Client/Server works this way. Client requests data, server can grant or not grant access to said data. It's like having a stranger coming up to you and nicely asking(without malice or threats) if you'd like to give them something(money, cellphone, newspaper, the time, etc..) you can say "yes or no". The server granted these people access without them breaking the law. The server could almost be viewed as an extension of the company or under license from the company to make these decisions to stream or not stream. If they had a password on the stream and people cracked it or multiple people were sharing an account authorized for only one stream, then yes that would be against the rules. If it is a public stream going out to anyone who asks, it's pretty much fair game.
We're in a situation where 1GB of data costs more than 1kW of energy, in some cases it can be drastically more. If grandma is paying $55/mn for internet and only uses 2GB that's about $22.50 per GB which is pretty outrageous. She uses less, she still pays the same. Even someone maxing out their 250GB Comcast cap is paying $0.22 per GB, which is still quite expensive.
Frankly all utilities need to be monitored and broken up when needed to ensure there are multiple competitors. Also there should be huge efforts towards preventing carriers from also producing content as well. For example Comcast owns Comcast Sports Net and has been trying to extort higher than average fees from other non-comcast providers to carry it. Also Comcast's merger with NBC should raise huge flags. They'll own the line, the TV service & the content on that service... How does that not sound like a monopoly?
NoSQL is not just key-value lookups. Take a look at Redis or MongoDB, there are novel ideas in both of them & yes they do bring new things to the table. They are NOT memcache. I am also not sure people are "sacrificing a lot of what is present in an RDBMS" by choosing NoSQL over an RDBMS. I think your gripe is with people who don't know what the hell they're doing, but you project that griping on to NoSQL in general. There are some things that RDBMSs are really good at, there are some things RDBMSs aren't so great at. The huge majority of people in the NoSQL communities and the users of these solutions know that loading a million objects client-side and discarding all but 5 is stupid & no one would suggest that is a failing of either RDBMS or NoSQL solutions, but squarely on the user.
I would have to say that NoSQL is more relatable to how people think about objects and their relation to each other. People don't easily boil their objects down into relational tables and how each of those tables should interact with each other. This takes skill & talent & can be a bit of a pain to dive into which is why we have a bunch of ORM solutions which add another layer of cruft on top of RDBMSs. NoSQL is basically getting rid of the ORM & the tables(though some still use table-like structures). For apps that use that would normally use ORM(a lot of web apps) extensively that's great. For newbies who don't have years and years under their belt designing, tweaking normalization, sharding/partitioning it can be easier to pickup. Some of the NoSQL solutions have clustering/horizontal scaling and/or replication built-in, with no or very little schema/query changes required.
So for some NoSQL will be a better solution and for others a RDBMS will be a better solution. I wouldn't knock either. Just because you can do something in an RDBMS doesn't mean it's better than a NoSQL solution & visa versa.
This is session cookie hijacking, it could be used to spoof your Slashdot credentials just as much as someone's Facebook account. Someone just put "Social Network" in the headline to make it seem more hip. Cookie spoofing has been known since the invention of Cookies.
Can you post some data showing the guaranteed worst-case write latency for a top of the line 15k SAS drive or SAS drives in raid configuration? I found some data showing an Intel SSD having around 80ms worst-case write latency.
Even for many enterprise scenarios, this is not a huge deal & for consumers this is completely moot.
OK, first this isn't a "file level" issue, this is a physical disk/sector level issue. When you have all your data in 512-byte sectors and the majority of the time in normal usage your data is NOT sequentially laid out, you will incur random writes & reads all over the platters of the disk. It doesn't matter if all the files you use are over 1MB or over 512KB, what matters is how those files are physically laid out. If you have a 1MB file but it's spread out with 50% mostly sequential sectors and 50% random sectors, you will get a hit on load times with a conventional disk. Due to usage patterns and at least how NTFS writes data, fragmentation will cause a large majority of read & write operations to be random, regardless of the size of the file you're reading. Even if you're using a file system that suffers less from fragmentation, that doesn't mean you won't need to read randomly from large files such as from a database.
You can stick with your HDDs and in many usage scenarios that may make the most sense, but from a performance perspective, SSDs will usually outclass HDDs in random/write unless you're going to crazy scenarios that cost similar or more than the SSD.
Plus this article was about SSD for consumer users and they usually do not have 1TB of data or if they do need it they can get both a smaller SSD and larger hard drive. The SSD will provide better response and faster boot than their single consumer level hard disk.
Have you personally dealt with a scenario where an SSD did not write data within the time frame you needed it to? How badly did the SSD do? What brand of SSD was it and what application were you using?
Well that does lead one to ask a lot of questions, such as "how many HDD are in a rack", if a rack means 5 or 6 15k rpm HDD that were short-stroked(at this point whining about the cost of SSDs goes out the window). Maybe that can beat the SSD. But 6 or even 12 consumer level HDD used to their full capacity, cannot defeat a single SSD at random reads/writes:
Yes it does show that 4 short stroked SAS 15K RPM drives are beating out a single SSD by getting 2500 IOPs in the DB test. But those are older SSDs. Compare to newer SSDs in similar/same benchmark:
3.8ms for HDD vs 0.1ms for SSD. That's still a big difference.
You are correct about the low level nature of flash memory, but there are many ways to mitigate this. SSD controllers use multiple channels to read/write banks of flash memory. They have large internal buffers & they also have "waste space", meaning they have extra flash memory that the device can read/write to so as to not hold up the drive. SandForce controllers also use compression and other methods to enhance performance. With proper trim support you do not need to run garbage collection, thus it's really not much different in operation than how a HDD will act, just faster. No modern SSD will make you wait one second while it erases a block of data, they just don't work that way now.
Even worst case random write speeds outpace standard HDDs. Yes if you short stroke your 15k RPM and buy 6 of them then yipee maybe you can exceed the performance of a single SSD, but then any arguments about saving money buying 6x $400 drives + controller so you can use 5% of their capacity go out the window.
Even the 10K high-end pricey WD Velociraptor has a pathetic 0.8MB/throughput for random writes. Meanwhile a value level Intel SSD gets 35MB/sec throughput for random writes? How do you think putting 4 of those high end HDD together will make then faster than an SSD? The fact is they won't.
Plus if you think an SSD is a waste, why would you even entertain the suggestion of capping a 1TB HDD to 32GB? Do you have an benchmarks proving this is faster and really drops seeks to near zero?
Random write speed on an SSD is still faster than a hard drive. Even value level SSDs now have comparable sequential write speeds to standard HDD. I am not sure where you're getting your "a lot slower" data from.
You don't have to have a disk full of files 4K to benefit from an SSD. 4K random read/writes give a good indication for drive responsiveness and it's ability to handle multiple IO tasks at the same time. For example reading & displaying header info from 1,000 image files then producing 1,000 tiny thumbnails(something Windows 7 likes to do) is much quicker on a SSD than on conventional HDD. Also a HDD will struggle if you're trying to do more than one major task at a time. Things like writing cache files, updating Firefox SQLite database, file searches, are quicker. Additional boot times are reduced greatly thanks to SSD many people see 25% to 50% improvement.
As for the 12 disk raid vs the SSD. The SSD is still going to kick the RAIDs ass at random read/writes, sorry but it's true. Maybe you see no point to it and want to hold on to your precious HDD. That doesn't mean there isn't a use for SSDs right now as a worthy upgrade for around $125. It will do more for your boot time and general computer usage than buying 12 HDDs.
HDDs are still useful for story large amounts of data cheaply, but they are by no means fast & are not getting any faster.
Let's invent a buzzword for SSDs like "PowerStream Boost w/ Turbo AI", makes no fucking sense but people will gobble it up even if they have no clue what it really means. Ultimately SSDs just need to be marketed correctly to educate customers that there is a performance improvement and that you do not need the larger hard drive. A lot of consumers could probably even get by with a 64 or 128GB SSD. So just market it as "20,000 Operations Per Second!!!! Thanks to PowerStream Turbo. Stores up to 20,000+ music files." People might ask "hmm how many Operations can that hard disk do" and if they find out it's only a few hundred, that might swing their purchasing decision.
Modern SSD tech has enough cycles to let you use it for many many years. They are most useful right now as an OS drive to speed up your general computing. They are not suited for file server/backup duties. They work via regular SATA connectors, so I am not sure where you're getting the idea you have to buy a special memory module? Why would you have to replace the whole laptop if only your drive went bad? Why do you think that writing all zeros is a secure erase, newsflash, it's NOT. Also that is a very a-typical situation most consumers will never find themselves in.
Ok and how many random 4K write/read IOPs could you get even if you did RAID-1? A few hundred. How many can you get with an SSD? 10,000+. Even if you took all 12 hard drives in your scenario and put them in a RAID configuration you'd still not match the performance of a single SSD. Also no one is saying go out and get a 512GB SSD which is on the bleeding edge of consumer SSD. You can easily find a 64GB SSD for around $125. Also no one needs to buy an Apple notebook.
Why should one have to rely on "other sources" to get an accurate death toll? Oh that's right because the US Military has been dodging pinning down a death toll that mirrored reality the entire war. They had the data right in front of them, they just didn't care or didn't want to the public to know of all the civilian deaths caused by this careless, ham-fisted war.
Firstly this can save money. Integrating the GPU into the CPU can create a lower cost part for an OEM than having to using two chips in separate packages. Second this is a fusion between x86 & GPGPU/OpenCL. Once a critical mass of CPUs have an integrated GPU then you will probably see GPGPU tech really start to become integrated into programs that can take advantage of it. Suddenly your low-end budget box CPU can encode & decode multiple HD streams from your camera or apply special effects in realtime. Your games can take advantage of the integrated GPU for physics or possibly the framerate will be playable compared to some of the other crap IGPs. Things like image/video/audio encoding/decoding/editing, gaming, compression & encryption can all benefit from GPGPU. This is basically the start of setting a GPU specifications floor. The question is will Intel/nVidia play along and implement quality OpenCL on their GPUs? I think nVidia will probably have to at some point, but Intel might be a stalwart as OpenCL, anything not x86 that performs general purpose instructions probably looks like a threat to them.
I am wondering how your Geforce4 Ti4600 got outclassed by a Voodoo3. The Voodoo3 was equaled or outclassed by the original GeForce256. Maybe your memory is fuzzy, but there would be some major issues if your Voodoo3 was faster than the Geforce4. Also multi-texturing was a big deal around the TnT2 & Voodoo2/3 days, the Geforce3/4 were way past that stage with the introduction of shaders. By the time the GeForce4 came around 3Dfx was already dead for two years after their Voodoo5 failed miserably against the Geforce2. In some cases, like going from a Geforce2 to a Geforce3/Radeon 8500, there wasn't much performance gain until shaders got involved. But I cannot see how a card that would have been 3 generations behind could have bested the Geforce4 even in lower-end non-shader games. The Voodoo3 would have been outclassed even from a raw clockrate, TMU, ROP standpoint.
I ultimately feel that the reason AMD is changing the model # is to save face so it doesn't look like they're putting out a low-end part first. This looks like they're still pumping out a leading edge flagship product. It's misleading, but that has never stopped the ill informed consumer from continuing to be ill informed.
There are actually quite a few morally questionable actions that are ignored everyday, performed by employees of corporations. Just because you have a legal job, doesn't mean you aren't screwing somebody over directly or part of a corporation that does it.
Not that this excuses thieves in general. Bad behavior is bad behavior.
Here's how it works. Upper level executives and marketing devise all sorts of complicated pricing structures and add-ons. The next step is to trickle this information down to 'code monkeys' to wrangle into their complex and cluttered billing system. Then the final step is to poorly inform and train their call center which usually has poor morale & high turnover so the chance of getting someone who knows what they're doing is slim. Rinse & repeat. Pretty much their mantra is "If there's a problem we'll fix it later. If the customer doesn't notice, all the better."
I would really love to know what state you're in that doesn't require you to purchase any form of auto-insurance. So you don't have to buy PIP or uninsured motorist coverage even? Nor do you have to fork over a fee(e.g. $500/yr fee) because you choose to remain uninsured? Are you also sure your state doesn't come down harder on the uninsured(e.g. suspending your license for an accident over $1,000 in damages then requiring you to obtain insurance for at least the next year)?
Would you revoke fire or police coverage if you had the choice? Does it bother you that you can't?
Slashdot Mirror
Client/Server works this way. Client requests data, server can grant or not grant access to said data. It's like having a stranger coming up to you and nicely asking(without malice or threats) if you'd like to give them something(money, cellphone, newspaper, the time, etc..) you can say "yes or no". The server granted these people access without them breaking the law. The server could almost be viewed as an extension of the company or under license from the company to make these decisions to stream or not stream. If they had a password on the stream and people cracked it or multiple people were sharing an account authorized for only one stream, then yes that would be against the rules. If it is a public stream going out to anyone who asks, it's pretty much fair game.
We're in a situation where 1GB of data costs more than 1kW of energy, in some cases it can be drastically more. If grandma is paying $55/mn for internet and only uses 2GB that's about $22.50 per GB which is pretty outrageous. She uses less, she still pays the same. Even someone maxing out their 250GB Comcast cap is paying $0.22 per GB, which is still quite expensive.
Frankly all utilities need to be monitored and broken up when needed to ensure there are multiple competitors. Also there should be huge efforts towards preventing carriers from also producing content as well. For example Comcast owns Comcast Sports Net and has been trying to extort higher than average fees from other non-comcast providers to carry it. Also Comcast's merger with NBC should raise huge flags. They'll own the line, the TV service & the content on that service... How does that not sound like a monopoly?
...and your hatred of Magento has what to do with NoSQL?
NoSQL is not just key-value lookups. Take a look at Redis or MongoDB, there are novel ideas in both of them & yes they do bring new things to the table. They are NOT memcache. I am also not sure people are "sacrificing a lot of what is present in an RDBMS" by choosing NoSQL over an RDBMS. I think your gripe is with people who don't know what the hell they're doing, but you project that griping on to NoSQL in general. There are some things that RDBMSs are really good at, there are some things RDBMSs aren't so great at. The huge majority of people in the NoSQL communities and the users of these solutions know that loading a million objects client-side and discarding all but 5 is stupid & no one would suggest that is a failing of either RDBMS or NoSQL solutions, but squarely on the user.
I would have to say that NoSQL is more relatable to how people think about objects and their relation to each other. People don't easily boil their objects down into relational tables and how each of those tables should interact with each other. This takes skill & talent & can be a bit of a pain to dive into which is why we have a bunch of ORM solutions which add another layer of cruft on top of RDBMSs. NoSQL is basically getting rid of the ORM & the tables(though some still use table-like structures). For apps that use that would normally use ORM(a lot of web apps) extensively that's great. For newbies who don't have years and years under their belt designing, tweaking normalization, sharding/partitioning it can be easier to pickup. Some of the NoSQL solutions have clustering/horizontal scaling and/or replication built-in, with no or very little schema/query changes required.
So for some NoSQL will be a better solution and for others a RDBMS will be a better solution. I wouldn't knock either. Just because you can do something in an RDBMS doesn't mean it's better than a NoSQL solution & visa versa.
Where do you live where they're improving your available bandwidth every 6 months? Some people are still stuck on 1.5mbit/sec or less crap DSL.
This is session cookie hijacking, it could be used to spoof your Slashdot credentials just as much as someone's Facebook account. Someone just put "Social Network" in the headline to make it seem more hip. Cookie spoofing has been known since the invention of Cookies.
Can you post some data showing the guaranteed worst-case write latency for a top of the line 15k SAS drive or SAS drives in raid configuration? I found some data showing an Intel SSD having around 80ms worst-case write latency.
Even for many enterprise scenarios, this is not a huge deal & for consumers this is completely moot.
OK, first this isn't a "file level" issue, this is a physical disk/sector level issue. When you have all your data in 512-byte sectors and the majority of the time in normal usage your data is NOT sequentially laid out, you will incur random writes & reads all over the platters of the disk. It doesn't matter if all the files you use are over 1MB or over 512KB, what matters is how those files are physically laid out. If you have a 1MB file but it's spread out with 50% mostly sequential sectors and 50% random sectors, you will get a hit on load times with a conventional disk. Due to usage patterns and at least how NTFS writes data, fragmentation will cause a large majority of read & write operations to be random, regardless of the size of the file you're reading. Even if you're using a file system that suffers less from fragmentation, that doesn't mean you won't need to read randomly from large files such as from a database.
You can stick with your HDDs and in many usage scenarios that may make the most sense, but from a performance perspective, SSDs will usually outclass HDDs in random/write unless you're going to crazy scenarios that cost similar or more than the SSD.
Plus this article was about SSD for consumer users and they usually do not have 1TB of data or if they do need it they can get both a smaller SSD and larger hard drive. The SSD will provide better response and faster boot than their single consumer level hard disk.
Have you personally dealt with a scenario where an SSD did not write data within the time frame you needed it to? How badly did the SSD do? What brand of SSD was it and what application were you using?
Well that does lead one to ask a lot of questions, such as "how many HDD are in a rack", if a rack means 5 or 6 15k rpm HDD that were short-stroked(at this point whining about the cost of SSDs goes out the window). Maybe that can beat the SSD. But 6 or even 12 consumer level HDD used to their full capacity, cannot defeat a single SSD at random reads/writes:
http://www.anandtech.com/show/2968/intel-s-x25-v-kingston-s-30gb-ssdnow-v-series-battle-of-the-125-ssds/6
Not quite:
http://www.tomshardware.com/reviews/short-stroking-hdd,2157-9.html
Yes it does show that 4 short stroked SAS 15K RPM drives are beating out a single SSD by getting 2500 IOPs in the DB test. But those are older SSDs. Compare to newer SSDs in similar/same benchmark:
http://www.xbitlabs.com/articles/storage/display/corsair-ssd-roundup_6.html#sect0
You will see they are bottoming out @ 4K IOPS in worst case scenarios.
Also contrary to what was suggested earlier, short stroking does not make HDD seek time negligible:
http://www.tomshardware.com/reviews/short-stroking-hdd,2157-5.html
3.8ms for HDD vs 0.1ms for SSD. That's still a big difference.
You are correct about the low level nature of flash memory, but there are many ways to mitigate this. SSD controllers use multiple channels to read/write banks of flash memory. They have large internal buffers & they also have "waste space", meaning they have extra flash memory that the device can read/write to so as to not hold up the drive. SandForce controllers also use compression and other methods to enhance performance. With proper trim support you do not need to run garbage collection, thus it's really not much different in operation than how a HDD will act, just faster. No modern SSD will make you wait one second while it erases a block of data, they just don't work that way now.
Even worst case random write speeds outpace standard HDDs. Yes if you short stroke your 15k RPM and buy 6 of them then yipee maybe you can exceed the performance of a single SSD, but then any arguments about saving money buying 6x $400 drives + controller so you can use 5% of their capacity go out the window.
Why do you lie about HDD vs SSD random read/writes:
http://www.anandtech.com/show/2968/intel-s-x25-v-kingston-s-30gb-ssdnow-v-series-battle-of-the-125-ssds/6
Even the 10K high-end pricey WD Velociraptor has a pathetic 0.8MB/throughput for random writes. Meanwhile a value level Intel SSD gets 35MB/sec throughput for random writes? How do you think putting 4 of those high end HDD together will make then faster than an SSD? The fact is they won't.
Plus if you think an SSD is a waste, why would you even entertain the suggestion of capping a 1TB HDD to 32GB? Do you have an benchmarks proving this is faster and really drops seeks to near zero?
Random write speed on an SSD is still faster than a hard drive. Even value level SSDs now have comparable sequential write speeds to standard HDD. I am not sure where you're getting your "a lot slower" data from.
You don't have to have a disk full of files 4K to benefit from an SSD. 4K random read/writes give a good indication for drive responsiveness and it's ability to handle multiple IO tasks at the same time. For example reading & displaying header info from 1,000 image files then producing 1,000 tiny thumbnails(something Windows 7 likes to do) is much quicker on a SSD than on conventional HDD. Also a HDD will struggle if you're trying to do more than one major task at a time. Things like writing cache files, updating Firefox SQLite database, file searches, are quicker. Additional boot times are reduced greatly thanks to SSD many people see 25% to 50% improvement.
As for the 12 disk raid vs the SSD. The SSD is still going to kick the RAIDs ass at random read/writes, sorry but it's true. Maybe you see no point to it and want to hold on to your precious HDD. That doesn't mean there isn't a use for SSDs right now as a worthy upgrade for around $125. It will do more for your boot time and general computer usage than buying 12 HDDs.
HDDs are still useful for story large amounts of data cheaply, but they are by no means fast & are not getting any faster.
Let's invent a buzzword for SSDs like "PowerStream Boost w/ Turbo AI", makes no fucking sense but people will gobble it up even if they have no clue what it really means. Ultimately SSDs just need to be marketed correctly to educate customers that there is a performance improvement and that you do not need the larger hard drive. A lot of consumers could probably even get by with a 64 or 128GB SSD. So just market it as "20,000 Operations Per Second!!!! Thanks to PowerStream Turbo. Stores up to 20,000+ music files." People might ask "hmm how many Operations can that hard disk do" and if they find out it's only a few hundred, that might swing their purchasing decision.
Modern SSD tech has enough cycles to let you use it for many many years. They are most useful right now as an OS drive to speed up your general computing. They are not suited for file server/backup duties. They work via regular SATA connectors, so I am not sure where you're getting the idea you have to buy a special memory module? Why would you have to replace the whole laptop if only your drive went bad? Why do you think that writing all zeros is a secure erase, newsflash, it's NOT. Also that is a very a-typical situation most consumers will never find themselves in.
Your rack of 15k rpm drives probably still can't match the random IO performance of a single quality SSD.
Ok and how many random 4K write/read IOPs could you get even if you did RAID-1? A few hundred. How many can you get with an SSD? 10,000+. Even if you took all 12 hard drives in your scenario and put them in a RAID configuration you'd still not match the performance of a single SSD. Also no one is saying go out and get a 512GB SSD which is on the bleeding edge of consumer SSD. You can easily find a 64GB SSD for around $125. Also no one needs to buy an Apple notebook.
Why should one have to rely on "other sources" to get an accurate death toll? Oh that's right because the US Military has been dodging pinning down a death toll that mirrored reality the entire war. They had the data right in front of them, they just didn't care or didn't want to the public to know of all the civilian deaths caused by this careless, ham-fisted war.
The proper measurement is actually metric asstons...
Firstly this can save money. Integrating the GPU into the CPU can create a lower cost part for an OEM than having to using two chips in separate packages. Second this is a fusion between x86 & GPGPU/OpenCL. Once a critical mass of CPUs have an integrated GPU then you will probably see GPGPU tech really start to become integrated into programs that can take advantage of it. Suddenly your low-end budget box CPU can encode & decode multiple HD streams from your camera or apply special effects in realtime. Your games can take advantage of the integrated GPU for physics or possibly the framerate will be playable compared to some of the other crap IGPs. Things like image/video/audio encoding/decoding/editing, gaming, compression & encryption can all benefit from GPGPU. This is basically the start of setting a GPU specifications floor. The question is will Intel/nVidia play along and implement quality OpenCL on their GPUs? I think nVidia will probably have to at some point, but Intel might be a stalwart as OpenCL, anything not x86 that performs general purpose instructions probably looks like a threat to them.
I am wondering how your Geforce4 Ti4600 got outclassed by a Voodoo3. The Voodoo3 was equaled or outclassed by the original GeForce256. Maybe your memory is fuzzy, but there would be some major issues if your Voodoo3 was faster than the Geforce4. Also multi-texturing was a big deal around the TnT2 & Voodoo2/3 days, the Geforce3/4 were way past that stage with the introduction of shaders. By the time the GeForce4 came around 3Dfx was already dead for two years after their Voodoo5 failed miserably against the Geforce2. In some cases, like going from a Geforce2 to a Geforce3/Radeon 8500, there wasn't much performance gain until shaders got involved. But I cannot see how a card that would have been 3 generations behind could have bested the Geforce4 even in lower-end non-shader games. The Voodoo3 would have been outclassed even from a raw clockrate, TMU, ROP standpoint.
I ultimately feel that the reason AMD is changing the model # is to save face so it doesn't look like they're putting out a low-end part first. This looks like they're still pumping out a leading edge flagship product. It's misleading, but that has never stopped the ill informed consumer from continuing to be ill informed.
3.999~G = 4G :)
There are actually quite a few morally questionable actions that are ignored everyday, performed by employees of corporations. Just because you have a legal job, doesn't mean you aren't screwing somebody over directly or part of a corporation that does it.
Not that this excuses thieves in general. Bad behavior is bad behavior.
Here's how it works. Upper level executives and marketing devise all sorts of complicated pricing structures and add-ons. The next step is to trickle this information down to 'code monkeys' to wrangle into their complex and cluttered billing system. Then the final step is to poorly inform and train their call center which usually has poor morale & high turnover so the chance of getting someone who knows what they're doing is slim. Rinse & repeat. Pretty much their mantra is "If there's a problem we'll fix it later. If the customer doesn't notice, all the better."
I would really love to know what state you're in that doesn't require you to purchase any form of auto-insurance. So you don't have to buy PIP or uninsured motorist coverage even? Nor do you have to fork over a fee(e.g. $500/yr fee) because you choose to remain uninsured? Are you also sure your state doesn't come down harder on the uninsured(e.g. suspending your license for an accident over $1,000 in damages then requiring you to obtain insurance for at least the next year)?
Would you revoke fire or police coverage if you had the choice? Does it bother you that you can't?