The company I work for sells far too many AIT, DDS and DLT drives than is healthy for the customers, but they just don't see the importance upgrading, as you say.
However, disk is expensive. Even iSCSI. Contrary to popular belief, tape is not dead. Tape is also a tenth to a fifth of the cost of disk. Adding more drive arrays is far less cost effective than tape.
Co-lo failover systems have their place, but trying to run a backup over the internet is going to be painful. A single LTO3 drive writes at about 70MB/sec, that's 560 Megabits/sec. Effectively, an OC-12. Those aren't cheap. I regularly deal with tape libraries that can hold 20 to 24 LTO3 drives. That's a lot of bandwidth. Granted, your SAN better have the balls to keep up, or you've got bigger problems.
An L700 equipped with 20 LTO3's can backup half a petabyte (500TB) in about seven and a half hours. And, no matter what Hollywood may think, four hard drives in a briefcase, just won't cut it.;-)
Okay, so first, I was merely offering another alternative. There are other free backup software solutions out there.
And second, how is using ntbackup any more crazy than using 212 DVD's to backup 1TB?
Remember when we used to laugh at how ridiculous it was that MS Office 4.3 came on 27 floppies? What you are suggesting is far worse. Face it, for four hundred bucks, anything you do will be far from ideal.
I would still rather deal with 5 to 10 tapes rather than 200+ DVD's.
Um. No. SDLT still sucks. We test them all day. And we hateses them. They are cranky, quirky, extremely heat sensitive and slow. Yes, they are better than DLT, but LTO still kicks their ass six ways from Sunday.
LTO Gen 1 xfer 14MB/sec LTO Gen 2 xfer 28-33MB/sec - yes, we actually fail/repair LTO2's for less than 27MB/s xfer rate. LTO Gen 3 xfer 68-75MB/sec LTO Gen 4 xfer 107-113MB/sec
SDLT320's xfer 10-12MB/sec SDLT600's xfer 22-24MB/sec - and will overheat 30sec from power on without active cooling! DLT/S-4's xfer dunno.
You can find used LTO2's on ebay between $500 and $600, if you dig deep enough. Retail is for consumers.
You mean, like this one? Say, that would be about the same price as a DVD burner! Hmmm.:-P
As for backup software, there are a number of FOSS solutions available if you're running *nix. And then there's always ntbackup, which comes free with 2K, XP, Server, etc. Actually, it may go back as far as NT 3.51, but I don't have a machine still running that to check.
A single tape drive doesn't need that much. If you've got a tape library, then things get a little more problematic.
I agree that online failover is good business practice. But it's not archiving, which is, literally, copying the data and storing it somewhere. I was arguing against his definition of archiving.
No! DDS-3 is the devil! And DLT is at least a lower demon.
DDS drives are the very reason why tape has such a bad reputation. They were junk when they were brand new and $3K. Then again, there was Exabyte. And QIC. And AIT.
Okay, but he's talking about archiving 100GB, which would definitely put him in the MozyPro category.
One, it's $50/mo for 100GB, not per year. And also, assuming he can actually get 1.54MB/sec upstream, it will take over 18 hours to upload. If he's getting closer to 150K/sec, it's more like 180 hours (7.5 days).
When you're talking about accounting data, inventory, policy manuals and other small footprint important stuff, then Mozy is a good idea. Until they get hacked.
Okay, so I'm having trouble understanding the solution you propose. Let me see if I get this right...
You should "constantly archive" which means that you regularly make copies of your data for long term storage (preferably off-site). But you shouldn't use any removable media. So you're left with Hard Drives as your only option (remember, USB keys are still removable storage).
Did I get that right? Or did you skip the idea of off-site storage all together?
What "formats you will always be able to use" are there? Hard drives? And what makes hard drives so special? They still require an interface to read the data. Someone posted that IDE has been around over 20 years. That's great, but it's also currently end of life. How long is SATA going to last? Fibre Channel? Hopefully your Linux distro in 2030 will still be able to read NTFS5 partitions. Hard drives are heavy and delicate.
Or are you talking about replicating your entire system to ensure that you can read the data? And then shipping the whole server to Iron Mountain? Heck, while you're at it you might as well plug the thing in, give it an internet connection and set it up for on-line fail-over.
Or maybe you have a different definition of "archive" than the rest of us.
DLT is considered end of life tech in the tape world. For the VS-160, the transfer rate is very poor (~6MB/s) compared to even LTO Gen 1 drives (~14MB/sec).
SDLT drives are nearly as unreliable as their ancestors and extremely sensitive to heat (your case fan dies and the drive is unusable). And, again, slower than the competing LTO drives.
A used LTO2 drive may be the sweet spot (200GB native per tape and 28-30MB/s transfer rate). I prefer HP and IBM drives over Certance/Quantum drives and half-height drives run at half speed. Even with a SCSI card (U160 or U320) you may be able to get set up for less than the $700 mentioned above.
Or you could always splurge for LTO3 (400GB native) or LTO4 (800GB native). I'm not listing the compressed stats (usually double) because video (usually already compressed) doesn't compress easily.
Or maybe emotions are naturally emergent properties of any sufficiently advanced neural network. I believe that that is indeed the case. Or more specifically, emotion is an emergent behavior of any sufficiently advanced self-aware neural network.
Self-awareness is feedback loops, lots and lots of feedback loops: about the status of the components of the system fed back into the system itself.
That is a very good explanation of neural nets and their history. Thank you.
I'd say that a lot of AGI research is to discover/understand/model all the "management functions" to that mass of neural nets crammed in our heads. My thought is that neural nets are mostly complex storage, rather than the system itself. Then again, I consider myself just a dilettante in the field of AI, so I could be full of complete crap.
I agree that a neural net, no matter how big, does not an intelligence make.
I have to disagree. I've read a fair chunk of Hofstadter and I've gained an appreciation for just how uber/ultra/mega-complex cognition really is. He's not introducing mysteries, but taking the time to point them out and attempt to define/explain those mysteries.
Humans have been attempting to define this consciousness/self-awareness/sentience/cognition/intelligence/"I"/spirit/soul/smarts/ubik since we've had this je ne sais quoi. Many of us have our own crackpot ideas on what that is and I would say mine are relatively close to Hofstadter's. But we haven't reached a point where we are actively testing the myriad of hypotheses yet. Mostly because, the mind, well, boggles the mind.
Kurzweil is not any more objective than the rest of us. Brilliant, yes. But, then again, so is Hofstatder. Of course, I'm not exactly objective on that opinion.;-)
AI's exist in a perfectly designed environment, they have humans feed them power & data and all they need to do is process. I think that if all the "system" does is process, it's not AI. As long as a system is merely condition/response, it can't be sentient. You must have a system that seeks new data and interaction, but does not require that input in order to perpetuate its existence/behavior/whatever. When the loops in the AI's mind cease to form new connections/concepts/ideas (it gets bored), it will look for new stimulus (goes out to play, calls up a friend, reads a book, etc).
The problem with a system like that, is that it is hard to tell if it's broken, or just doesn't want to talk.
So, I guess this would be part of my definition of AI (or AGI, Artificial General Intelligence), but certainly not the whole thing.
I see facial recognition and natural language processing as the precursors to sensory modalities for AGI's, rather than the core(s) of AGI's. Still quite important, though.
Public transportation is a pain here in Phoenix, AZ, US. It takes me 2.5hrs to get to work by bus and 3 hrs to get home. If I drive, it takes 25 minutes each way. I work 19 miles from home. Thankfully, I live in Central Phoenix and work in the suburbs, so I'm going opposite of most traffic. I'd move, but the difference in rent (getting a great deal, 10+ yrs & few hikes) might not offset the savings in gas. Especially since I've got friends all over the Valley.
From the east end to west end, the Phoenix Metro Area spans more than 100 miles (~160km). You have to go pretty far just to get to "rural".
solar power -> through existing electric infrastructure -> to the battery of your electric car/mower/series of tubes 90% loss -> 50% loss -> 50% loss = 2.5% efficiency.
Solar Panels: 22% efficiency SunPower Electrical Transmission: 92.8% efficiency (by jtoomim (217124) Alter Relationship on Thursday June 12, @06:10AM (#23761407)) Battery Charging: 85% efficiency John W. Stevens and Garth P. Corey Electric Motor: 90% efficiency
78% loss -> 7.2% loss -> 15% loss -> 10% loss = 15.6% efficiency, not 2.5%.
If you don't factor in the "loss" in the efficiency of the panel to collect the rather amazing amount of energy produced by the sun (of which the Earth only intercepts a tiny fraction), the efficiency is in the 70% range.
I don't know the exact conversion ratio here, but I do know charging batteries is a very, very inefficient process.
Numbers Here. Actually, if you can keep your total charge to 90% of the battery's rated capacity, you can easily get 85% efficiency. And that's on Lead-Acid. Lithium-ion batteries are even more efficient. It all comes down to proper power management.
It's still better than 20-30% efficiency on a combustion engine.
It all depends on what you consider a reasonable range for a car. Based on what time period? An hour? A day? A week?
A Tesla roadster is supposed to get about 200 miles per charge. It's storage capacity is 53KW/h. At the current rates I'm paying for electricity, that's about $5 per charge.
Based on my current milage (19 miles to work, friends all over the Phoenix Valley), I would need to charge the car 5 times a month. So, $25/month to keep it running. And it's a freakin' sports car!
Compare that to $280 a month I'm paying now for a vehicle that gets 19mpg.
That's a net savings of $3060 a year. That savings can rent me a really nice gas-powered vehicle for my yearly road-trip to Coachella (and my friend will pay for half the gas!). All this because I'm willing to plug my EV in five nights a week.
Granted, the savings per year doesn't justify buying a $100K+ sports car, but then again the Aveta at $29K is starting to look pretty cool.
My complaint about corn-based ethanol is its lack of efficiency. That's it.
Ethanol production efficiency (courtesy of wikipedia): Corn, 320/420 Gallons per acre (there's some inconsistency in the numbers) Cane, 800 Gallons per acre
Sugarcane/sugarbeets are far more efficient crops if your goal is creating ethanol. The other interesting thing about cane-ethanol, is that the 800 gallons per acre is derived after the table sugar and molasses are removed through normal processing. I don't know if this is also true for sugar beets.
Now, once we get the kinks worked out of cellulosic conversion, the corn efficiency goes way up and my complaint about it goes away. We may even be able to harvest the kernals for normal use and convert the remaining plant into ethanol.
It also brings our lovable (and on topic!) switch-grass into the mix as well, which can be grown and harvested on marginal land.
Although, after all's said and done, I think solar is ultimately the way to go.
Okay, dude. You really need to take a good look at your power bills.
Here's mine electric usage here in Phoenix as an example. In the past year, my monthly KWh usage has varied between 891 (Mar '08) and 2183 (Sep '07).
On average, I run put about 200 miles a week on my car, so charging up once a week is about right. At 50KWh a week, for a total of 200KWh a month, it's really not that big an addition to my electric bill. Based on the time of year, it ends up being a 10-20% larger bill.
Strictly based on the average price I pay for electricity, that equates to about $20 a month to put 800 miles on my car, instead of ~$200 a month right now. In actuality, it would be cheaper for me, since I am on a flex plan and am charged less (I think about $.07 KWh) at night.
Unfortunately, that $2160 of savings each year isn't enough to justify buying a >$100,000 car. I still want one, though.
Oh yeah. IIRC, one of the requirements to buying one of these beauties is that you need have a 480V connection installed in your house to charge the car. Mostly to speed up the charging process.
One of the items of note that I gleaned from other forums was that Creative currently has ZERO products currently on the market that run fully in Vista. All those Vista Certified cards are really Vista Crippled. It's kind of hard to get people to upgrade when you don't have anything for them to upgrade to.
The only thing that I can think of is that it's a temper tantrum at MS because Vista is different and they don't want to have to work that hard.
They probably mean "on the songs I like, power chords are often played at the same time as loud bass and bass drums".
I think you're right. For me, I've always called it the "kick". It's the burst from all the instruments after the solo in Yes's "Roundabout"; the concussive blast after the intro in "Name of the Game" by The Crystal Method.
Compression (both types) shove everything up into the higher range, so that when you want that POW! there is simply nothing left. It's like watching a Michael Bay movie and seeing a massive budget-breaking explosion and yet hearing a duck fart.
I noticed the lack of punch in "Name of the Game" at a DJ gig a few years ago. Most of what we play is uncompressed, but the only version we happened to have at the time was.mp3 (192/48K). The kick in that song is awesome at concert volume: ass-spankin', baby-slappin', window-shakin' kinda cool. Blade II had just come out and I couldn't wait for that kick to hit the crowd... But it sounded like static. No edge, no power, it just wasn't that great.
The disappointment in that one tiny little duck fart led me to re-evaluate the codec I was using to store all my music. What I discovered, was that.mp3 could not reproduce that Crystal Method kick, regardless of the bitrate or codec. Even at 320 with 48KHz over-sampling. It always sounded like static..ogg, on the other hand, reproduced it perfectly at 192. With a decent pair of headphones, I couldn't tell the difference between.ogg and.wav. So now, most of my library has been re-ripped and converted to.ogg.
For me, it is not that compression handles music well in most cases, but that when you need to push it to eleven, it's not there.
Ultimately, music is about emotion and contrast. When you flatten the recording, you flatten the emotion.
Slashdot Mirror
The company I work for sells far too many AIT, DDS and DLT drives than is healthy for the customers, but they just don't see the importance upgrading, as you say.
However, disk is expensive. Even iSCSI. Contrary to popular belief, tape is not dead. Tape is also a tenth to a fifth of the cost of disk. Adding more drive arrays is far less cost effective than tape.
Co-lo failover systems have their place, but trying to run a backup over the internet is going to be painful. A single LTO3 drive writes at about 70MB/sec, that's 560 Megabits/sec. Effectively, an OC-12. Those aren't cheap. I regularly deal with tape libraries that can hold 20 to 24 LTO3 drives. That's a lot of bandwidth. Granted, your SAN better have the balls to keep up, or you've got bigger problems.
An L700 equipped with 20 LTO3's can backup half a petabyte (500TB) in about seven and a half hours. And, no matter what Hollywood may think, four hard drives in a briefcase, just won't cut it. ;-)
Okay, so first, I was merely offering another alternative. There are other free backup software solutions out there.
And second, how is using ntbackup any more crazy than using 212 DVD's to backup 1TB?
Remember when we used to laugh at how ridiculous it was that MS Office 4.3 came on 27 floppies? What you are suggesting is far worse. Face it, for four hundred bucks, anything you do will be far from ideal.
I would still rather deal with 5 to 10 tapes rather than 200+ DVD's.
Um. No. SDLT still sucks. We test them all day. And we hateses them. They are cranky, quirky, extremely heat sensitive and slow. Yes, they are better than DLT, but LTO still kicks their ass six ways from Sunday.
LTO Gen 1 xfer 14MB/sec
LTO Gen 2 xfer 28-33MB/sec - yes, we actually fail/repair LTO2's for less than 27MB/s xfer rate.
LTO Gen 3 xfer 68-75MB/sec
LTO Gen 4 xfer 107-113MB/sec
SDLT320's xfer 10-12MB/sec
SDLT600's xfer 22-24MB/sec - and will overheat 30sec from power on without active cooling!
DLT/S-4's xfer dunno.
You can find used LTO2's on ebay between $500 and $600, if you dig deep enough. Retail is for consumers.
You mean, like this one? Say, that would be about the same price as a DVD burner! Hmmm. :-P
As for backup software, there are a number of FOSS solutions available if you're running *nix. And then there's always ntbackup, which comes free with 2K, XP, Server, etc. Actually, it may go back as far as NT 3.51, but I don't have a machine still running that to check.
A single tape drive doesn't need that much. If you've got a tape library, then things get a little more problematic.
I agree that online failover is good business practice. But it's not archiving, which is, literally, copying the data and storing it somewhere. I was arguing against his definition of archiving.
No! DDS-3 is the devil! And DLT is at least a lower demon.
DDS drives are the very reason why tape has such a bad reputation. They were junk when they were brand new and $3K. Then again, there was Exabyte. And QIC. And AIT.
Trust me. LTO is your friend.
Ebay listing
No, I am not related to the party above, it was just the first listing that caught my eye and matched (mostly) your criteria.
I'd pick LTO over DVD any day. But, then again, tape drives are my day job.
Okay, but he's talking about archiving 100GB, which would definitely put him in the MozyPro category.
One, it's $50/mo for 100GB, not per year. And also, assuming he can actually get 1.54MB/sec upstream, it will take over 18 hours to upload. If he's getting closer to 150K/sec, it's more like 180 hours (7.5 days).
When you're talking about accounting data, inventory, policy manuals and other small footprint important stuff, then Mozy is a good idea. Until they get hacked.
Okay, so I'm having trouble understanding the solution you propose. Let me see if I get this right...
You should "constantly archive" which means that you regularly make copies of your data for long term storage (preferably off-site). But you shouldn't use any removable media. So you're left with Hard Drives as your only option (remember, USB keys are still removable storage).
Did I get that right? Or did you skip the idea of off-site storage all together?
What "formats you will always be able to use" are there? Hard drives? And what makes hard drives so special? They still require an interface to read the data. Someone posted that IDE has been around over 20 years. That's great, but it's also currently end of life. How long is SATA going to last? Fibre Channel? Hopefully your Linux distro in 2030 will still be able to read NTFS5 partitions. Hard drives are heavy and delicate.
Or are you talking about replicating your entire system to ensure that you can read the data? And then shipping the whole server to Iron Mountain? Heck, while you're at it you might as well plug the thing in, give it an internet connection and set it up for on-line fail-over.
Or maybe you have a different definition of "archive" than the rest of us.
DLT is considered end of life tech in the tape world. For the VS-160, the transfer rate is very poor (~6MB/s) compared to even LTO Gen 1 drives (~14MB/sec).
SDLT drives are nearly as unreliable as their ancestors and extremely sensitive to heat (your case fan dies and the drive is unusable). And, again, slower than the competing LTO drives.
A used LTO2 drive may be the sweet spot (200GB native per tape and 28-30MB/s transfer rate). I prefer HP and IBM drives over Certance/Quantum drives and half-height drives run at half speed. Even with a SCSI card (U160 or U320) you may be able to get set up for less than the $700 mentioned above.
Or you could always splurge for LTO3 (400GB native) or LTO4 (800GB native). I'm not listing the compressed stats (usually double) because video (usually already compressed) doesn't compress easily.
Self-awareness is feedback loops, lots and lots of feedback loops: about the status of the components of the system fed back into the system itself.
That is a very good explanation of neural nets and their history. Thank you.
I'd say that a lot of AGI research is to discover/understand/model all the "management functions" to that mass of neural nets crammed in our heads. My thought is that neural nets are mostly complex storage, rather than the system itself. Then again, I consider myself just a dilettante in the field of AI, so I could be full of complete crap.
I agree that a neural net, no matter how big, does not an intelligence make.
I have to disagree. I've read a fair chunk of Hofstadter and I've gained an appreciation for just how uber/ultra/mega-complex cognition really is. He's not introducing mysteries, but taking the time to point them out and attempt to define/explain those mysteries.
;-)
Humans have been attempting to define this consciousness/self-awareness/sentience/cognition/intelligence/"I"/spirit/soul/smarts/ubik since we've had this je ne sais quoi. Many of us have our own crackpot ideas on what that is and I would say mine are relatively close to Hofstadter's. But we haven't reached a point where we are actively testing the myriad of hypotheses yet. Mostly because, the mind, well, boggles the mind.
Kurzweil is not any more objective than the rest of us. Brilliant, yes. But, then again, so is Hofstatder. Of course, I'm not exactly objective on that opinion.
The problem with a system like that, is that it is hard to tell if it's broken, or just doesn't want to talk.
So, I guess this would be part of my definition of AI (or AGI, Artificial General Intelligence), but certainly not the whole thing.
I see facial recognition and natural language processing as the precursors to sensory modalities for AGI's, rather than the core(s) of AGI's. Still quite important, though.
D'oh!
Yeah, I meant Aptera.
Public transportation is a pain here in Phoenix, AZ, US. It takes me 2.5hrs to get to work by bus and 3 hrs to get home. If I drive, it takes 25 minutes each way. I work 19 miles from home. Thankfully, I live in Central Phoenix and work in the suburbs, so I'm going opposite of most traffic. I'd move, but the difference in rent (getting a great deal, 10+ yrs & few hikes) might not offset the savings in gas. Especially since I've got friends all over the Valley.
From the east end to west end, the Phoenix Metro Area spans more than 100 miles (~160km). You have to go pretty far just to get to "rural".
90% loss -> 50% loss -> 50% loss = 2.5% efficiency.
Solar Panels: 22% efficiency SunPower
Electrical Transmission: 92.8% efficiency (by jtoomim (217124) Alter Relationship on Thursday June 12, @06:10AM (#23761407))
Battery Charging: 85% efficiency John W. Stevens and Garth P. Corey
Electric Motor: 90% efficiency
78% loss -> 7.2% loss -> 15% loss -> 10% loss = 15.6% efficiency, not 2.5%.
If you don't factor in the "loss" in the efficiency of the panel to collect the rather amazing amount of energy produced by the sun (of which the Earth only intercepts a tiny fraction), the efficiency is in the 70% range.
Numbers Here.
Actually, if you can keep your total charge to 90% of the battery's rated capacity, you can easily get 85% efficiency. And that's on Lead-Acid. Lithium-ion batteries are even more efficient. It all comes down to proper power management.
It's still better than 20-30% efficiency on a combustion engine.
Okay, but...
It all depends on what you consider a reasonable range for a car. Based on what time period? An hour? A day? A week?
A Tesla roadster is supposed to get about 200 miles per charge. It's storage capacity is 53KW/h. At the current rates I'm paying for electricity, that's about $5 per charge.
Based on my current milage (19 miles to work, friends all over the Phoenix Valley), I would need to charge the car 5 times a month. So, $25/month to keep it running. And it's a freakin' sports car!
Compare that to $280 a month I'm paying now for a vehicle that gets 19mpg.
That's a net savings of $3060 a year. That savings can rent me a really nice gas-powered vehicle for my yearly road-trip to Coachella (and my friend will pay for half the gas!). All this because I'm willing to plug my EV in five nights a week.
Granted, the savings per year doesn't justify buying a $100K+ sports car, but then again the Aveta at $29K is starting to look pretty cool.
Ethanol production efficiency (courtesy of wikipedia):
Corn, 320/420 Gallons per acre (there's some inconsistency in the numbers)
Cane, 800 Gallons per acre
Sugarcane/sugarbeets are far more efficient crops if your goal is creating ethanol. The other interesting thing about cane-ethanol, is that the 800 gallons per acre is derived after the table sugar and molasses are removed through normal processing. I don't know if this is also true for sugar beets.
Now, once we get the kinks worked out of cellulosic conversion, the corn efficiency goes way up and my complaint about it goes away. We may even be able to harvest the kernals for normal use and convert the remaining plant into ethanol.
It also brings our lovable (and on topic!) switch-grass into the mix as well, which can be grown and harvested on marginal land.
Although, after all's said and done, I think solar is ultimately the way to go.
Okay, dude. You really need to take a good look at your power bills.
Here's mine electric usage here in Phoenix as an example. In the past year, my monthly KWh usage has varied between 891 (Mar '08) and 2183 (Sep '07).
On average, I run put about 200 miles a week on my car, so charging up once a week is about right. At 50KWh a week, for a total of 200KWh a month, it's really not that big an addition to my electric bill. Based on the time of year, it ends up being a 10-20% larger bill.
Strictly based on the average price I pay for electricity, that equates to about $20 a month to put 800 miles on my car, instead of ~$200 a month right now. In actuality, it would be cheaper for me, since I am on a flex plan and am charged less (I think about $.07 KWh) at night.
Unfortunately, that $2160 of savings each year isn't enough to justify buying a >$100,000 car. I still want one, though.
Oh yeah. IIRC, one of the requirements to buying one of these beauties is that you need have a 480V connection installed in your house to charge the car. Mostly to speed up the charging process.
So, if you don't mind my asking, who did you go with instead of Xiotech?
Okay, somebody please correct me if I am wrong...
One of the items of note that I gleaned from other forums was that Creative currently has ZERO products currently on the market that run fully in Vista. All those Vista Certified cards are really Vista Crippled. It's kind of hard to get people to upgrade when you don't have anything for them to upgrade to.
The only thing that I can think of is that it's a temper tantrum at MS because Vista is different and they don't want to have to work that hard.
I think you're right. For me, I've always called it the "kick". It's the burst from all the instruments after the solo in Yes's "Roundabout"; the concussive blast after the intro in "Name of the Game" by The Crystal Method.
Compression (both types) shove everything up into the higher range, so that when you want that POW! there is simply nothing left. It's like watching a Michael Bay movie and seeing a massive budget-breaking explosion and yet hearing a duck fart.
I noticed the lack of punch in "Name of the Game" at a DJ gig a few years ago. Most of what we play is uncompressed, but the only version we happened to have at the time was .mp3 (192/48K). The kick in that song is awesome at concert volume: ass-spankin', baby-slappin', window-shakin' kinda cool. Blade II had just come out and I couldn't wait for that kick to hit the crowd... But it sounded like static. No edge, no power, it just wasn't that great.
The disappointment in that one tiny little duck fart led me to re-evaluate the codec I was using to store all my music. What I discovered, was that .mp3 could not reproduce that Crystal Method kick, regardless of the bitrate or codec. Even at 320 with 48KHz over-sampling. It always sounded like static. .ogg, on the other hand, reproduced it perfectly at 192. With a decent pair of headphones, I couldn't tell the difference between .ogg and .wav. So now, most of my library has been re-ripped and converted to .ogg.
For me, it is not that compression handles music well in most cases, but that when you need to push it to eleven, it's not there.
Ultimately, music is about emotion and contrast. When you flatten the recording, you flatten the emotion.
No big deal, just wait 'til Service Pack 2. Then it'll be fine.
Just don't burn your fingers on the new firewall. And hopefully the new blue screen in the TV studio won't kill you.