Wrong comparison. Instead hark back to 1950s era telephone companies where they required that you had to rent/buy their telephone, and you got charged for the number of telephones you had, even if you had an illicit phone that the lineman didn't get to see (voltage checks, Ringer Equivalencies, etc).
This isn't about what you can say/do, but what you can use. All in the name of safety of course. An illegal/badly made phone could start a fire! It could damage other properly working telephone equipment!
No, you cannot do this. Otherwise you would have a kernel->userspace information leak. Any/all registers are accessible to an application when it is running. Additionally, it has the potential to cause data corruption. Thus it is that registers are saved and restored on context-switches, plus-or-minus the returns from syscalls of course.
Worse than that, most datacentres, if they provide AMD boxes for lease, use the cheapest POS motherboards. SiS chipsets, usually. Then when the ethernet falls over b/c of a DoS, they blame it on AMD. No, it's the cheap motherboards they pair with the processors. Stick a real eepro NIC in an SiS motherboard, and it will sit pretty and take all the abuse. Then again, the SiS sucks anyway, and I recommend VIA chips to be paired with AMD processors.
And yes, this is personal experience with multiple datacentres. One was kind enough to replace the NIC with an EEPRO for free, and nary a problem since. Others would just blame it on AMD when we said our box was falling over for no good reason.
So, these Intel fanatics see AMD, see it fail, and spread their (technically truthful, given their experiences) FUD. Say it's cheap, you get what you pay for. This is true... but it's not about AMD.
Wrong. the drive has a very good reason to cache data. Only the drive knows which sectors are nearest to where [why? a) hotfix sectors b) all drives now use LBA. CHS has been a hack since 500MB hard-drives. the actual geometry is hidden from the OS, thus the OS can only do very limited re-ordering of reads/writes.].
That is the point of TCQ, and why drives should buffer reads and writes and execute out of order.
I think you have successfully missed the point. He means 8 bits per channel. 10 to 12 bits per CHANNEL. so it's not about 24 bit colour... it would be 30 to 36 bits. for example, the PNG file format can do 16 bits per channel, or 48bit colour. Don't confuse the people, and don't pollute the waters by spewing what you don't understand.
funny, i'm in Michigan and in SBC country, and they don't allow (last i checked) us to get DSL service w/o SBC Phone service... you have to have an SBC phone number. same requirement for the local ISP that runs on SBC's lines.
are you sure? -O3 can enable unsafe optimizations... plus march=i686 and mcpu=i586 would produce programs with instructions for 686 but instruction scheduling for 586. -march should promote -mcpu (tho it might not... UNIX does give you enough rope to shoot yourself in the foot. However it would be a STUPID thing to do).
If I were to give you the benefit of a doubt, I'd assume you mean -march=i586 and -mcpu=i686 which will limit it to instructions that exist on the P5, but will schedule it for the PPro and PII series processors.
at least in Michigan, food and prescription drugs are not taxed, as they are classed as "necessary for living". and I know that Michigan is not alone, tho it may not be the same in all states.
However, it is not. a) there will be a separate 2.6.12-pre1. b) a -pre release is entirely a 'testing' release for adding patches, but not considered 'stable'.
So a m.n.x-pre is an unstable release, while a m.n.x.y will be considered a patchfix/more-stable release.
Possibly b/c I have many "role" accounts? I have half a dozen different domains I have accounts under. Mind you, they all forward to the same mailbox, but I have to be able to send back email using the same From that it was sent To.
And fwiw, the problem you speak of is handled quite nicely with SPF and/or SPF/PRA. And I can even deal with it by setting my MX as a valid relay for all of those "role" accounts.
Re:Crippling DNS? How much does DNS suck?
on
Spammers' Upend DNS
·
· Score: 1
And you have missed the point as well. Even assuming that the negative result is cached, that is still per cache. So although you DO reduce the load somewhat (tho chances are, the negative is being cached already) you are forgetting breadth. You still have this spam sent to 10million email addrs, and to say 100,000 domains.
Say there is an average of 2 domains per MX server (number pulled out of ass. however it is probably reasonable), you now have 50,000 requests. per mailing.
And these requests will likely all be made over a one hour, mebbe two hour period. Which doesn't sound so bad. But now consider the following. Now add in the fact that there are, say, 1000 of these emails sent out with different domains. You now have 50million requests.
Which probably will have a spread of (for example, this is another number pulled out of ass) maybe 6 hours.
True, this might not be so bad, but it certainly is not good. and you certainly didn't remember this problem. Depth isn't our only concern, but also breadth.
Actually, pretty much all do, but it's really just for accounting purposes. Tho it should be noted, that at least in linux, the 'idle process' can actually be doing work. Servicing hardware interrupts. However it should also be noted that _any_ process can get charged for hard-irq servicing. Of course, the cost of the hard-irq service is kept to a minimum, and as much as possible is pushed into a deffered 'soft-irq'.
This is done mostly b/c at hard-irq time, it is largely impossible to know beforehand which process the work really belongs to, but in theory a CPU hog should be causing most of the IRQ work as well. Obviously this is not always the case, but it is the fairest that is possible given the architecture, as well as getting hardirqs serviced as quickly as possible.
NAT rewrites outgoing packets and maintains a connection tracking table for outgoing packets so that incoming packets may be routed to the correct internal host.
However, this also means that packets coming in that do not match an entry in the conntrack table will simply be... ignored. aka dropped. After all, short of you setting up a default host for random packets to go to, a NAT device does not know what to do with random packets.
Eastern Standard Tribe by Cory Doctorow has this concept.
Eastern Standard Tribe
Aparrently instead of paying a toll for the New Jersey Turnpike, one downloads music, then pays for it at the gate. Don't ask me tho...
That method may get you marked as a spammer yourself (think high quantities of unwanted email == you getting reported).
Plus the fact that you may end up holding bounced-spams in your queue for 3-4 days until it expires, and greatly increasing the load on your mail-server. The trick is to bounce before the SMTP transaction is over. If you determine it's spam/bad after the SMTP transaction is over, just drop the email. File to/dev/null.
And as said before, this also just ends up doing what is basically a reverse joe-job. If the sender address is a real email address, you end up filling up their box with all the bounces. This is not kind either.
Well, dang. I had a parenthetical in there "(I always called it W-2 employment)", but removed it as I had thought it irrelevant and maybe I was wrong. so much for that.
1099 is (not from the form name, but what it is used for) "Temporary/Contract worker", and the employee is responsible for all taxes, benefits, etc. However, a 1099 employer rarely pays enough to pay for benefits, and it's legally NOT THEIR PROBLEM. As a 1099 employee, you are legally self-employed as a contractor/subcontractor.
W-4 employment is where you are legally on their payroll, they handle all the payroll taxes, benefits, etc, and you are a permanent worker (well as permanent as you can get).
Hopefully this will help with any confusion. And I can be assured to not get a karma boost, as the moderators are much furthe down the page by now;)
Given the lossiness of JPEG compression, and the way they artifact, I doubt you could use JPEG either. I don't think even B&W JPEG. This is yet another case for PNG, probably. Possibly reducible to 1 bit, with careful filtering.
I have no clue where you got the idea that ext3 is based off XFS. ext3 doesn't even use extents. Tho there have been discussions of an extension/adaptation of it using extents. there is also the discussion of htree support. personally, I think it should be named ext4 then...
XFS doesn't [have to] blockalign files either. ext2/3 does. ext3 uses full blocks for directories (to store dentries). XFS doesn't (though it will, after a couple dozen dentries). Reiser allocates incrementally, never going for a full block for directories.
Anyway, I don't see how ext3 relates to XFS except both are journaled FSs and may use similar theories. At which point one may as well claim its related to NTFS or JFS or even JFFS
Slashdot Mirror
KE = (m/2)*v^2
or the more common form: KE = 1/2mv^2
This incidentally does make it in the same class as E=m*c^2 (due to magnitude/exponentiation) but that's more or less irrelevant.
Wrong comparison. Instead hark back to 1950s era telephone companies where they required that you had to rent/buy their telephone, and you got charged for the number of telephones you had, even if you had an illicit phone that the lineman didn't get to see (voltage checks, Ringer Equivalencies, etc).
This isn't about what you can say/do, but what you can use. All in the name of safety of course. An illegal/badly made phone could start a fire! It could damage other properly working telephone equipment!
No, you cannot do this. Otherwise you would have a kernel->userspace information leak. Any/all registers are accessible to an application when it is running. Additionally, it has the potential to cause data corruption. Thus it is that registers are saved and restored on context-switches, plus-or-minus the returns from syscalls of course.
Worse than that, most datacentres, if they provide AMD boxes for lease, use the cheapest POS motherboards. SiS chipsets, usually. Then when the ethernet falls over b/c of a DoS, they blame it on AMD. No, it's the cheap motherboards they pair with the processors. Stick a real eepro NIC in an SiS motherboard, and it will sit pretty and take all the abuse. Then again, the SiS sucks anyway, and I recommend VIA chips to be paired with AMD processors.
And yes, this is personal experience with multiple datacentres. One was kind enough to replace the NIC with an EEPRO for free, and nary a problem since. Others would just blame it on AMD when we said our box was falling over for no good reason.
So, these Intel fanatics see AMD, see it fail, and spread their (technically truthful, given their experiences) FUD. Say it's cheap, you get what you pay for. This is true... but it's not about AMD.
Wrong. the drive has a very good reason to cache data. Only the drive knows which sectors are nearest to where [why? a) hotfix sectors b) all drives now use LBA. CHS has been a hack since 500MB hard-drives. the actual geometry is hidden from the OS, thus the OS can only do very limited re-ordering of reads/writes.].
That is the point of TCQ, and why drives should buffer reads and writes and execute out of order.
Wal*Mart doesn't have 99c price-points. It's a policy even.
.x5 and no .x9 price-points. I'd try to give more info but I haven't had access to Pipeline since '01.
Policy is no
I think you have successfully missed the point. He means 8 bits per channel. 10 to 12 bits per CHANNEL. so it's not about 24 bit colour... it would be 30 to 36 bits. for example, the PNG file format can do 16 bits per channel, or 48bit colour. Don't confuse the people, and don't pollute the waters by spewing what you don't understand.
funny, i'm in Michigan and in SBC country, and they don't allow (last i checked) us to get DSL service w/o SBC Phone service... you have to have an SBC phone number. same requirement for the local ISP that runs on SBC's lines.
are you sure? -O3 can enable unsafe optimizations... plus march=i686 and mcpu=i586 would produce programs with instructions for 686 but instruction scheduling for 586. -march should promote -mcpu (tho it might not... UNIX does give you enough rope to shoot yourself in the foot. However it would be a STUPID thing to do).
If I were to give you the benefit of a doubt, I'd assume you mean -march=i586 and -mcpu=i686 which will limit it to instructions that exist on the P5, but will schedule it for the PPro and PII series processors.
at least in Michigan, food and prescription drugs are not taxed, as they are classed as "necessary for living". and I know that Michigan is not alone, tho it may not be the same in all states.
the new ATi CATALYST driver installer claims to need the .NET Framework. probably not for the driver but for the control panel.
However, it is not. a) there will be a separate 2.6.12-pre1. b) a -pre release is entirely a 'testing' release for adding patches, but not considered 'stable'.
So a m.n.x-pre is an unstable release, while a m.n.x.y will be considered a patchfix/more-stable release.
"Dick van Dyke" show anyone?
Possibly b/c I have many "role" accounts? I have half a dozen different domains I have accounts under. Mind you, they all forward to the same mailbox, but I have to be able to send back email using the same From that it was sent To.
And fwiw, the problem you speak of is handled quite nicely with SPF and/or SPF/PRA. And I can even deal with it by setting my MX as a valid relay for all of those "role" accounts.
And you have missed the point as well. Even assuming that the negative result is cached, that is still per cache. So although you DO reduce the load somewhat (tho chances are, the negative is being cached already) you are forgetting breadth. You still have this spam sent to 10million email addrs, and to say 100,000 domains.
Say there is an average of 2 domains per MX server (number pulled out of ass. however it is probably reasonable), you now have 50,000 requests. per mailing.
And these requests will likely all be made over a one hour, mebbe two hour period. Which doesn't sound so bad. But now consider the following.
Now add in the fact that there are, say, 1000 of these emails sent out with different domains. You now have 50million requests.
Which probably will have a spread of (for example, this is another number pulled out of ass) maybe 6 hours.
True, this might not be so bad, but it certainly is not good. and you certainly didn't remember this problem. Depth isn't our only concern, but also breadth.
Actually, pretty much all do, but it's really just for accounting purposes.
Tho it should be noted, that at least in linux, the 'idle process' can actually be doing work. Servicing hardware interrupts. However it should also be noted that _any_ process can get charged for hard-irq servicing. Of course, the cost of the hard-irq service is kept to a minimum, and as much as possible is pushed into a deffered 'soft-irq'.
This is done mostly b/c at hard-irq time, it is largely impossible to know beforehand which process the work really belongs to, but in theory a CPU hog should be causing most of the IRQ work as well. Obviously this is not always the case, but it is the fairest that is possible given the architecture, as well as getting hardirqs serviced as quickly as possible.
Both true and false.
NAT rewrites outgoing packets and maintains a connection tracking table for outgoing packets so that incoming packets may be routed to the correct internal host.
However, this also means that packets coming in that do not match an entry in the conntrack table will simply be... ignored. aka dropped. After all, short of you setting up a default host for random packets to go to, a NAT device does not know what to do with random packets.
Eastern Standard Tribe Aparrently instead of paying a toll for the New Jersey Turnpike, one downloads music, then pays for it at the gate. Don't ask me tho...
That method may get you marked as a spammer yourself (think high quantities of unwanted email == you getting reported).
/dev/null.
Plus the fact that you may end up holding bounced-spams in your queue for 3-4 days until it expires, and greatly increasing the load on your mail-server. The trick is to bounce before the SMTP transaction is over. If you determine it's spam/bad after the SMTP transaction is over, just drop the email. File to
And as said before, this also just ends up doing what is basically a reverse joe-job. If the sender address is a real email address, you end up filling up their box with all the bounces. This is not kind either.
Well, dang. I had a parenthetical in there "(I always called it W-2 employment)", but removed it as I had thought it irrelevant and maybe I was wrong. so much for that.
He means W-4 vs 1099 employment.
;)
1099 is (not from the form name, but what it is used for) "Temporary/Contract worker", and the employee is responsible for all taxes, benefits, etc. However, a 1099 employer rarely pays enough to pay for benefits, and it's legally NOT THEIR PROBLEM. As a 1099 employee, you are legally self-employed as a contractor/subcontractor.
W-4 employment is where you are legally on their payroll, they handle all the payroll taxes, benefits, etc, and you are a permanent worker (well as permanent as you can get).
Hopefully this will help with any confusion. And I can be assured to not get a karma boost, as the moderators are much furthe down the page by now
that are all behind the same DSL^Wdialup modem, sitting in your mother's basement.
Given the lossiness of JPEG compression, and the way they artifact, I doubt you could use JPEG either. I don't think even B&W JPEG. This is yet another case for PNG, probably. Possibly reducible to 1 bit, with careful filtering.
Can't be. there was a WinAMP 3. It just sucked. So Justin (and crew) worked on WinAMP 5. which was so named b/c it was a combination of 2 + 3.
Then again, most ppl I know still prefer 2.9x
I have no clue where you got the idea that ext3 is based off XFS. ext3 doesn't even use extents. Tho there have been discussions of an extension/adaptation of it using extents. there is also the discussion of htree support. personally, I think it should be named ext4 then...
XFS doesn't [have to] blockalign files either. ext2/3 does. ext3 uses full blocks for directories (to store dentries). XFS doesn't (though it will, after a couple dozen dentries). Reiser allocates incrementally, never going for a full block for directories.
Anyway, I don't see how ext3 relates to XFS except both are journaled FSs and may use similar theories. At which point one may as well claim its related to NTFS or JFS or even JFFS