Do I need a license to stream MP3s from system RAM to the MP3 player? From my Hard Drive to RAM? From my File Server to my machine?
Don't give them any ideas!
There is only one bright side the the above nightmare scenario: Since every time your system move a song, or part one one, between RAM and swap would count as a horrid, horrid copyright infringement at 150k a pop, it should be comparatively easy to convince Legal that upgrading your workstation to 256GB of RAM is money well spent...
I, for one, do not want to be the expert witness who has to parse out the differences between storing multiple copies, symlinks, hardlinks, file-level deduplication, block-level deduplication, whatever jiggery-pokery Amazon's EBS volumes use, whatever different jiggery-pokery supports objects and buckets on S3, to a layman jury and a judge who may or may not have gone to lawyer school back when glass ttys were pretty cool stuff...
It may well be even worse because, since copyright law was written more or less entirely without any reference to contemporary storage, the decision might well hinge on some combination of how the backend works and what abstraction or abstractions are used to present the data to the user.
Probably about the same thing that happens to the families/friends/etc of people who get fired for serious workplace negligence, or who get sent to jail for some crime or other; only getting a new cert is easier and cheaper than replacing a person.
It is, unfortunately, true that nuking them as a trusted CA will have some negative effects on innocent parties. However, there is essentially no form of punishment/consequences, whether leveled against a corporation or a person, that does not affect some innocent bystanders. Somehow, given that the alternative would be the abandonment of consequences, we manage to accept that.
So is "rolling out a new two factor authentication system" code for "our last two-factor authentication system consisted of 'something you know', your username, and 'something you know, your password; because, despite the fact that we are a fucking CA we just can't be bothered"?
Other than inertia, is there any reason to give these guys a second chance, rather than just drop them from the default trusted CAs list and let the company sell itself for scrap? Generating SSL certs is technologically trivial, anybody can do it at home with commonly available free software. Essentially, the only purpose of a CA is to be competent and trustworthy about who they generate certs for. CAs aren't really software or technology companies, they are much closer to the position of escrow services or trust companies. Generating certs is just the minor 'paperwork'. Generating only the right certs for only the right people is the job. If they can't do that, they are worse than useless.
You sound like you were raised by Steve Ballmer and rocked to sleep each night by a loving marketing brochure. Lay it on a bit thicker, will you?
That said, disk encryption(almost certainly full disk; because you Do Not Want to have to puzzle out all the possible locations that a modern OS and suite of common programs may stash temporary files, caches, etc.) is more or less a must for sensitive information that leaves the site. It reduces the hazards of sloppy disposal even for desktops that are only supposed to leave the building at EOL.
You can get disks that do it in hardware, there are a variety of software options; but it is pretty much the bare minimum of responsible handling of sensitive data. Even better, of course, is never actually having the data on the device in the first place. With the comparatively low cost of broad internet coverage today, forcing people working on really sensitive stuff to do so only in a terminal session that actually lives on a nice cozy server back in your locked cage, with only pictures and input device events going back and forth over the (SSL secured) wire is fairly practical and means that even a badly rooted client is limited to some screengrabs and a stolen client gets nothing but a stock OS with one of the terminal clients installed.
Is there any flavor of malfeasance or negligence that we should punish, or do we have to live in constant terror of our precious businesses taking their ball and going home no matter what?
"A single Raven costs about $35,000 and the total system costs $250,000. The RQ-11B Raven UAV weighs about 1.9 kg (4.2 lb), has a flight endurance of 60–90 minutes and an effective operational radius of approximately 10 km (6.2 miles)."
Either military contractors are doing what they do best, or this particular model airplane has some kind of fancy tech tricks up its sleeve...
"Free as in markets" often has comparatively little to do with "free as in freedom", and most corporations don't even care about "free as in markets"; because that implies low barriers to entry for their competitors...
In terms of things like ergonomics, number of buttons you can actually use/remap, etc. but the bottom line is that optical sensors have gotten pretty good, even at the low end.
It is true that the fancy laser stuff will let you mouse on surfaces where basic LED mice won't; but even laser diodes aren't all that costly, though they are used as a price discrimination feature.
Beyond mere ergonomic satisfaction, which is something of a matter of taste, and utility of extra buttons, which is a combination of taste and design, the only place that really dramatic differences jump out at you is with the wireless stuff. It is harder, though still entirely possible, to buy some really dire wireless mice. Slow refresh, shuts down to save power at the worst possible times and then spends 10 seconds waking up again, that sort of thing(and bluetooth? Pay double or insert dongle...)
For your basic rat on a string, though, it is hard to get too worked up about the differences between modern sensors.
I suspect that most of the public reaction is, indeed, a visceral response to the current incident. Emotional, and not likely to last all that long(particularly given that, with incomes flat or declining among the bottom 4 or 4.5 quintiles, and energy costs rising, people are going to grasp at anything that pretends that they will be able to keep on living their familiar suburban existence.
What I find disconcerting about the whole thing is not so much that a given 60's era reactor design didn't cope all that well when exposed to atypically gigantic earthquake and tsunami conditions; but that plant HQ has, apparently, been slimy and dubiously transparent about their somewhat cavalier risk management practices for decades, they've only just had it bite them public-ally.
The "zOMG, nuclear power always causes 3-eyed rats and flipper babies made of pure cancer!" brigade is out to lunch. However, unfortunately enough, the "nuclear power has the potential to be safe; but its operation always seems to end up in the hands of penny-pinching scumweasels who do their best to fail to live up to that promise." is more history than hypothesis.
Until the engineers manage the historic leap of creating a design that managers can't fuck up, certain concerns will remain entirely valid(to be fair, most of those concerns validated, often with grotesque callousness, on a daily basis in other forms of power generation, just ask a coal miner...); but it is true that nuclear designs tend to underperform their theoretical engineering maximums for reasons that come down to frankly untrustworthy management.
The problem is that symbian, while basically just plain unfit for purpose as a modern smartphone OS, can and does run as intended on hardware that would make its next-get contenders cry bitter tears very, very, slowly.
Unless they have something else in mind, "no more symbian" = "We aren't even going to try on the low end"
They recent symbian "flagship" phones have been pretty sad, roughly the same price as a decent android for yesterday's OS and hardware specs; but it'd be sad to see Nokia's classic low-ends, the dumb candybars that get a week on a charge, don't care about being dropped, and just keep on trucking, go.
Unless you are something of a gambler, all SMART errors actually mean "you need a new hard drive". I count myself lucky when an HDD bothers to display any SMART anomalies at all before keeling over. Never trust a hard drive.
I'd say that there are some very interesting potential applications for an external PCIe run with usable cable characteristics(and something resembling actual adoption: there are at least 5, mutually incompatible, relatively obscure systems in limited use at present). If Intel can deliver, great.
I'm just going to be interested to see how the peripheral market shakes out. Pretty much all the other external busses in common use are at least one level of abstraction higher, which makes them less flexible; but tends to reduce the variability in peripheral performance a bit. As you noted, for instance, USB bridge chips frequently end up quietly Doing The Wrong Thing behind your back. On the other hand, because they are behind a USB class and a USB host chip, they are typically free to puke up their brains with minimal disruption to the host system.
Thunderbolt stuff will be more or less as close to your CPU and memory as anything else on your motherboard. On the plus side, this will, one hopes, allow vendors who care to implement cool products that allow you to add otherwise impossible peripheral setups with a single standard cable, the sort of thing that normally requires one of those 200+ pin proprietary docking station connectors. On the minus side, if the OS's standard driver for whatever POS SATA controller was cheapest doesn't take hotplugging into account, it should be eminently possible to bring your kernel to a screaming halt just by plugging in that disk array... Time will tell how good the peripheral market ends up being.
It will be interesting, possibly morbidly so, to see whether that potential is realized.
My understanding is that, effectively, thunderbolt allows you to run a 4x PCIe line through external cabling of usable length and hot-swappability. This will, presumably, mean that (unlike USB) thunderbolt peripherals will be packages that include a PCIe version of whatever type of chip is appropriate(SATA controller for external disk arrays, USB/FW controller to support a monitor with a bunch of ports and a cardreader, etc.) On the plus side, assuming your OS supports any special sauce needed by the thunderbolt chip, it will have pretty much the same access to the PCIe peripheral on the far end of the cable that it would were that same chip on an internal card. On the minus side, if support for whatever chip the manufacturer used is lousy or nonexistent, there will be no abstraction to save you.
USB, on the contrary, doesn't provide direct access to a lot of what a PCIe connection could; but it has a number of useful generic abstractions(the assorted device classes) that, after some ugly teething issues, frequently Just Work. On the other hand, a combination of being able to hide behind these classes and good old price pressure means that a lot of USB devices get a bit dodgy under pressure.
It will be interesting to see if, with Thunderbolt peripherals, the experience is generally equivalent to plugging in a high-quality PCIe expansion card of the relevant type, only with external cabling and hot-plug, or whether vendors will take advantage of being able to just put "Thunderbolt!" on the box while shoving the nastiest PCIe peripheral chips they can find into the product... In the case of storage, say, it is true that a fair percentage of USB2.0/SATA bridge chips pretty much suck. And, since they rarely put the chip used on the box, you don't always know until you buy. What we don't yet know is whether thunderbolt devices will typically feature good varieties of their respective peripheral chips, or if an analogous class of nasty but inexpensive PCIe chips will end up being used.
eSATA is in the somewhat uncomfortable position of simultaneously being a fairly good replacement for storage tasks that used to require springing for expensive SCSI gear and being about as mature as USB 1 when it comes to things like "most products on the shelves will, in fact, work properly" and "hotplugging, autodetection, hubs('port expanders' in eSATA parlance), and such will actually not fall over in a screaming heap even if you don't do copious research and find the one true chipset combination of Antioch."
Presumably the US could just ask MS nicely for a neat digest of accounts of interest, delivered from their US-located datacenters, rather than asking them nicely to turn off SSL, and then having to MITM a whole bunch of people in a variety of largely hostile locales...
SSL doesn't exactly keep Microsoft from reading your hotmail, it just keeps those between you and them from doing so(terms and restrictions may apply...)
I'm genuinely curious what the logic is. "zOMG the Feds!!!" seems unlikely(because Microsoft doesn't exactly have to crack the SSL connection between you and itself to watch you and provide whatever information they wish...) It also seems somewhat unlikely that they received a "disable SSL or we block you" ultimatum, in silence, from a veritable laundry list of undesirable locations at the same time. Those countries also represent a reasonably broad spectrum of different flavors of repressive fucked-upness, and a fair variety of different levels of "they may be dictators with blood on their hands; but they serve our interests", everything from "They are our good buddies who let us headquarter the 5th fleet" to "we would really prefer if they died in a fire.."
That makes it sort of tricky to assign a foreign-policy based incentive behind Microsoft's activities. Economics, though, isn't obviously more helpful. That list represents one hell of a GDP spread, from "barely subsisting" to "oil plutocracy", so it doesn't seem to be a straightforward 'eh, you guys just aren't worth the SSL costs, fuck it." cutoff.
The trouble is, it isn't the cost of connectors that doomed ubiquitous firewire(FW connectors are a little bit more complex than USB ones, FW800 a bit more complex again, but the price difference isn't too drastic); but the fact that including firewire always meant including an additional chip and supporting circuitry. Essentially every chipset, and a wide variety of SoCs and whatnot, ended up supporting at least 1 USB port, often substantially more. Firewire always meant an extra chip.
If thunderbolt ends up embedded in Intel chipsets, its future is likely assured(unless they really run away with using it as a price discrimination feature). If it remains a distinct chip with almost the same die area as an upper midrange discrete mobile Radeon, or one of Intel's platform controller hubs, that sucker is going to remain a premium feature.
If anything, the fact that Intel decided to combine their high speed data link port with displayport could make things trickier: If it were a data-only port, a PCIe 2.0 4x port, quite common on nicer motherboards, would neatly support adding a thunderbolt chip. However, that wouldn't support video-out without some hairy and bandwidth intensive cooperation with the video card. That will make adoption in the desktop and workstation market rather more all-or-nothing: unless we go down the delightful route of having "thunderbolt with video" and "thunderbolt without video; but looks the same", the only way to get it will either be on-motherboard(with potential compatibility headaches on the desktop/workstation segment if you want a discrete video card) or on video card, which will tie the thunderbolt chip to a component that is commonly either omitted entirely or replaced extremely frequently.
It will be interesting to see if it overcomes that; but I would be less than entirely surprised if it ends up as the next firewire.
Slashdot Mirror
Yup. It would be embarrassing if it were a couple of teenagers breaking in, therefore it is not a couple of teenagers breaking in.
Perhaps it is the same mysterious "Advanced Persistent Threat" that hit RSA a little while ago...
Do I need a license to stream MP3s from system RAM to the MP3 player? From my Hard Drive to RAM? From my File Server to my machine?
Don't give them any ideas!
There is only one bright side the the above nightmare scenario: Since every time your system move a song, or part one one, between RAM and swap would count as a horrid, horrid copyright infringement at 150k a pop, it should be comparatively easy to convince Legal that upgrading your workstation to 256GB of RAM is money well spent...
I, for one, do not want to be the expert witness who has to parse out the differences between storing multiple copies, symlinks, hardlinks, file-level deduplication, block-level deduplication, whatever jiggery-pokery Amazon's EBS volumes use, whatever different jiggery-pokery supports objects and buckets on S3, to a layman jury and a judge who may or may not have gone to lawyer school back when glass ttys were pretty cool stuff...
It may well be even worse because, since copyright law was written more or less entirely without any reference to contemporary storage, the decision might well hinge on some combination of how the backend works and what abstraction or abstractions are used to present the data to the user.
Unfortunately, the price isn't listed; but one of their salesdrones would probably give you a number upon request.
Probably about the same thing that happens to the families/friends/etc of people who get fired for serious workplace negligence, or who get sent to jail for some crime or other; only getting a new cert is easier and cheaper than replacing a person.
It is, unfortunately, true that nuking them as a trusted CA will have some negative effects on innocent parties. However, there is essentially no form of punishment/consequences, whether leveled against a corporation or a person, that does not affect some innocent bystanders. Somehow, given that the alternative would be the abandonment of consequences, we manage to accept that.
The world is truly lucky that the man with the experience of 1,000 hackers has not yet discovered steroids...
So is "rolling out a new two factor authentication system" code for "our last two-factor authentication system consisted of 'something you know', your username, and 'something you know, your password; because, despite the fact that we are a fucking CA we just can't be bothered"?
Other than inertia, is there any reason to give these guys a second chance, rather than just drop them from the default trusted CAs list and let the company sell itself for scrap? Generating SSL certs is technologically trivial, anybody can do it at home with commonly available free software. Essentially, the only purpose of a CA is to be competent and trustworthy about who they generate certs for. CAs aren't really software or technology companies, they are much closer to the position of escrow services or trust companies. Generating certs is just the minor 'paperwork'. Generating only the right certs for only the right people is the job. If they can't do that, they are worse than useless.
Is to predict and prevent unnecessary hospitalizations.
Predict and prevent?
Prevent with extreme prejudice.
You sound like you were raised by Steve Ballmer and rocked to sleep each night by a loving marketing brochure. Lay it on a bit thicker, will you?
That said, disk encryption(almost certainly full disk; because you Do Not Want to have to puzzle out all the possible locations that a modern OS and suite of common programs may stash temporary files, caches, etc.) is more or less a must for sensitive information that leaves the site. It reduces the hazards of sloppy disposal even for desktops that are only supposed to leave the building at EOL.
You can get disks that do it in hardware, there are a variety of software options; but it is pretty much the bare minimum of responsible handling of sensitive data. Even better, of course, is never actually having the data on the device in the first place. With the comparatively low cost of broad internet coverage today, forcing people working on really sensitive stuff to do so only in a terminal session that actually lives on a nice cozy server back in your locked cage, with only pictures and input device events going back and forth over the (SSL secured) wire is fairly practical and means that even a badly rooted client is limited to some screengrabs and a stolen client gets nothing but a stock OS with one of the terminal clients installed.
Takeoff and landing might be a bit tricky...
Is there any flavor of malfeasance or negligence that we should punish, or do we have to live in constant terror of our precious businesses taking their ball and going home no matter what?
"A single Raven costs about $35,000 and the total system costs $250,000. The RQ-11B Raven UAV weighs about 1.9 kg (4.2 lb), has a flight endurance of 60–90 minutes and an effective operational radius of approximately 10 km (6.2 miles)."
Either military contractors are doing what they do best, or this particular model airplane has some kind of fancy tech tricks up its sleeve...
Does having to file a flight plan with the FAA before use count as an infringement on your right to bear arms?
"Free as in markets" often has comparatively little to do with "free as in freedom", and most corporations don't even care about "free as in markets"; because that implies low barriers to entry for their competitors...
In terms of things like ergonomics, number of buttons you can actually use/remap, etc. but the bottom line is that optical sensors have gotten pretty good, even at the low end.
It is true that the fancy laser stuff will let you mouse on surfaces where basic LED mice won't; but even laser diodes aren't all that costly, though they are used as a price discrimination feature.
Beyond mere ergonomic satisfaction, which is something of a matter of taste, and utility of extra buttons, which is a combination of taste and design, the only place that really dramatic differences jump out at you is with the wireless stuff. It is harder, though still entirely possible, to buy some really dire wireless mice. Slow refresh, shuts down to save power at the worst possible times and then spends 10 seconds waking up again, that sort of thing(and bluetooth? Pay double or insert dongle...)
For your basic rat on a string, though, it is hard to get too worked up about the differences between modern sensors.
I suspect that most of the public reaction is, indeed, a visceral response to the current incident. Emotional, and not likely to last all that long(particularly given that, with incomes flat or declining among the bottom 4 or 4.5 quintiles, and energy costs rising, people are going to grasp at anything that pretends that they will be able to keep on living their familiar suburban existence.
What I find disconcerting about the whole thing is not so much that a given 60's era reactor design didn't cope all that well when exposed to atypically gigantic earthquake and tsunami conditions; but that plant HQ has, apparently, been slimy and dubiously transparent about their somewhat cavalier risk management practices for decades, they've only just had it bite them public-ally.
The "zOMG, nuclear power always causes 3-eyed rats and flipper babies made of pure cancer!" brigade is out to lunch. However, unfortunately enough, the "nuclear power has the potential to be safe; but its operation always seems to end up in the hands of penny-pinching scumweasels who do their best to fail to live up to that promise." is more history than hypothesis.
Until the engineers manage the historic leap of creating a design that managers can't fuck up, certain concerns will remain entirely valid(to be fair, most of those concerns validated, often with grotesque callousness, on a daily basis in other forms of power generation, just ask a coal miner...); but it is true that nuclear designs tend to underperform their theoretical engineering maximums for reasons that come down to frankly untrustworthy management.
The problem is that symbian, while basically just plain unfit for purpose as a modern smartphone OS, can and does run as intended on hardware that would make its next-get contenders cry bitter tears very, very, slowly.
Unless they have something else in mind, "no more symbian" = "We aren't even going to try on the low end"
They recent symbian "flagship" phones have been pretty sad, roughly the same price as a decent android for yesterday's OS and hardware specs; but it'd be sad to see Nokia's classic low-ends, the dumb candybars that get a week on a charge, don't care about being dropped, and just keep on trucking, go.
How can you tell that a marketing guy is lying? His lips are moving.
How can you tell that a marketing guy is omitting important information? His lips aren't moving.
Unless you are something of a gambler, all SMART errors actually mean "you need a new hard drive". I count myself lucky when an HDD bothers to display any SMART anomalies at all before keeling over. Never trust a hard drive.
I'd say that there are some very interesting potential applications for an external PCIe run with usable cable characteristics(and something resembling actual adoption: there are at least 5, mutually incompatible, relatively obscure systems in limited use at present). If Intel can deliver, great.
I'm just going to be interested to see how the peripheral market shakes out. Pretty much all the other external busses in common use are at least one level of abstraction higher, which makes them less flexible; but tends to reduce the variability in peripheral performance a bit. As you noted, for instance, USB bridge chips frequently end up quietly Doing The Wrong Thing behind your back. On the other hand, because they are behind a USB class and a USB host chip, they are typically free to puke up their brains with minimal disruption to the host system.
Thunderbolt stuff will be more or less as close to your CPU and memory as anything else on your motherboard. On the plus side, this will, one hopes, allow vendors who care to implement cool products that allow you to add otherwise impossible peripheral setups with a single standard cable, the sort of thing that normally requires one of those 200+ pin proprietary docking station connectors. On the minus side, if the OS's standard driver for whatever POS SATA controller was cheapest doesn't take hotplugging into account, it should be eminently possible to bring your kernel to a screaming halt just by plugging in that disk array... Time will tell how good the peripheral market ends up being.
It will be interesting, possibly morbidly so, to see whether that potential is realized.
My understanding is that, effectively, thunderbolt allows you to run a 4x PCIe line through external cabling of usable length and hot-swappability. This will, presumably, mean that (unlike USB) thunderbolt peripherals will be packages that include a PCIe version of whatever type of chip is appropriate(SATA controller for external disk arrays, USB/FW controller to support a monitor with a bunch of ports and a cardreader, etc.) On the plus side, assuming your OS supports any special sauce needed by the thunderbolt chip, it will have pretty much the same access to the PCIe peripheral on the far end of the cable that it would were that same chip on an internal card. On the minus side, if support for whatever chip the manufacturer used is lousy or nonexistent, there will be no abstraction to save you.
USB, on the contrary, doesn't provide direct access to a lot of what a PCIe connection could; but it has a number of useful generic abstractions(the assorted device classes) that, after some ugly teething issues, frequently Just Work. On the other hand, a combination of being able to hide behind these classes and good old price pressure means that a lot of USB devices get a bit dodgy under pressure.
It will be interesting to see if, with Thunderbolt peripherals, the experience is generally equivalent to plugging in a high-quality PCIe expansion card of the relevant type, only with external cabling and hot-plug, or whether vendors will take advantage of being able to just put "Thunderbolt!" on the box while shoving the nastiest PCIe peripheral chips they can find into the product... In the case of storage, say, it is true that a fair percentage of USB2.0/SATA bridge chips pretty much suck. And, since they rarely put the chip used on the box, you don't always know until you buy. What we don't yet know is whether thunderbolt devices will typically feature good varieties of their respective peripheral chips, or if an analogous class of nasty but inexpensive PCIe chips will end up being used.
eSATA is in the somewhat uncomfortable position of simultaneously being a fairly good replacement for storage tasks that used to require springing for expensive SCSI gear and being about as mature as USB 1 when it comes to things like "most products on the shelves will, in fact, work properly" and "hotplugging, autodetection, hubs('port expanders' in eSATA parlance), and such will actually not fall over in a screaming heap even if you don't do copious research and find the one true chipset combination of Antioch."
Presumably the US could just ask MS nicely for a neat digest of accounts of interest, delivered from their US-located datacenters, rather than asking them nicely to turn off SSL, and then having to MITM a whole bunch of people in a variety of largely hostile locales...
SSL doesn't exactly keep Microsoft from reading your hotmail, it just keeps those between you and them from doing so(terms and restrictions may apply...)
I'm genuinely curious what the logic is. "zOMG the Feds!!!" seems unlikely(because Microsoft doesn't exactly have to crack the SSL connection between you and itself to watch you and provide whatever information they wish...) It also seems somewhat unlikely that they received a "disable SSL or we block you" ultimatum, in silence, from a veritable laundry list of undesirable locations at the same time. Those countries also represent a reasonably broad spectrum of different flavors of repressive fucked-upness, and a fair variety of different levels of "they may be dictators with blood on their hands; but they serve our interests", everything from "They are our good buddies who let us headquarter the 5th fleet" to "we would really prefer if they died in a fire.."
That makes it sort of tricky to assign a foreign-policy based incentive behind Microsoft's activities. Economics, though, isn't obviously more helpful. That list represents one hell of a GDP spread, from "barely subsisting" to "oil plutocracy", so it doesn't seem to be a straightforward 'eh, you guys just aren't worth the SSL costs, fuck it." cutoff.
Any ideas?
The trouble is, it isn't the cost of connectors that doomed ubiquitous firewire(FW connectors are a little bit more complex than USB ones, FW800 a bit more complex again, but the price difference isn't too drastic); but the fact that including firewire always meant including an additional chip and supporting circuitry. Essentially every chipset, and a wide variety of SoCs and whatnot, ended up supporting at least 1 USB port, often substantially more. Firewire always meant an extra chip.
If thunderbolt ends up embedded in Intel chipsets, its future is likely assured(unless they really run away with using it as a price discrimination feature). If it remains a distinct chip with almost the same die area as an upper midrange discrete mobile Radeon, or one of Intel's platform controller hubs, that sucker is going to remain a premium feature.
If anything, the fact that Intel decided to combine their high speed data link port with displayport could make things trickier: If it were a data-only port, a PCIe 2.0 4x port, quite common on nicer motherboards, would neatly support adding a thunderbolt chip. However, that wouldn't support video-out without some hairy and bandwidth intensive cooperation with the video card. That will make adoption in the desktop and workstation market rather more all-or-nothing: unless we go down the delightful route of having "thunderbolt with video" and "thunderbolt without video; but looks the same", the only way to get it will either be on-motherboard(with potential compatibility headaches on the desktop/workstation segment if you want a discrete video card) or on video card, which will tie the thunderbolt chip to a component that is commonly either omitted entirely or replaced extremely frequently.
It will be interesting to see if it overcomes that; but I would be less than entirely surprised if it ends up as the next firewire.