While not a perfect measure of a company, currently Apples market cap (159.5 Billion) is greater than IBM's (144.9 Billion). AT&T is currently at 231.7 Billion market, cap, so by that commonuly used measure, AT&T is still bigger.
I would wager that IBM didn't blow off Apple, but that IBM really couldn't deliver a performance competitive in a form with a TDP appropriate for laptops, with the final straw being Intel releasing Core2, for all intents and purposes erasing the instructions per clock advantage the PPC architecture had. (I know Apple made the jump before that, but I guarantee you that Intel shared the Core2 info with Apple).
Apple smartened up and realized that even when IBM made up for it, the simple fact was that Apple wasn't able to consistently differentiate themselves on hardware performance (and it really wasn't one of their goals now anyway), so they decided to play in the same market as their competitors, ensuring that they wouldn't appear to be left behind at any point in time. Extra bonus of Windows compatibility in the face of the market reality of desktop software. They chose to differentiate on brand, styling, and software (to an extent).
My experience has been most people stumble upon my resume while not looking for resumes at all. They'll google for answers to weird esoteric problems with esoteric stuff they are using, and my resume shows up on the first page of results... A coworker of mine decided it was a sign to skip googling when his google attempt just landed my resume anyway and called me about the problem instead.
I use Gaim OTR, and my buddy used Trillian OTR (without him even realizing it incidently). There was a Gaim encryption plugin before the OTR plugin, but I don't know anyone using that anymore.
It's nothing new for that much moneys worth of equipment to be in a single truck. Quite often I know trucks full of a datacenter's worth of racks drive to the destination..
That said, I wonder if the 'portable' or 'modular' aspect of it is really useful/cost saving. "Because it's a small, contained environment, cooling costs are far less than for traditional data centers", but why is it the case that a on-site constructed datacenter *must* be larger? I look at the pictures and it seems more like the 8' wide restriction imposed by the trailer width is used as a sort of excuse for making things much harder to service. When constructing a datacenter, you could probably build an 8'x40' room and be equally inconvenient, but probably achieve the same cooling situation that benefits these installs. It also has no place to service systems, or for people to actually work. All these things should be possible outside the strict cooling zone suitable for the racks, but a lot of datacenter design likes to provide for it in one place. Another thing is that in a traditional data center, OHSA dictates hot aisles be at least 3 feet wide, and floor tiles being about 2 foot wide that generally leads to 4 feet of space between the rear of racks, when 2 feet would have sufficed.
So is the lesson to make aisles between rear of racks impossibly narrow to get around OHSA, drop ceilings and have narrow service aisles, and the portable datacenters help by being so inconvenient as to force these issues?
My experience is that some of the newer drives are pretty decent on noise. But, getting out of local disks isn't hard at all even with pretty run-of-the-mill motherboards, at least not with linux. You could setup up a box with the gobs of storage somewhere, put some linux on it (CentOS 5.1 might be a good choice) with software target (CentOS 5.1 ought to get that incuded, since RHEL5.1 did, otherwise google for iet). CentOS/RHEL support a fairly nomral install to an iSCSI software target with the software initiator. Since most desktop systems don't support iSCSI boot, you'll have to access the SCSI target elsewhere post install and set up your network boot server to serve up that kernel and initrd, or, use a USB flash disk as/boot and not worry about network booting at all.
On the Windows side, I'm not sure if MS has any way at all to install and boot appropriately for iSCSI without iSCSI firmware/hardware available. In fact, the trend seems to be that most of the fubar firmware implementations exists mainly because Windows needs it (the commonly known example is the significant amount of 'fakeraid' cards that satisfy BIOS calls until the 'real' driver loads which contains the vendor's chosen software RAID implementation, Windows needs it to boot because they can't bootstrap flexibly while Linux could have a normal or RAID-1/boot and other than that small piece can use the standard software RAID implementation).
Except they *can't* make such assumptions at a higher layer. For example, the generic layer has to anticipate that a single presented block device could represent a RAID-0 of devices. The specific driver, would, of course, be written for that card or else it would probably emulate a controller with hardware raid-0 capability.
The performance of the BitMicro SSD isn't even up to a SATA-II 3.0 Gb/s speed. And to top it off, they only seem to support FC, which is currently commonly 4.0 Gb/s and moving toward 8.0 Gb/s. If you have more than one drive, yes, you'll exceed 3.0, but you'll also be using more SATA ports, so in the aggregate, still staying below the SATA standard limits. Maybe a different standard will come, but for now, SATA is perfectly capable. The likes of PCI-e has not been extended in a standardized way to allow for significant trace (I guess in this situiation cable) length. Unless PCI-e is extended in such a way *and* drives can cheaply implement the hardware/firmware necessary to speak the more generalized protocol, it will not be feasible and the current state of the technology is also meaning it is pointless to bother. By the time the SSDs get above 3.0 Gb/s per port making sense, I'd wager an iteration of the SATA architecture will be available that would suit the speed and permit reasonable backward and forward compatibility with SATA devices.
The other thing is just a bunch of servers with lots of DIMM slots and a 25 minute UPS marketed as a SAN solution, so no need to sweat that detail with respect to this. (You'd only buy such a thing to hook up to your infiniband or fibre channel fabric, not direct-attached to a single system).
The Texas Memory Systems datasheet claims 24 GB/second of random sustainable data bandwidth which is much higher than the Fusion IO card but it looks like they are serializing this possibly across multiple drives. They also claim higher (3.2 million) operations per second. The Texas memory product is cramming a bunch of ram in 24U and putting a whopping 25 minutes of battery backup. If you disconnect power for more thon 25 minutes, the only thing left is whatever was committed to 'real' peristant storage. They provide Infiniband and FC ports, so it's more akin to an EMC or Engenio storage controller than a hard drive. 1 TB/24U is actually kind of sad when hard drives can easily yield 3 TB/U nowadays. There is a place for this (ramdisk performance is pretty nifty), but it's not even remotely relevant to anything a normal person would think of when they hear SSD (they picture a drive intended to connect directly to a system some how, not participate in a SAN directly.
The BitMicro drive is groin grabbingly amazing in size but claims only 55k operations per second & sustained data transfer rate over 230MB/sec. And it *actually* is flash based storage, meaning it can fairly be called persistant storage. Of course, the clear hint is there when they talk IOps and only mention FC connection that they are targetting only deep pockets with the product as of this press release or whatever. 55k operations and 230 MB/s is ludicrously insane performance for a single drive relative to current spinning disks. You can fit 144 of these into a 24U space and have a theoretical aggregation that exceeds the ram based system specs. Of course, RAM should be able to trounce it so the limiting factor is a controller setup to push the IOPs and throughput, so both solutions would probably perform comprably.
So what I would wager is that PCIe might provide more throughput than SATA but don't quote me on that. I'm interested to see where this goes & also curious to see whether we continue dumping drives on channels like the Texas Memory solution or if it just goes back to a server with a ton of PCIe slots on it and hot pluggable card swapping for 'drives.' Well, considering that SATA controllers at best currently use PCI-e as the method to communicate with the chipset, PCI-e slots better be capable of better than SATA... Ok, it's over-simplyfying, a 1x PCIe first gen slot yields about 2.5 Gb/s or so, and a SATA II port is 3.0 Gb/s, so a single lane PCI-e slot would be slower than a SATA port. However commonly PCIe appears in 4/8/16 lane configurations, I assume you meant 880 MB/s, which would point to PCIe 4x slot sort of throughput, which makes sense, it's not unreasonable to expect at least a 4x lane slot to be free, requiring anything more could waste limited hardware resources.
So key things when building a beowulf cluster are performance per dollar and performance per watt. Hit sweet spots in those and you can adjust the rest through node count.
In this case, buying 4 of these boards would probably suck down more power than a single quad-core Intel planar+processor. The cost of 4 boards (plus memory, etc) would probably be not much cheaper than consolidating all of that into one chassis. So if expecting a significant lifetime out of it, it's not really worth it. You can put together similar budget systems for maybe 30% more money, but with probably 50-70% more performance and better performance per watt.
Now, if it offered say, 15% less performance per planar, but 25% less power consumption at 30% lower price, then yes, it may make sense to buy it and increase node count to offset the difference. But my understanding is that the difference in performance is quite drastic, more drastic than the cost savings or power consumption figures. I heard the VIA platform takes two clock cycles to execute a single-precision floating-point operation. For comparison, current Intel archictecture acheives 4 double-precision flops for every clock cycle (theoretical max). This is of key interest if wanting to compare your setup against the Top500.
Now if your intent is simply to learn the in and outs of clustering, with no practical work expected, and you lack old hardware to hobble together, it may be educational. However, it's likely that any old hardware lying around would be on the order of the same amount of educational value.
The target market is obviously areas where they won't need more than one, in which case scaling back power and cost at the expense of performance is a no-brainer. Other interesting places for VIA products ('just-enough' processing at a low TDP) are embedded. I've been wanting to piece together a Car-PC, but haven't overcome my laziness. A car-pc is an excellent target for VIA based products.
From my brief look at nVidia's site, it looks like it's more an internal standard than external. IPMI stops at the service processor. The service processor, in turn, used as many arbitrary busses/protocols internally to gather the data and report it back (and while the service processor *might* help you with fan-control versus throttling using OEM commands, the sensor interface strictly speaking only mandates reporting, not controlling). Such a system, for example, would do perfectly fine at reporting on all components explicitly integrated by the manufacturer. However, stick in an arbitrary graphics card, and no sensor value for it's temp/fan will appear as an IPMI-reported device. If a hypothetical standard existed within the case, monitoring may work better without having to buy a single-vendor solution that planned for all that, and you could hypothetically monitor and control arbitrary installed devices. Also, if intelligent about it, more tradeoffs can be made. When you can send instruction to both the GPU and it's cooling device, then for performance you could request that the cooling ramp up whatever it takes, but then go to acoustic and it would throttle the GPU instead of ramping up fans. An IPMI compliant BMC that was also ESA compliant might generate SDRs based on some sort of ESA-discovery.
nVidia isn't the only game in town talking about internal standards for this, Intel for one definitely has been pushing their own ideas in terms of moving standardization to the interior of the systems.
Entropy in a system can be reversed, you just have to increase entropy by more than the same amount outside of the system.
In this case, entropy is decreased every time a battery is recharged, with respect to the system that is the battery. Theoretically speaking, a battery that for its entire life could be charged to full capacity and drained does not go against the rationale of entropy. Just everyt ime you charge it, you induce more entropy in the universe.
After being bombarded for the past decade with seemingly endless presentations, I'm certain that overall it has brought down the quality of presentations and discussions.
The first obvious problem is if people think they need lots of 'features' in presentation software (i.e. effects), they are 100% doomed to make a useless piece of trash. The core of the presentation if it must be done should be simple and clean, not Myspace-style crap. Some font selection and subtle bacgrounds can assist, but any intra-slide animations (text sliding in or appearing bullet-point-by-bullet-point) are killer and inter-slide animations if used generally are horrible, long, and cheesy. I could see some subtle, hypothetical sub 200 ms transitions being less jarring than simple screen replacement, but I never see such things happen.
A more critical flaw is people begin intrinsically worrying about the presentation file itself rather than being more broadly prepared. It's a fixation that leads them to the path of more or less parroting the slides, perhaps with some emphasis.
Further taxing things, is I've started to see presentation files used as the medium of choice for more general transaction. I get information files and product summaries as a powerpoint file too often. It's the worst of all worlds. On the one hand, the medium is targetted at large-font display, so content is limited, and thus they omit important information to fit the format. On the other hand, they truly cannot trim enough information, and as such end up with unpresentable crowded pages despite trimming useful information. Additionally, breaks between slides always are awkward. It's just bad.
Not to mention the effect it has on the nature of discourse. Without a presentation for the general audience, the discourse can be bidirectional and free-flowing. The presenter may have private notes that can be consulted at will, but it doesn't constrict the nature of the discourse. With a presentation, by and large people feel obligated to follow the flow dictated by the big screen, rather than engaging in more constructive methods.
So far, I've gotten it to suspend, and resume successfully in Gutsy, *but* the catch that makes it useless, is I essentially make/etc/acpi/resumed/65-console.sh *not* attempt a single chvt, and so I can never get back to X. So the kernel seems fine, but X won't come back and chvt will hang in the process. Have tried saving vbestate and post_video, and neither, but of those two there is no success. The graphics did work with the vbestate saving and post_video in text consoles though....
The SLUB problem was that laptop would never complete suspend. The manifistation of it was that the half-moon would just keep blinking and never go solid. Now with 8.42 (mind you, hacked for a FireGL PCIID), it completes suspend. Resuming however, has yet to return me a working screen. However, if I kill X with alt-sysrq-k, and I can set capslock and change VTs, indicating it almost works. I'll play with the resume scripts, but it appears that the SLUB-blocking-suspend was addressed, but for some strange reason, it still doesn't resume right.
I'm running Ubuntu gutsy, and the T60p out of the box suspended fine. If I allow the binary ATI drivers onto my box, suspend breaks because their driver has a bug with the SLUB allocator, preventing sleep.
The other frustration is that I'm pretty much stuck using ndiswrapper, the madwifi driver is way too flaky on this laptop.
If I had a choice of equipment, would've gone with an nVidia graphics and Intel wireless, but oh well.
It won't necessarily always suck or even have to suck today. The only question to ask is whether the hardware vendors behind the various components in your system take Linux seriously. The same goes for any operating system and platform. Can you expect OSX to run well and support absolutely everything on a set of hardware that Apple never blessed? Can you expect NT4 to install and run on a brand new system with SATA controllers from a manufacturer who decided NT4 wasn't worth the time to implement?
By the same token, you can expect a number of systems and devices to be taken seriously even today. If the hardware is common with any Tier one server equipment, it's essentially guaranteed to work. I personally have had good luck with printers, though I confess desktop printers I have not seen evidence of actively trying to support linux. If Linux adoption was ~10%, a vast majority of hardware would absolutely work correctly under the platform. The problem may seem like chicken and the egg, but the truth is that it's at least reached the point where hardware vendors have had to take it seriously, some for the consumer and others who realize the value as an embedded platform, and also internationally there is interest in not funneling so much money to an American software company. The conditions are not perfect, but are surprisingly decent for the right things to occur for linux as a platform.
On Vista, it's more applicable to consider the whole Windows environment. You forget that for the people who would've bought Vista, the reason was not so much it explicitly sucking, but that XP suits their needs fine and they don't see the point of going to a slightly different setup and spend money to do so when they have a working solution today. For the Windows people that took the time to evaluate it and declare that it sucks, it's generally because either drivers for Vista were lower quality than XP, and the resources the Vista fluff takes up drags the system down more than XP. So in essence, Vista if anything is losing to XP, not driving people significantly to non-Windows systems. The people blowing it up in the media and declaring Vista sucks are these people plus the people who dislike Windows no matter the incarnation, but are there to revel in what failings they can.
On Linux, I disagree that for most common non-gaming people, that Linux has to be too geeky anymore. There is the great potential on a Linux system to geek out to extraordinary degrees, but if you don't elect to and the hardware vendor providing your platform explicitly tries to work with Linux, the experience can be quite straightforward to people who never want to 'pop the hood' so to speak. Claiming Linux is too geeky is to an extent like claiming OSX is too geeky because of the BSD core, the fact you can start terminal and get a *nix shell, and it uses the NeXT defaults system for configuration. A vast majority of OSX users may never realize these facts (or if they do, what they mean except to bring up to defend their platform), and the same can be true for desktop linux users. The exception is when trying to use hardware whose drivers are off the beaten path, and the way the Linux market goes, it's far more common to have the system vendor not paying attention to Linux, and therefore pushing this evaluation to the consumer I have seen in the x86 world system vendors switch components because the Windows drivers the hardware vendor wrote not be able to perform reliably. That's a huge part of what's biting Vista today (people ugrading their systems may have components where someone couldn't have possibly known the Vista driver quality for). I can pre-select a set of hardware, assemble it barebones, and hand the install disc to a non-technical person, and they can be up working with documents and surfing the web without significant assistance.
I could write a client that keeps track of which pieces I downloaded from whom. A swarm of these clients could ensure they got full copies from each of the targets. Ah ha! My client which downloads the whole thing, but only allows itself to seed certain random parts but never certain parts would foil you!
But seriously,RIAA's strategy as some people have pointed out is as flimsy as screenshots with filenames. If that's acheiving success, they don't have to try so hard...
Whatever it was. I've pretty much thrown up my hands at trying to say something 'looks' green or orange or anything if it is in the realm I can't discern, because I recognize I simply have no comparative base to describe what something looks like to me in order to match a normal color vision person's perception.
However, I do feel confident that it's accurately characterized in my case as a dramatic insensitivity to red, so it is a decent bet that to me that everything looks less red (i.e. brown looks green is my logical guess, but no way of knowing I map green to the same thing other people map green to, etc.)
Not many, but in numa, there is benefit for allocations and execution occur on the same memory 'node'. Of course, the OS scheduler should know what to deal with this...
Slashdot Mirror
While not a perfect measure of a company, currently Apples market cap (159.5 Billion) is greater than IBM's (144.9 Billion). AT&T is currently at 231.7 Billion market, cap, so by that commonuly used measure, AT&T is still bigger.
I would wager that IBM didn't blow off Apple, but that IBM really couldn't deliver a performance competitive in a form with a TDP appropriate for laptops, with the final straw being Intel releasing Core2, for all intents and purposes erasing the instructions per clock advantage the PPC architecture had. (I know Apple made the jump before that, but I guarantee you that Intel shared the Core2 info with Apple).
Apple smartened up and realized that even when IBM made up for it, the simple fact was that Apple wasn't able to consistently differentiate themselves on hardware performance (and it really wasn't one of their goals now anyway), so they decided to play in the same market as their competitors, ensuring that they wouldn't appear to be left behind at any point in time. Extra bonus of Windows compatibility in the face of the market reality of desktop software. They chose to differentiate on brand, styling, and software (to an extent).
My experience has been most people stumble upon my resume while not looking for resumes at all. They'll google for answers to weird esoteric problems with esoteric stuff they are using, and my resume shows up on the first page of results... A coworker of mine decided it was a sign to skip googling when his google attempt just landed my resume anyway and called me about the problem instead.
I use Gaim OTR, and my buddy used Trillian OTR (without him even realizing it incidently). There was a Gaim encryption plugin before the OTR plugin, but I don't know anyone using that anymore.
It's nothing new for that much moneys worth of equipment to be in a single truck. Quite often I know trucks full of a datacenter's worth of racks drive to the destination..
That said, I wonder if the 'portable' or 'modular' aspect of it is really useful/cost saving. "Because it's a small, contained environment, cooling costs are far less than for traditional data centers", but why is it the case that a on-site constructed datacenter *must* be larger? I look at the pictures and it seems more like the 8' wide restriction imposed by the trailer width is used as a sort of excuse for making things much harder to service. When constructing a datacenter, you could probably build an 8'x40' room and be equally inconvenient, but probably achieve the same cooling situation that benefits these installs. It also has no place to service systems, or for people to actually work. All these things should be possible outside the strict cooling zone suitable for the racks, but a lot of datacenter design likes to provide for it in one place. Another thing is that in a traditional data center, OHSA dictates hot aisles be at least 3 feet wide, and floor tiles being about 2 foot wide that generally leads to 4 feet of space between the rear of racks, when 2 feet would have sufficed.
So is the lesson to make aisles between rear of racks impossibly narrow to get around OHSA, drop ceilings and have narrow service aisles, and the portable datacenters help by being so inconvenient as to force these issues?
Try decades! The good old days of Unix even had salts (even if they were just two bytes)
My experience is that some of the newer drives are pretty decent on noise. But, getting out of local disks isn't hard at all even with pretty run-of-the-mill motherboards, at least not with linux. You could setup up a box with the gobs of storage somewhere, put some linux on it (CentOS 5.1 might be a good choice) with software target (CentOS 5.1 ought to get that incuded, since RHEL5.1 did, otherwise google for iet). CentOS/RHEL support a fairly nomral install to an iSCSI software target with the software initiator. Since most desktop systems don't support iSCSI boot, you'll have to access the SCSI target elsewhere post install and set up your network boot server to serve up that kernel and initrd, or, use a USB flash disk as /boot and not worry about network booting at all.
/boot and other than that small piece can use the standard software RAID implementation).
On the Windows side, I'm not sure if MS has any way at all to install and boot appropriately for iSCSI without iSCSI firmware/hardware available. In fact, the trend seems to be that most of the fubar firmware implementations exists mainly because Windows needs it (the commonly known example is the significant amount of 'fakeraid' cards that satisfy BIOS calls until the 'real' driver loads which contains the vendor's chosen software RAID implementation, Windows needs it to boot because they can't bootstrap flexibly while Linux could have a normal or RAID-1
It looked to me like a wide-angle lens shot. In which case, it could be much better than pictured. Insufficient detail to be able to know...
Except they *can't* make such assumptions at a higher layer. For example, the generic layer has to anticipate that a single presented block device could represent a RAID-0 of devices. The specific driver, would, of course, be written for that card or else it would probably emulate a controller with hardware raid-0 capability.
The performance of the BitMicro SSD isn't even up to a SATA-II 3.0 Gb/s speed. And to top it off, they only seem to support FC, which is currently commonly 4.0 Gb/s and moving toward 8.0 Gb/s. If you have more than one drive, yes, you'll exceed 3.0, but you'll also be using more SATA ports, so in the aggregate, still staying below the SATA standard limits. Maybe a different standard will come, but for now, SATA is perfectly capable. The likes of PCI-e has not been extended in a standardized way to allow for significant trace (I guess in this situiation cable) length. Unless PCI-e is extended in such a way *and* drives can cheaply implement the hardware/firmware necessary to speak the more generalized protocol, it will not be feasible and the current state of the technology is also meaning it is pointless to bother. By the time the SSDs get above 3.0 Gb/s per port making sense, I'd wager an iteration of the SATA architecture will be available that would suit the speed and permit reasonable backward and forward compatibility with SATA devices.
The other thing is just a bunch of servers with lots of DIMM slots and a 25 minute UPS marketed as a SAN solution, so no need to sweat that detail with respect to this. (You'd only buy such a thing to hook up to your infiniband or fibre channel fabric, not direct-attached to a single system).
So key things when building a beowulf cluster are performance per dollar and performance per watt. Hit sweet spots in those and you can adjust the rest through node count.
In this case, buying 4 of these boards would probably suck down more power than a single quad-core Intel planar+processor. The cost of 4 boards (plus memory, etc) would probably be not much cheaper than consolidating all of that into one chassis. So if expecting a significant lifetime out of it, it's not really worth it. You can put together similar budget systems for maybe 30% more money, but with probably 50-70% more performance and better performance per watt.
Now, if it offered say, 15% less performance per planar, but 25% less power consumption at 30% lower price, then yes, it may make sense to buy it and increase node count to offset the difference. But my understanding is that the difference in performance is quite drastic, more drastic than the cost savings or power consumption figures. I heard the VIA platform takes two clock cycles to execute a single-precision floating-point operation. For comparison, current Intel archictecture acheives 4 double-precision flops for every clock cycle (theoretical max). This is of key interest if wanting to compare your setup against the Top500.
Now if your intent is simply to learn the in and outs of clustering, with no practical work expected, and you lack old hardware to hobble together, it may be educational. However, it's likely that any old hardware lying around would be on the order of the same amount of educational value.
The target market is obviously areas where they won't need more than one, in which case scaling back power and cost at the expense of performance is a no-brainer. Other interesting places for VIA products ('just-enough' processing at a low TDP) are embedded. I've been wanting to piece together a Car-PC, but haven't overcome my laziness. A car-pc is an excellent target for VIA based products.
From my brief look at nVidia's site, it looks like it's more an internal standard than external. IPMI stops at the service processor. The service processor, in turn, used as many arbitrary busses/protocols internally to gather the data and report it back (and while the service processor *might* help you with fan-control versus throttling using OEM commands, the sensor interface strictly speaking only mandates reporting, not controlling). Such a system, for example, would do perfectly fine at reporting on all components explicitly integrated by the manufacturer. However, stick in an arbitrary graphics card, and no sensor value for it's temp/fan will appear as an IPMI-reported device. If a hypothetical standard existed within the case, monitoring may work better without having to buy a single-vendor solution that planned for all that, and you could hypothetically monitor and control arbitrary installed devices. Also, if intelligent about it, more tradeoffs can be made. When you can send instruction to both the GPU and it's cooling device, then for performance you could request that the cooling ramp up whatever it takes, but then go to acoustic and it would throttle the GPU instead of ramping up fans. An IPMI compliant BMC that was also ESA compliant might generate SDRs based on some sort of ESA-discovery.
nVidia isn't the only game in town talking about internal standards for this, Intel for one definitely has been pushing their own ideas in terms of moving standardization to the interior of the systems.
Entropy in a system can be reversed, you just have to increase entropy by more than the same amount outside of the system.
In this case, entropy is decreased every time a battery is recharged, with respect to the system that is the battery. Theoretically speaking, a battery that for its entire life could be charged to full capacity and drained does not go against the rationale of entropy. Just everyt ime you charge it, you induce more entropy in the universe.
After being bombarded for the past decade with seemingly endless presentations, I'm certain that overall it has brought down the quality of presentations and discussions.
The first obvious problem is if people think they need lots of 'features' in presentation software (i.e. effects), they are 100% doomed to make a useless piece of trash. The core of the presentation if it must be done should be simple and clean, not Myspace-style crap. Some font selection and subtle bacgrounds can assist, but any intra-slide animations (text sliding in or appearing bullet-point-by-bullet-point) are killer and inter-slide animations if used generally are horrible, long, and cheesy. I could see some subtle, hypothetical sub 200 ms transitions being less jarring than simple screen replacement, but I never see such things happen.
A more critical flaw is people begin intrinsically worrying about the presentation file itself rather than being more broadly prepared. It's a fixation that leads them to the path of more or less parroting the slides, perhaps with some emphasis.
Further taxing things, is I've started to see presentation files used as the medium of choice for more general transaction. I get information files and product summaries as a powerpoint file too often. It's the worst of all worlds. On the one hand, the medium is targetted at large-font display, so content is limited, and thus they omit important information to fit the format. On the other hand, they truly cannot trim enough information, and as such end up with unpresentable crowded pages despite trimming useful information. Additionally, breaks between slides always are awkward. It's just bad.
Not to mention the effect it has on the nature of discourse. Without a presentation for the general audience, the discourse can be bidirectional and free-flowing. The presenter may have private notes that can be consulted at will, but it doesn't constrict the nature of the discourse. With a presentation, by and large people feel obligated to follow the flow dictated by the big screen, rather than engaging in more constructive methods.
That's news to me... Where's the Impress comparison?
So far, I've gotten it to suspend, and resume successfully in Gutsy, *but* the catch that makes it useless, is I essentially make /etc/acpi/resumed/65-console.sh *not* attempt a single chvt, and so I can never get back to X. So the kernel seems fine, but X won't come back and chvt will hang in the process. Have tried saving vbestate and post_video, and neither, but of those two there is no success. The graphics did work with the vbestate saving and post_video in text consoles though....
The SLUB problem was that laptop would never complete suspend. The manifistation of it was that the half-moon would just keep blinking and never go solid. Now with 8.42 (mind you, hacked for a FireGL PCIID), it completes suspend. Resuming however, has yet to return me a working screen. However, if I kill X with alt-sysrq-k, and I can set capslock and change VTs, indicating it almost works. I'll play with the resume scripts, but it appears that the SLUB-blocking-suspend was addressed, but for some strange reason, it still doesn't resume right.
Ahh... closed source drivers.
I'm running Ubuntu gutsy, and the T60p out of the box suspended fine. If I allow the binary ATI drivers onto my box, suspend breaks because their driver has a bug with the SLUB allocator, preventing sleep.
The other frustration is that I'm pretty much stuck using ndiswrapper, the madwifi driver is way too flaky on this laptop.
If I had a choice of equipment, would've gone with an nVidia graphics and Intel wireless, but oh well.
It won't necessarily always suck or even have to suck today. The only question to ask is whether the hardware vendors behind the various components in your system take Linux seriously. The same goes for any operating system and platform. Can you expect OSX to run well and support absolutely everything on a set of hardware that Apple never blessed? Can you expect NT4 to install and run on a brand new system with SATA controllers from a manufacturer who decided NT4 wasn't worth the time to implement?
By the same token, you can expect a number of systems and devices to be taken seriously even today. If the hardware is common with any Tier one server equipment, it's essentially guaranteed to work. I personally have had good luck with printers, though I confess desktop printers I have not seen evidence of actively trying to support linux. If Linux adoption was ~10%, a vast majority of hardware would absolutely work correctly under the platform. The problem may seem like chicken and the egg, but the truth is that it's at least reached the point where hardware vendors have had to take it seriously, some for the consumer and others who realize the value as an embedded platform, and also internationally there is interest in not funneling so much money to an American software company. The conditions are not perfect, but are surprisingly decent for the right things to occur for linux as a platform.
On Vista, it's more applicable to consider the whole Windows environment. You forget that for the people who would've bought Vista, the reason was not so much it explicitly sucking, but that XP suits their needs fine and they don't see the point of going to a slightly different setup and spend money to do so when they have a working solution today. For the Windows people that took the time to evaluate it and declare that it sucks, it's generally because either drivers for Vista were lower quality than XP, and the resources the Vista fluff takes up drags the system down more than XP. So in essence, Vista if anything is losing to XP, not driving people significantly to non-Windows systems. The people blowing it up in the media and declaring Vista sucks are these people plus the people who dislike Windows no matter the incarnation, but are there to revel in what failings they can.
On Linux, I disagree that for most common non-gaming people, that Linux has to be too geeky anymore. There is the great potential on a Linux system to geek out to extraordinary degrees, but if you don't elect to and the hardware vendor providing your platform explicitly tries to work with Linux, the experience can be quite straightforward to people who never want to 'pop the hood' so to speak. Claiming Linux is too geeky is to an extent like claiming OSX is too geeky because of the BSD core, the fact you can start terminal and get a *nix shell, and it uses the NeXT defaults system for configuration. A vast majority of OSX users may never realize these facts (or if they do, what they mean except to bring up to defend their platform), and the same can be true for desktop linux users. The exception is when trying to use hardware whose drivers are off the beaten path, and the way the Linux market goes, it's far more common to have the system vendor not paying attention to Linux, and therefore pushing this evaluation to the consumer I have seen in the x86 world system vendors switch components because the Windows drivers the hardware vendor wrote not be able to perform reliably. That's a huge part of what's biting Vista today (people ugrading their systems may have components where someone couldn't have possibly known the Vista driver quality for). I can pre-select a set of hardware, assemble it barebones, and hand the install disc to a non-technical person, and they can be up working with documents and surfing the web without significant assistance.
But seriously,RIAA's strategy as some people have pointed out is as flimsy as screenshots with filenames. If that's acheiving success, they don't have to try so hard...
Whatever it was. I've pretty much thrown up my hands at trying to say something 'looks' green or orange or anything if it is in the realm I can't discern, because I recognize I simply have no comparative base to describe what something looks like to me in order to match a normal color vision person's perception.
However, I do feel confident that it's accurately characterized in my case as a dramatic insensitivity to red, so it is a decent bet that to me that everything looks less red (i.e. brown looks green is my logical guess, but no way of knowing I map green to the same thing other people map green to, etc.)
But then again, being color blind makes a lot of things look the same that shouldn't be...
Not many, but in numa, there is benefit for allocations and execution occur on the same memory 'node'. Of course, the OS scheduler should know what to deal with this...