
Also, if I rewrote an app written in an imperative language to a FP one like Haskell, would I see a that much of a difference on a multi-core processor?
That depends on a great many things, including the quality of the compilers, your skill at writing code into the different languages, and the nature of the particular app.
"Only drawback is they need to be supercooled, something that may be addressed by improving the materials used." - that last part is a bit of an understatement. We're still decades (centuries?) away from room temperature superconductors.
Even getting it up to liquid nitrogen temperatures would probably be "good enough" for non-portable uses. I'm pretty sure that's a heck of a lot cheaper than liquid helium.
I'm posting anonymously because I don't want to have people at my company know who I am.
And also because you're trolling?
But it seems to me that Linux while cheap to buy is not cheap to keep patched and secure, particularly in a fleet of inhomogeneous platforms and users and network,printer, or disk sharing conditions in different buildings and subnets.
So don't make it so heterogeneous, then.
But it takes effort, dicsipline and an above avegage IT guy. And if you lose that person, you are screwed. Even a new equally skilled guy probably can't get all the scripts and stuff the last guy used to manage to work.
Effort and discipline, yes, but doing anything well requires those. But if the new guy can't get things to work it means someone wasn't required to document things or use a versioning repository. I'd also question the "above average" requirement, unless you're saying that "average" means "insists on known-bad practices even when told otherwise".
With windows, you can take a balow average imbecile, get them through a certification course, and they become almost interchangable monkeys. you need a lot of them since you will constantly be fighting fires or hunting down the right driver for the given brand of computer, but they can do it and it will work.
Or you can take someone who knows what they're doing, and things will work about as well as your Linux system does with people who mostly know what they're doing.
Moreover, and this is the critical part, a manager who is not an expert can tell if his monkies are keeping up with patches. MS tells him what he need to do. With Linux you can't really tell if the IT guy is doing it all, or if your pants are around your ankles.
[citation needed]
So it's not enough to use Linux to reduce TCO. you need to have a company like IBM telling you how to manage your configuration. Not because a skillful IT can't. But because a manager will know that IBM has his back.
No, you just need to treat your config files and admin scripts like any other programming project and demand good practices like source control and adequate documentation.
saddly a mediocre virus prone Windows network is, to a manager, much easier to sleep at night, than a well run Linux system that's tight as a ducks Ass, simply because he knows it's reasonably safe from an industry standard point of view.
[citation needed]
people will trade, extremes (linux) for mediocre, if they can limit thier risks.
People would rather flee the unknown and stay in their comfort zone, instead of learning about something new. With hard times forcing people to make an effort, maybe they'll find out that Linux isn't as scary and risky as they've been told.
So yes, we are probably seeing the beginning of the end of performance gains using general-purpose CPU interconnects and have to go back to vector processing. Unless we are somehow able to jump the heat dissipation barrier and start raising GHz again.
That's what the superconducting FETs are for, just wait a few years / couple decades for them to get something that can be made on an IC and works at liquid nitrogen temperatures.
A dual-channel memory controller used to be the exception, not the rule, but now the idea is very common. In time, a 32 channel memory controller will be the standard even in an average home computer. How those channels are used to talk to memory of course remains to be seen, but you get the idea.
Where are you going to find enough pins for that, or the space for the DIMM slots?
I'd expect either (1) completely on-die or in-package memory (and fixed memory per core) or (2) some sort of stackable chips (how would this interact with heatsinks?) where your CPU has a grid of contacts on top and the memory has a grid on the bottom, which should allow for much higher speeds because you're not driving some hugely long set of wires.
Slashdot Mirror
In order to stop spam, we need to stop using SMTP and switch to a protocol that rejects mail by default.
I wasn't aware that SMTP was incompatible with whitelisting. In fact, I'm pretty sure I've heard of setups that do just that, result in an SMTP server that rejects connections from people it doesn't know.
There's a trivial technological means to fight spam. It just requires abandoning SMTP and moving to a new protocol with the following requirements.
You forgot one:
With that, spam is basically dead. As soon as you require those restrictions, suddenly spammers have to actually own a domain name and provide a working DNS server in order to deliver spam, and that DNS server must contain up-to-date mappings for those hosts to IP numbers. That pretty much obliterates the use of zombies for delivering mail.
Unless they can 0wn a DNS server, or have the zombies send through the owner's legitimate outbound email accounts, or can get a steady supply of disposable domains somewhere (zombie-XXXXXX.disposable-20081217.com, etc).
It also means that there is now a domain name, which by ICANN policy, is required to have a valid postal address, phone number, and other contact information associated with it.
And when the spammers don't follow the policy? Sure the domains might get shut down after someone realized (and got the registrar to verify) that the contact info was bogus, but that's a bit too late.
If Google were more libertarian than liberal, I would expect them to be proposing a referendum in California to sweep away all of the franchising laws so that there are no local or state limits on who can enter what Internet or TV market.
Part of the logic behind franchising laws is that they give more revenue to local governments, but so what?
Might be kinda difficult right now, isn't California (among many others) having some budged difficulties?
The unregulated so-called "free market" will take care of everything, right?
A working free market requires a number of assumptions, such as that people know what they're buying, that nobody has too much market power, and that efficient matching of buyers and sellers is actually the desired outcome. This last item especially doesn't hold for (at least basic) health care, where being universal is probably more important than being economically efficient.
Just look at what great shape our economy is in...
That's partly a case of people (well, banks) not understanding what they were buying, and I think partly a case of the nation approaching its collective credit limit. Different (not more, different) regulations would have helped with the former, and the government not price-fixing the credit markets would probably have helped with the latter.
Why??
RAM is far faster. simply pump your system up to 4-16 Gigs of ram and call it done. Why do you want a kludge like a second drive?
RAM is more expensive per GB, except for the very high-end SSDs. RAM also isn't persistent across reboots.
It's good that you bring that on, because it also illustrates a security problem: You can not wipe your drive anymore... All your data will be readable by anyone.
Better to just encrypt everything from the start, wiping it only works if you know in advance that your drive is about to change hands. The key can be a boot password or a hardware dongle (USB stick, SD card, etc) or something.
What would bother me more than the high price is the limited number of writes. Sure, there are ways to limit writes to the disk, like disable swapping and delaying writes whenever possible, but I would still rather go with a reliable HDD over a SDD. I never dropped one of my notebooks until today, but then again I never had one that looked like a toy...
Based on what I've heard an SSD is actually much more reliable, even if the lifetime is a bit shorter than the average lifetime for an HDD (which I don't think it necessarily is, especially if you get one of the SLC (faster write) drives).
The question should be "is this the year that SSDs will be price competitive with hard drives?" Until that day comes, SSDs will only sell in small quantities.
Some of them are (might be just the Intel ones?). People just don't know this, because "I/Os per second" isn't one of the commonly given numbers like read/write speeds and capacity are. (Probably because most people don't know why it matters.)
4) Energy security (not sure what this means, sounds like nationalism)
5) Water supply (not sure what this is about, obviously, if water is scarce, it would quickly become very expensive, leading to the power plant using it to close down, assuming a free-market environment)
I do know that there were a couple cases where nuclear plants had to shut down due to drought conditions during summer when people really need electricity for air conditioning (because the rivers they were taking cooling water from started to run low). Not sure if this is what that's about or not, it should really apply to pretty much any highly concentrated power plant.
I"m wondering if their evaluation of nukes..was based on the current 'laws and regulations' in the US (encacted by Carter I think?), that pretty much prohibit things like breeder reactors, that 'can' be used to manufacture weapons grade stuff, but, also can allow the fuel to be used much more efficiently, leaving much less waste than the first run we currently do?
From my limited understanding, if we repealed those laws...we could really stretch the nuclear fuel in a massive way, and have much, much less radioactive waste to have to manage, that has a much lower half life, etc.
THey do assume that. (that's the "HTML article" in the second story link, in case they don't like direct links like this).
They also try to calculate how much use of nuclear electric plants would increase the chance of a nuclear war (by giving more groups access to various nuclear technologies), and the environmental impact such a war might have.
This turns out to be rather insignificant, at least as far a carbon emissions are concerned (table 3). "Lifecycle" emissions for nuclear are "9-70" which is about equivalent to Solar PV or Geothermal, and somewhat worse than the 10-20-ish range of most others. Wind is significantly better at <10. They also calculate an "Opportunity cost", which they have much higher for nuclear because it takes so long to build (partly regulatory issues again).
I think centralization/concentration also works out as a negative, even if not counted directly.
Hydro plants scored about the same as nuclear, only wind/sun/ocean powered systems came out ahead.
Do the violated children get the money?
I thought it was the parents who were "violated", by not getting the required assistance in keeping track of what their children are doing online (because putting the computer where they can see it is too hard)?
Maybe linux will break .8% of the market with this groundbreaking advance.. *snicker*
Windows 37.4% (268), Linux 34.6% (248), Unknown 19.2% (138), Macintosh 7.6% (55), FreeBSD 0.5% (4), Solaris 0.4% (3).
I think it depends on which market you're talking about.
Now, in BOTH cases, you lose your data. But in one case you lose anyone else's data AND the OS, and in the other one, only your data.
This only matters on a shared system, where the different users actually use different accounts and set nontrivial passwords. Any envioronment with users clueful enough to do this, or administrators clueful enough to enforce this, will also have users clueful enough to avoid trojans or administrators clueful enough to secure their systems and make regular backups.
Plus I think the default umask is world-readable, so any one user running a spyware will in fact compromise everyone's data. Malware these days isn't so intent on causing corruption, because that isn't profitable.
But the one where you lose ALL security is the better one????
WTF are you talking about? The only "better" security model I mentioned is per-application access limitations, which I do not believe any mainstream system offers in a usable manner, and which will still be circumvented by a sufficiently motivated (chance of free porn or cute puppies) idiot.
Microsoft just spent $9 billion and many years to create Vista, so it does not sound reasonable that some new alternative could just snap into existence overnight like that. IBM tried, and spent a huge amount of money developing OS/2 but could never keep up with Windows.
Is this a joke?
Or maybe it's just someone who doesn't understand that less buggy software is easier (cheaper/faster) to build than buggy software.
Even though users can have their files easily restored in minutes from a backup?
What backup?
And since the malware can't infect at the system level it is then a simple matter for an administrator to nix the offending files?
Is "administrator" a fancy term for "geeky neighbor kid"?
The only files that matter are the user's files, everything else can be fixed with apt-get and a livecd. If those files are messed up, it does not matter that the stock OS files are still intact. The *nix security model is good for protecting users from eachother, while malware protection requires protecting users from themselves. Probably the only ways to get the latter are some unmaintainable SELinux config or a highly inconvenient browser-in-a-VM and email-in-a-different-VM setup, and even those can't ever be idiotproof.
But to be quite frank, I lost all faith in any open source business application after I tried using Zen Cart. Those guys fucking deleted the *god damn row from the database* when you removed an order!!! Now I ain't no lawyer or professional accountant, but I do know you *never remove anything from the database unless you are planing some scheme to commit fraud!*. Hide it, yes, but you don't delete the fucking thing forever!
How did you come by this knowledge?
Materials? We're talking about web development here, which should be a pure SERVICE. If you bill for any kind of materials, you are considered a retailer/reseller and have to deal with specific sales tax, B&O tax, recycling & disposal fees, storage and transport laws etc related to physical goods in cities and states that have them. Also, very few people live in neighborhoods zoned for business so you can only do limited services when working out of your home. You generally can't warehouse materials without running afoul of local zoning ordinances.
Make the customer buy their own materials or have them pay you for them separately as an nontaxable reimbursement. If someone needs to buy software or hardware to do the job, YOU buy it and bill them for using it by the hour - that's "time and equipment" and it's how you bill when you are a pure service contractor. You get to amortize the price of the equipment over one or more years and you don't have to deal with most of the stuff that physical contractors have to put up with.
That might not work, at least if the hosting provider is in the same state as you...
While I'm aware that machine code can be optimized slightly better than C code, no one programs in it. (Maybe only electrical engineers when working on very specific types of embedded systems.)
I've seen multiple x86 assembler versions of SHA-1 in the same program, optimized for different processor families (old Intel vs old AMD vs things with SSE2). There's also a tiny notepad replacement that some crazy person wrote entirely in assembly ("grown up notepad" or something, I think). I also know that a few small parts of the Linux kernel must be done is assembly, but AIUI this is more often due to "cannot be expressed in cross-platform languages" rather than performance.
Most software engineering techniques rely heavily upon testing to provide quality assurances.
I'm not sure such a lack of proper design really counts an "engineering".
With parallel algorithms, testing provides almost not guarantees because of the combinatorial explosion in the test cases that you will need to test all the timing paths.
That means that you're doing it wrong. It should at all times be clear what data is owned by what thread. Race conditions should be designed out and the code reviewed for adherence to the design, or you could use one of the languages that don't have a way to express race conditions.
And no, engineers don't deal with stuff like that all the time because if they did they simple would fail get the job done correctly.
And because they're legally required to know what they're doing.
CSP? I think you mean CPS (Continuation Passing Style).
Communicating Sequential Processes
So a quick question before I go and master FP...does the compiler automatically compile the code that can be done in parallel in the proper "way", or do I have to specify something?
I would imagine that this depends on the language and compiler you use.
Also, if I rewrote an app written in an imperative language to a FP one like Haskell, would I see a that much of a difference on a multi-core processor?
That depends on a great many things, including the quality of the compilers, your skill at writing code into the different languages, and the nature of the particular app.
"Only drawback is they need to be supercooled, something that may be addressed by improving the materials used." - that last part is a bit of an understatement. We're still decades (centuries?) away from room temperature superconductors.
Even getting it up to liquid nitrogen temperatures would probably be "good enough" for non-portable uses. I'm pretty sure that's a heck of a lot cheaper than liquid helium.
I'm posting anonymously because I don't want to have people at my company know who I am.
And also because you're trolling?
But it seems to me that Linux while cheap to buy is not cheap to keep patched and secure, particularly in a fleet of inhomogeneous platforms and users and network,printer, or disk sharing conditions in different buildings and subnets.
So don't make it so heterogeneous, then.
But it takes effort, dicsipline and an above avegage IT guy. And if you lose that person, you are screwed. Even a new equally skilled guy probably can't get all the scripts and stuff the last guy used to manage to work.
Effort and discipline, yes, but doing anything well requires those. But if the new guy can't get things to work it means someone wasn't required to document things or use a versioning repository. I'd also question the "above average" requirement, unless you're saying that "average" means "insists on known-bad practices even when told otherwise".
With windows, you can take a balow average imbecile, get them through a certification course, and they become almost interchangable monkeys. you need a lot of them since you will constantly be fighting fires or hunting down the right driver for the given brand of computer, but they can do it and it will work.
Or you can take someone who knows what they're doing, and things will work about as well as your Linux system does with people who mostly know what they're doing.
Moreover, and this is the critical part, a manager who is not an expert can tell if his monkies are keeping up with patches. MS tells him what he need to do. With Linux you can't really tell if the IT guy is doing it all, or if your pants are around your ankles.
[citation needed]
So it's not enough to use Linux to reduce TCO. you need to have a company like IBM telling you how to manage your configuration. Not because a skillful IT can't. But because a manager will know that IBM has his back.
No, you just need to treat your config files and admin scripts like any other programming project and demand good practices like source control and adequate documentation .
saddly a mediocre virus prone Windows network is, to a manager, much easier to sleep at night, than a well run Linux system that's tight as a ducks Ass, simply because he knows it's reasonably safe from an industry standard point of view.
[citation needed]
people will trade, extremes (linux) for mediocre, if they can limit thier risks.
People would rather flee the unknown and stay in their comfort zone, instead of learning about something new. With hard times forcing people to make an effort, maybe they'll find out that Linux isn't as scary and risky as they've been told.
So yes, we are probably seeing the beginning of the end of performance gains using general-purpose CPU interconnects and have to go back to vector processing. Unless we are somehow able to jump the heat dissipation barrier and start raising GHz again.
That's what the superconducting FETs are for, just wait a few years / couple decades for them to get something that can be made on an IC and works at liquid nitrogen temperatures.
A dual-channel memory controller used to be the exception, not the rule, but now the idea is very common. In time, a 32 channel memory controller will be the standard even in an average home computer. How those channels are used to talk to memory of course remains to be seen, but you get the idea.
Where are you going to find enough pins for that, or the space for the DIMM slots?
I'd expect either (1) completely on-die or in-package memory (and fixed memory per core) or (2) some sort of stackable chips (how would this interact with heatsinks?) where your CPU has a grid of contacts on top and the memory has a grid on the bottom, which should allow for much higher speeds because you're not driving some hugely long set of wires.