Well, here's a key on exceptions - don't use them.:)
I know that's a trivial statement, but seriously, I've worked with well-written code that only used existing exceptions where absolutely necessary, had few try-catch blocks, and would almost never throw a RunTime exception because data was validated before being operated on.
Then I had the counter-experience: a massively overly complicated "Exception Framework" complete with layered translation layers all spec'd out via Spring. Why? Because some nimrod read about exception frameworks being the next best thing in some drivel rag, and gave the struts exception framework as an example. What happened? We had individual exception classes written for every exception, complete with translator classes etc. Over 200 classes for something that at most required a single class with a data member indicating type of application exception. I say at most, because many of these were potential NPEs and the like that could have been checked and handled gracefully or via the default exceptions, but instead were layered in a "framework" that made debugging virtually impossible, as the stacktrace was lost through the translation layers - all without logging, of course. I did state that this was a nimrod's brain fart.
Well, we can start with MS implementation and usage of the FS, which is basically "designed to fragment" from the get go. Then continue on with the archaic start at cylinder 1 (effectively) mode of writing. Then continue with absolutely no optimization of file writes when file sizes are known (part of that inherently fragmenting design).
As to safely recover from power failure, you have to be kidding me. I can't count the number of times I've had file corruption on my NTFS drives after a hard reset due to a lockup or failure to shutdown properly. (power failures generally aren't an issue - UPSes deal with that issue) I've even lost 3 drive's contents this way, separately. No they weren't RAIDED.
I don't know how well XFS, JFS, or ReiserFS deal with those, but my ext2/3 boxes seem to have no problems. ReiserFS when I used it had other issues, especially under heavy load. To be fair, it wasn't completed and tested at the time, being a RC release.
I was speaking of my personal experiences there. I can't vouch for the workstation aspect, as I don't use it as such at this time.
btw - Win 2K server and even the workstation is pretty much considered the zenith of MS OSes as far as reliability, performance, and general architecture goes. Everything since has been a bit of a downhill slide down one of the criteria for good reliable OSes.
I started some work with C# once, and quickly decided that given the choice, I wouldn't use it precisely because of that incredibly bad decision in language architecture. That alone kills the third party library market, unless they make everything explicitly overridable. Considering most third party libraries don't even come with reasonable documentation to make even the simple samples work....
To me it seems as if no real design went into VB.Net in contrast to C# which seems like a lot of thought went into how to do things and how not to do things.
Most of that "design" and "thought" was "how did Sun do it with Java". And they still borked it up with the explicit override, virtual, and new keywords for methods. What's the point of an OO language's polymorphism if you have to go back and change the super classes in a third party library if you need to override a method? (And yes, I'm aware that C++ follows the same paradigm - who said they were right?)
Interestingly enough, as I was OC'ing my X2 3800+ yesterday, it's done via what's effectively FSB. Bump the CPU frequency up 10%, and HTT and memory gets bumped 10%. Makes for cool OC options though, as they're independently settable. Now if I could only find DDR 480 or DDR 500 RAM...:)
I'm kidding, DDR2 would be the arena for that. Remarkably, this cheap (ECS) motherboard handles the 10% bump without issue, the 20% bump causes it to require a hard reboot though, hangs otherwise, even if you lower the memory bandwidth. Guess the HTT speeds are too high. I haven't messed with lowering the multiplier yet, that's next if this board/CPU combo let me do that. The CPU runs under 50C @ 2.4GHz with superPi.
ATI was bought by AMD. That was part of the reasoning for the all in one CPU. Imagine the price drop if you could just add a GPU core to an AMD 64 X2G chip. Heck, video RAM could just be another slot on the motherboard. Want to upgrade your video to 512MB? No problem, just buy a DIMM...
In truth, I must confess this isn't even that original a thought. There was speculation about AMD's multi-core approach as soon as they announced the architecture in that you could have a couple of general purpose CPU type cores, coupled with special DSP, GPU, or other cores, all working at HTT speeds.
I think that AI and physics co-processors have a better future as part of the CPU, rather than an add-in board. Perhaps as additional core(s) on an AMD processor, with full access and feedback to the CPU proper?
I agree that it'd be awesome if the mini had a PCIe slot. (AGP, come on...) However, where would you install the card? You're not buying a $600 pre-built minature box for a performance workstation. I will agree that the option to buy something like the mini in a mini tower format with at least a couple of PCI and at least 1 PCIe slot would be very nice.
As for drive expansion, try external FW drives - USB works too, but FW is awesome! Best yet, works with laptops and all your workstations. After resisting the external HD for a long time (more cables etc) but once I did, the freedom of just plugging in the HD into multiple systems and having everything just there is awesome. Not to mention the backup ability of external drives. Actually, what sold me was going FW, USB 2.0 just has too many issues.
Now, here's a good question - why would you need a new wireless card? 99.9% of us will never have the need to swap out a working wireless card. I will posit that this is a non-issue.
The Mac mini's CPU/Memory can be upgraded. Those are the 2 most important things for this class of machine.
The Mac Pro equivalent to a $300 PC? How do you figure? You can upgrade CPUs (it'd be nice if it came in a single CPU config with the option to add a second later). You can also upgrade RAM, the GPU, and hard drives. You can connect external components. What else do you want? A cool-cathode tube in a window?
I'm guessing from the rest of your post, you want the equiavalent of a single proc Mac Pro for around $500-600.
I understand your point. The title of the film aside, this is something hundreds or even thousands of divers do every day. From further stories on the topic, it appears that this incident/accident is exceedingly rare and just extraordinary bad luck. Check Boing-Boing for a summary of several sources, including Australian articles. They only have 2 recorded deaths by stingray, Steve Irwin is the third.
To trace this back to your analogy about driving, are you going to be hyper-alert for planes hitting your car? That happens more often than stingray attacks. I've personally seen the aftermath of just such a wreck only moments after it occurred, killing 5 people.
You can't be safe from everything. Now, in Steve Irwin's case, he courted danger. For him to die the way he did I propose is not too dissimilar from a motorcycle daredevil being hit by freak lightning on a golf course, or perhaps slipping between the platform and train of a subway. (I believe that happened in NY and Japan both, too lazy to go look it up at the moment.)
At the time he was just doing a simple dive. From the detective who reviewed the film of Steve Irwin swimming in about 6-7 feet of water.
"There is no evidence that Mr Irwin was intimidating or threatening the stingray," Queensland Police superintendent Michael Keating said. "My advice is that he was observing the stingray."
Earlier reports indicated that the stingray was covered with sand and that he may not even have seen it.
Check the story. He unknowingly swam over a buried stingray that reacted to his prescense with a defense mechanism. He was extremely unlucky in this case, and the magnitude of that bad luck is raised to ironic levels by his everday (for him) activities.
It's like a motorcycle daredevil getting hit by freak lightning on a clear day on the golf course.
Your statement vaguely reminds me of college days where several friends would attempt to insist to me that going to a party and getting drunk off my ass followed by puking and waking up with a huge hangover was "the fun part".
It is the fun part for everyone but you! If you puke and/or wake up with a hang-over, you've done it wrong and need more practice.
You're just an Intel fanboy. (Wow, that was fun, now that the name-calling's done, let's get down to real discussion)
Let's tackle your FSB assertion first. Woodcrest shares a FSB between 2 CPUs. To state that they can scale just by increasing the bus clock is the same argument used for the now thoroughly discredited P4. You can't keep increasing FSB without making trade-offs that will usually run into one wall or the other (heat or interference are the 2 main ones for FSB speed)
When did Intel get NUMA? That's pretty much what's required for CPUs to have their own FSB and still work together. (If you're only looking for virtualization type systems, this isn't important.)
The Sun T1 has 8 threads. Using 4 memory controllers is probably a good thing. I think you're confused with Intel chips. AMD chips have 1 controller per core. As for upgrading, when's the last time you upgraded a memory controller? Heck, when's the last time the memory controller was a problem in an AMD system?
Intel has double the cache, because that's the only way they can even be in the ballpark, until recently with the Core 2. As for L3, Intel dropped L3 because it was too expensive. Anything can be done, it's merely a question of cost. As for process tech, Intel's got 65nm right now, AMD competed with them on 90nm because of superior tech - SOI. When AMD hits 65nm, and Intel goes 45nm, I think you'll find AMD's performance will still be neck and neck at best for Intel. Why @ best? Because Intel's got nothing else up their sleeve.
Re:Just like there will never be another Doom
on
Can Anyone Beat WoW?
·
· Score: 1
You're still working under the current concept of a MMOG. There are other ways to deal with these issues, including the one I drafted in 93-94, when I was coding for one of the more successful MUD's. The draft addressed 3 main concerns, including how to increase the number of simultaneous players connected, how to decrease lag, and how to provide areas for that increased number of players. The architecture was drawn up that addressed all these concerns, but things like marriage, births, family, and job changes (ie, life) got in the way of us actually coding the changes required before MUDs faded.
Let's say that the solution required thinking way outside the box, and was addressed by technology that only now is being discussed as "new" and "cutting edge" for the gaming industry.
They're behind? Until Intel's just released Core 2 chips, Intel was a generation behind. AMD will be releasing their next gen in about 6-9 months. Intel's next gen won't be out for a long while. Don't forget, Core 2 is a fallback to the old P-III/M architecture utilizing a shared cache. It's Intel's Plan B, and they're still scrambling. Furthermore, initial tests on Woodcrest indicate that Intel's already hit the FSB wall with a mere 2P system.
AMD will be utilizing a shared L3 cache on their next CPUs, don't suffer from FSB issues, and are pretty effectively infinitely scalable in number of processors (and cores, for that matter). I'd say Intel's toast. I'll give them their last hurrah for these 6-9 months though. Core 2 has the crown.
In single threaded apps, I think AMD will at least be equal to Core 2 with their 65nm K8L architecture.
In the server realm, Intel's a has-been. Their market share's falling faster than a rock through air. If you need proof, check out who all announced AMD powered servers recently (HP, IBM, Sun, and even Dell).
As for the quad performance numbers, I think Intel's going to wilt when AMD's quads come out and are tested against multi-threaded benchmarks.
Well damn, looking up the specs and doing some quick back of the napkin calcs, you could almost put a whole Lord of the Rings episode on a single layer disk in full HD on Blu-ray. So an AC actually had some real value add.
BTW, H.264 is MPEG4, which is about 8-10 times higher compression of the stream than standard MPEG 2 for a set quality. It's still lossy.
Re:Just like there will never be another Doom
on
Can Anyone Beat WoW?
·
· Score: 1
Sure they could...;) w/ 1 server or 1 system, it doesn't matter.
Slashdot Mirror
Well, here's a key on exceptions - don't use them. :)
I know that's a trivial statement, but seriously, I've worked with well-written code that only used existing exceptions where absolutely necessary, had few try-catch blocks, and would almost never throw a RunTime exception because data was validated before being operated on.
Then I had the counter-experience: a massively overly complicated "Exception Framework" complete with layered translation layers all spec'd out via Spring. Why? Because some nimrod read about exception frameworks being the next best thing in some drivel rag, and gave the struts exception framework as an example. What happened? We had individual exception classes written for every exception, complete with translator classes etc. Over 200 classes for something that at most required a single class with a data member indicating type of application exception. I say at most, because many of these were potential NPEs and the like that could have been checked and handled gracefully or via the default exceptions, but instead were layered in a "framework" that made debugging virtually impossible, as the stacktrace was lost through the translation layers - all without logging, of course. I did state that this was a nimrod's brain fart.
Funny.
AMD X2 3800+ with X1900XT. Not planning on buying a new one this year. Maybe next year I'll replace the CPU/MB with the quad AM2 CPU. Maybe.
It needs to be able to display HD content over the component inputs on my HDTV monitor.
It needs to render on my PC. The one I have now.
If not, I won't be buying it.
Well, we can start with MS implementation and usage of the FS, which is basically "designed to fragment" from the get go. Then continue on with the archaic start at cylinder 1 (effectively) mode of writing. Then continue with absolutely no optimization of file writes when file sizes are known (part of that inherently fragmenting design).
As to safely recover from power failure, you have to be kidding me. I can't count the number of times I've had file corruption on my NTFS drives after a hard reset due to a lockup or failure to shutdown properly. (power failures generally aren't an issue - UPSes deal with that issue) I've even lost 3 drive's contents this way, separately. No they weren't RAIDED.
I don't know how well XFS, JFS, or ReiserFS deal with those, but my ext2/3 boxes seem to have no problems. ReiserFS when I used it had other issues, especially under heavy load. To be fair, it wasn't completed and tested at the time, being a RC release.
I was speaking of my personal experiences there. I can't vouch for the workstation aspect, as I don't use it as such at this time.
btw - Win 2K server and even the workstation is pretty much considered the zenith of MS OSes as far as reliability, performance, and general architecture goes. Everything since has been a bit of a downhill slide down one of the criteria for good reliable OSes.
I started some work with C# once, and quickly decided that given the choice, I wouldn't use it precisely because of that incredibly bad decision in language architecture. That alone kills the third party library market, unless they make everything explicitly overridable. Considering most third party libraries don't even come with reasonable documentation to make even the simple samples work....
I can add a third and a fourth:
OS/2 - HPFS was a great approach to a filesystem.
Linux - it just runs and runs, although it's a server, not a workstation.
MS could replicate these successes by junking NTFS and getting a real file system.
Most of that "design" and "thought" was "how did Sun do it with Java". And they still borked it up with the explicit override, virtual, and new keywords for methods. What's the point of an OO language's polymorphism if you have to go back and change the super classes in a third party library if you need to override a method? (And yes, I'm aware that C++ follows the same paradigm - who said they were right?)
What's the problem? Just write in them starting from the back. It'll be backwards for us righties, but then, so is your writing style ;)
Interestingly enough, as I was OC'ing my X2 3800+ yesterday, it's done via what's effectively FSB. Bump the CPU frequency up 10%, and HTT and memory gets bumped 10%. Makes for cool OC options though, as they're independently settable. Now if I could only find DDR 480 or DDR 500 RAM... :)
I'm kidding, DDR2 would be the arena for that. Remarkably, this cheap (ECS) motherboard handles the 10% bump without issue, the 20% bump causes it to require a hard reboot though, hangs otherwise, even if you lower the memory bandwidth. Guess the HTT speeds are too high. I haven't messed with lowering the multiplier yet, that's next if this board/CPU combo let me do that. The CPU runs under 50C @ 2.4GHz with superPi.
ATI was bought by AMD. That was part of the reasoning for the all in one CPU. Imagine the price drop if you could just add a GPU core to an AMD 64 X2G chip. Heck, video RAM could just be another slot on the motherboard. Want to upgrade your video to 512MB? No problem, just buy a DIMM...
In truth, I must confess this isn't even that original a thought. There was speculation about AMD's multi-core approach as soon as they announced the architecture in that you could have a couple of general purpose CPU type cores, coupled with special DSP, GPU, or other cores, all working at HTT speeds.
I think that AI and physics co-processors have a better future as part of the CPU, rather than an add-in board. Perhaps as additional core(s) on an AMD processor, with full access and feedback to the CPU proper?
I agree that it'd be awesome if the mini had a PCIe slot. (AGP, come on...) However, where would you install the card? You're not buying a $600 pre-built minature box for a performance workstation. I will agree that the option to buy something like the mini in a mini tower format with at least a couple of PCI and at least 1 PCIe slot would be very nice.
As for drive expansion, try external FW drives - USB works too, but FW is awesome! Best yet, works with laptops and all your workstations. After resisting the external HD for a long time (more cables etc) but once I did, the freedom of just plugging in the HD into multiple systems and having everything just there is awesome. Not to mention the backup ability of external drives. Actually, what sold me was going FW, USB 2.0 just has too many issues.
Now, here's a good question - why would you need a new wireless card? 99.9% of us will never have the need to swap out a working wireless card. I will posit that this is a non-issue.
The Mac mini's CPU/Memory can be upgraded. Those are the 2 most important things for this class of machine.
The Mac Pro equivalent to a $300 PC? How do you figure? You can upgrade CPUs (it'd be nice if it came in a single CPU config with the option to add a second later). You can also upgrade RAM, the GPU, and hard drives. You can connect external components. What else do you want? A cool-cathode tube in a window?
I'm guessing from the rest of your post, you want the equiavalent of a single proc Mac Pro for around $500-600.
I understand your point. The title of the film aside, this is something hundreds or even thousands of divers do every day. From further stories on the topic, it appears that this incident/accident is exceedingly rare and just extraordinary bad luck. Check Boing-Boing for a summary of several sources, including Australian articles. They only have 2 recorded deaths by stingray, Steve Irwin is the third.
To trace this back to your analogy about driving, are you going to be hyper-alert for planes hitting your car? That happens more often than stingray attacks. I've personally seen the aftermath of just such a wreck only moments after it occurred, killing 5 people.
You can't be safe from everything. Now, in Steve Irwin's case, he courted danger. For him to die the way he did I propose is not too dissimilar from a motorcycle daredevil being hit by freak lightning on a golf course, or perhaps slipping between the platform and train of a subway. (I believe that happened in NY and Japan both, too lazy to go look it up at the moment.)
Earlier reports indicated that the stingray was covered with sand and that he may not even have seen it.
Check the story. He unknowingly swam over a buried stingray that reacted to his prescense with a defense mechanism. He was extremely unlucky in this case, and the magnitude of that bad luck is raised to ironic levels by his everday (for him) activities.
It's like a motorcycle daredevil getting hit by freak lightning on a clear day on the golf course.
It is the fun part for everyone but you! If you puke and/or wake up with a hang-over, you've done it wrong and need more practice.
You're just an Intel fanboy. (Wow, that was fun, now that the name-calling's done, let's get down to real discussion)
Let's tackle your FSB assertion first. Woodcrest shares a FSB between 2 CPUs. To state that they can scale just by increasing the bus clock is the same argument used for the now thoroughly discredited P4. You can't keep increasing FSB without making trade-offs that will usually run into one wall or the other (heat or interference are the 2 main ones for FSB speed)
When did Intel get NUMA? That's pretty much what's required for CPUs to have their own FSB and still work together. (If you're only looking for virtualization type systems, this isn't important.)
The Sun T1 has 8 threads. Using 4 memory controllers is probably a good thing. I think you're confused with Intel chips. AMD chips have 1 controller per core. As for upgrading, when's the last time you upgraded a memory controller? Heck, when's the last time the memory controller was a problem in an AMD system?
Intel has double the cache, because that's the only way they can even be in the ballpark, until recently with the Core 2. As for L3, Intel dropped L3 because it was too expensive. Anything can be done, it's merely a question of cost. As for process tech, Intel's got 65nm right now, AMD competed with them on 90nm because of superior tech - SOI. When AMD hits 65nm, and Intel goes 45nm, I think you'll find AMD's performance will still be neck and neck at best for Intel. Why @ best? Because Intel's got nothing else up their sleeve.
You're still working under the current concept of a MMOG. There are other ways to deal with these issues, including the one I drafted in 93-94, when I was coding for one of the more successful MUD's. The draft addressed 3 main concerns, including how to increase the number of simultaneous players connected, how to decrease lag, and how to provide areas for that increased number of players. The architecture was drawn up that addressed all these concerns, but things like marriage, births, family, and job changes (ie, life) got in the way of us actually coding the changes required before MUDs faded.
Let's say that the solution required thinking way outside the box, and was addressed by technology that only now is being discussed as "new" and "cutting edge" for the gaming industry.
They're behind? Until Intel's just released Core 2 chips, Intel was a generation behind. AMD will be releasing their next gen in about 6-9 months. Intel's next gen won't be out for a long while. Don't forget, Core 2 is a fallback to the old P-III/M architecture utilizing a shared cache. It's Intel's Plan B, and they're still scrambling. Furthermore, initial tests on Woodcrest indicate that Intel's already hit the FSB wall with a mere 2P system.
AMD will be utilizing a shared L3 cache on their next CPUs, don't suffer from FSB issues, and are pretty effectively infinitely scalable in number of processors (and cores, for that matter). I'd say Intel's toast. I'll give them their last hurrah for these 6-9 months though. Core 2 has the crown.
In single threaded apps, I think AMD will at least be equal to Core 2 with their 65nm K8L architecture.
In the server realm, Intel's a has-been. Their market share's falling faster than a rock through air. If you need proof, check out who all announced AMD powered servers recently (HP, IBM, Sun, and even Dell).
As for the quad performance numbers, I think Intel's going to wilt when AMD's quads come out and are tested against multi-threaded benchmarks.
that still doesn't explain how though.
Well damn, looking up the specs and doing some quick back of the napkin calcs, you could almost put a whole Lord of the Rings episode on a single layer disk in full HD on Blu-ray. So an AC actually had some real value add.
BTW, H.264 is MPEG4, which is about 8-10 times higher compression of the stream than standard MPEG 2 for a set quality. It's still lossy.
Sure they could... ;) w/ 1 server or 1 system, it doesn't matter.