> to trace and identify the source of a communication - Source IP address. 4 byte (for IPv4 - more for v6) > to trace and identify the destination of a communication - Destination IP address Another 4 byte > to identify the date, time and duration of a communication - Time until TCP connection dies or UDP session ends A date/time (4 bytes) plus another time - assume another 4 byte, to take it wide > to identify the type of communication - being able to determine if it was TCP, UDP or some other packet type was considered sufficient. A few bits, say one byte. I daresay the protocol will also be required, though, another 2 byte for the port number. > to identify the communication device - In the case of mobile phones and mobile phone networks Mmm, harder. Assume storing the Mac - that should give a reasonable idea. 6 bytes. > to identify the location of mobile communication equipment - Obvious. Hmmm... Let's assume GPS coordinates, here, stored conveniently in 2 4-byte floats.
That makes 33 byte per connection. Peanuts, I agree. Now for the tough part of the calculation: how many connections does an ISP handle per day. Remember, we're talking connections, not customers, not even web pages. I'm going to be extrapolating heavily here, bear with me.
Way back when I was part of Skynet (no, not that one) we had about 3 million pageviews per day on our portal. At a rate of about 30 connections per pageview - not unreasonable, take into account CSS, JS, split images et cetera - that makes 90 million connections per day, just for our own portal. Our main competitor in the portal area was MSN, say the same amount, plus maybe twice that for all the other portals combined; makes 260 million connections just for web portals. That makes 8 gigabytes of logging - overhead for record headers and indices not included.
Now, I think it's a pretty reasonable assumptions that people don't just surf portals. Totally out of the blue, let's assume non-portal web traffic is twenty times that - I honestly don't think that's overrated, far from it. Another 160 gigabytes.
Let's give FTP, NNTP and the rest of the lesser-known protocols about 100 million connections - totally out of the blue, admittedly, but that's the same as our own portal, probably not exaggerated. 3 gigabytes.
That makes for 171 gigabytes of logging per day. Assume you're not going to do a lot of querying there, so ignore indices. Compress the data, so say 50 gig per day. Times 365 makes 18 tera, times 7 years makes roughly 130 terabytes. Not exceedingly unlikely, I agree, but still costly. You're not allowed to keep the data for longer (privacy law), so you need to clean it up on a regular basis, so you will actually need some indices on the datefield. Make it 150 terabyte. Encryption will also be needed, probably, although that mostly impacts the processing power needed, not the compression or data size.
Now for the heavy bastard, though: peer-to-peer. Given the nature of the beast, several hundred to thousand connections per download per user. Very very hard to estimate, but I'm sure you can see where this is going. Given the popularity of peer-to-peer, I honestly wouldn't be surprised if storage requirements for the simple, easy logging you requested exceed a petabyte. I'm not renting that from Amazon, dude.
And then you're going by the European directive, which is very likely to be "improved" by local authorities and MAFIAA-lobbyists who feel they need to actually prove you were transferring actual illegal materials, too.
Of course, searching through that amount of compressed, encrypted data will take absolutely ages, so after the first few requests, law enforcement will get deadlines for information retrieval that pretty much force you to index everything, so double your storage again.
Oh, Skynet is one of the two major ISPs in Belgium, a 10-million people country. Just how big do ISPs get in America ?
One of the things I remember seeing this being demoed with quite some time ago, was using the rear camera to do augmented reality stuff - your fingers on the touchpad basically appear on top of what the camera sees, creating the illusion of a see-through device.
As one possible example, multitouch while holding the device like a console, with both hands. Might be interesting, will probably be both wrongly and over-used by shitty developers.
I'm not saying it's a *better* way, but the traditional procedure is waiting until you're so sick of the regime that you just can't stand it anymore, and then run out on the street attacking random avatars of said regime. If the time is right for it, major parts of the populace will then also realise they've had enough and join you.
If the time isn't right, however, well, let's just say your revolution will go out in a very brief bang.
This has worked for centuries past, be it slowly and painfully.
It's not nust about splitting log events. It's also about adding a lot more logging (you don't routinely log every connection from every customer, let alone the contents of them); but also the keeping of records for seven years. Given the amount of traffic ISPs push, that requires ungodly amounts of storage, which in turn requires ungodly amounts of power.
It is ? Don't get me wrong, sounds like a good strategy, from their point of view, but I've never heard of it. How do they keep the second bomb/guy from being blown up by the first one ? If he's off-scene, how do they get him on-scene through the security the response team brings in ? When/where has this happened before ?
Cast metal ? There's several brands of 3D printers out there that simply PRINT metal (SLS, or Selective Laser Sintering). Get with the program, community hobbyist.
It's been a while, but I seem to recall you can run Bugzilla under ModPerl, which keeps the compiled versions in memory and simply re-executes them - yes, kinda like it should be - and thus gives you rather impressive speedups.
ModPerl has some gotchas re. variable scope and cleanup, though, so I may be wrong about it Just Working.
You're touching on why this may well have a lot less effect than desired: I have found, over the years, that a lot of people fail to make the connection between "the government steals my monies" and "my, how convenient that there's a hardened road between my house and the shops".
I have no idea why this is, though - perhaps the fact that it's always been there, like the sun and the trees ? Maybe the perception that politicians fill their pockets with that money plays, too, true or false as it may be.
Oh, right, the main reason I need to switch between X and console, is that XBMC regularly hangs on that box. I have no proof that that is related to GL or the Ati drivers in general; but when I pkill xbmc on the console, that console immediately and consistently goes corrupt, and no command I type fixes that. The corruption only clears after I've switched back to X.
I'll be switching out that card for an NVidia one soon. The VDPAU will be a nice bonus for the XBMC, too.
Oh, I wasn't targeting NVidia - they're pretty much the only big name who does have mostly decent drivers, as also pointed out somewhere else; apparently they're the only ones whitelisted iirc.
The Intel chipsets are also pretty decently supported, but they're not exactly the high end of the graphics market, and I can't much speak for their GL implementation. I do know that my current HP laptop with Intel graphics tends to lock up when I use GL screensavers.
Ati's drivers were more than horrible last time I touched them, to the point where I would reboot into windows, come back to Linux and find my resolution quasi-permanently stuck to 800x600; shit like that. Rebooting into Windows would of course magically work fine. The Linux resolution would somehow unstick when I reinstalled the system (yes, with the same distro, so no updated drivers), but that's not really a useful strategy.
I currently have a single Ati box left, attached to my HDMI TV; and for that I had to write a small script that calls the Ati tools to reset the origin and size of the picture so it actually uses the full screen size. Have to run that script every boot, but also every time I switch back to X from a console.
Slashdot Mirror
Now you're assuming burgers can exist within Stallman's gravitational field.
Popular is not the same as important.
Woosh.
I'm not the guy the AC knows, but...
> to trace and identify the source of a communication - Source IP address.
4 byte (for IPv4 - more for v6)
> to trace and identify the destination of a communication - Destination IP address
Another 4 byte
> to identify the date, time and duration of a communication - Time until TCP connection dies or UDP session ends
A date/time (4 bytes) plus another time - assume another 4 byte, to take it wide
> to identify the type of communication - being able to determine if it was TCP, UDP or some other packet type was considered sufficient.
A few bits, say one byte. I daresay the protocol will also be required, though, another 2 byte for the port number.
> to identify the communication device - In the case of mobile phones and mobile phone networks
Mmm, harder. Assume storing the Mac - that should give a reasonable idea. 6 bytes.
> to identify the location of mobile communication equipment - Obvious.
Hmmm... Let's assume GPS coordinates, here, stored conveniently in 2 4-byte floats.
That makes 33 byte per connection. Peanuts, I agree. Now for the tough part of the calculation: how many connections does an ISP handle per day. Remember, we're talking connections, not customers, not even web pages. I'm going to be extrapolating heavily here, bear with me.
Way back when I was part of Skynet (no, not that one) we had about 3 million pageviews per day on our portal. At a rate of about 30 connections per pageview - not unreasonable, take into account CSS, JS, split images et cetera - that makes 90 million connections per day, just for our own portal. Our main competitor in the portal area was MSN, say the same amount, plus maybe twice that for all the other portals combined; makes 260 million connections just for web portals. That makes 8 gigabytes of logging - overhead for record headers and indices not included.
Now, I think it's a pretty reasonable assumptions that people don't just surf portals. Totally out of the blue, let's assume non-portal web traffic is twenty times that - I honestly don't think that's overrated, far from it. Another 160 gigabytes.
Let's give FTP, NNTP and the rest of the lesser-known protocols about 100 million connections - totally out of the blue, admittedly, but that's the same as our own portal, probably not exaggerated. 3 gigabytes.
That makes for 171 gigabytes of logging per day. Assume you're not going to do a lot of querying there, so ignore indices. Compress the data, so say 50 gig per day. Times 365 makes 18 tera, times 7 years makes roughly 130 terabytes. Not exceedingly unlikely, I agree, but still costly. You're not allowed to keep the data for longer (privacy law), so you need to clean it up on a regular basis, so you will actually need some indices on the datefield. Make it 150 terabyte. Encryption will also be needed, probably, although that mostly impacts the processing power needed, not the compression or data size.
Now for the heavy bastard, though: peer-to-peer. Given the nature of the beast, several hundred to thousand connections per download per user. Very very hard to estimate, but I'm sure you can see where this is going. Given the popularity of peer-to-peer, I honestly wouldn't be surprised if storage requirements for the simple, easy logging you requested exceed a petabyte. I'm not renting that from Amazon, dude.
And then you're going by the European directive, which is very likely to be "improved" by local authorities and MAFIAA-lobbyists who feel they need to actually prove you were transferring actual illegal materials, too.
Of course, searching through that amount of compressed, encrypted data will take absolutely ages, so after the first few requests, law enforcement will get deadlines for information retrieval that pretty much force you to index everything, so double your storage again.
Oh, Skynet is one of the two major ISPs in Belgium, a 10-million people country. Just how big do ISPs get in America ?
One of the things I remember seeing this being demoed with quite some time ago, was using the rear camera to do augmented reality stuff - your fingers on the touchpad basically appear on top of what the camera sees, creating the illusion of a see-through device.
As one possible example, multitouch while holding the device like a console, with both hands. Might be interesting, will probably be both wrongly and over-used by shitty developers.
Sooo... port forwarding over SSH ?
His point exactly, one might suspect.
I'm not saying it's a *better* way, but the traditional procedure is waiting until you're so sick of the regime that you just can't stand it anymore, and then run out on the street attacking random avatars of said regime. If the time is right for it, major parts of the populace will then also realise they've had enough and join you.
If the time isn't right, however, well, let's just say your revolution will go out in a very brief bang.
This has worked for centuries past, be it slowly and painfully.
It's not nust about splitting log events. It's also about adding a lot more logging (you don't routinely log every connection from every customer, let alone the contents of them); but also the keeping of records for seven years. Given the amount of traffic ISPs push, that requires ungodly amounts of storage, which in turn requires ungodly amounts of power.
> The only way not to have a facebook account "at all" is not to have had a facebook account in the first place.
And check.
It is ? Don't get me wrong, sounds like a good strategy, from their point of view, but I've never heard of it. How do they keep the second bomb/guy from being blown up by the first one ? If he's off-scene, how do they get him on-scene through the security the response team brings in ? When/where has this happened before ?
Darling, I'm gay. I don't need an excuse to behave like an asshat.
(and, yes, I really am, so I'm allowed to make fun. nuh. *sticks out tongue*)
> IRS Nails CPA For Copying Steve Jobs, Google Execs
So where can I get my chinese Sergey clone ?
* [...] gods [...]
* [...] conservatives [...]
See the error in your reasoning, there ?
Maybe it has pictures of Mussol- err, Berlusconi on a bad hair day or something similarly incriminating.
Very likely true. However, it only has to be strong enough to remove the screws once, so you can replace them with sensible ones, no ? :-p
Besides, when it breaks you print a new one.
Cast metal ? There's several brands of 3D printers out there that simply PRINT metal (SLS, or Selective Laser Sintering). Get with the program, community hobbyist.
*snore*
Years of experience.
That being said, though, Apple is on a different path with a similar destination, though.
It's been a while, but I seem to recall you can run Bugzilla under ModPerl, which keeps the compiled versions in memory and simply re-executes them - yes, kinda like it should be - and thus gives you rather impressive speedups.
ModPerl has some gotchas re. variable scope and cleanup, though, so I may be wrong about it Just Working.
You're touching on why this may well have a lot less effect than desired: I have found, over the years, that a lot of people fail to make the connection between "the government steals my monies" and "my, how convenient that there's a hardened road between my house and the shops".
I have no idea why this is, though - perhaps the fact that it's always been there, like the sun and the trees ? Maybe the perception that politicians fill their pockets with that money plays, too, true or false as it may be.
Oh, right, the main reason I need to switch between X and console, is that XBMC regularly hangs on that box. I have no proof that that is related to GL or the Ati drivers in general; but when I pkill xbmc on the console, that console immediately and consistently goes corrupt, and no command I type fixes that. The corruption only clears after I've switched back to X.
I'll be switching out that card for an NVidia one soon. The VDPAU will be a nice bonus for the XBMC, too.
Oh, I wasn't targeting NVidia - they're pretty much the only big name who does have mostly decent drivers, as also pointed out somewhere else; apparently they're the only ones whitelisted iirc.
The Intel chipsets are also pretty decently supported, but they're not exactly the high end of the graphics market, and I can't much speak for their GL implementation. I do know that my current HP laptop with Intel graphics tends to lock up when I use GL screensavers.
Ati's drivers were more than horrible last time I touched them, to the point where I would reboot into windows, come back to Linux and find my resolution quasi-permanently stuck to 800x600; shit like that. Rebooting into Windows would of course magically work fine. The Linux resolution would somehow unstick when I reinstalled the system (yes, with the same distro, so no updated drivers), but that's not really a useful strategy.
I currently have a single Ati box left, attached to my HDMI TV; and for that I had to write a small script that calls the Ati tools to reset the origin and size of the picture so it actually uses the full screen size. Have to run that script every boot, but also every time I switch back to X from a console.
Non-ideological enough for you ?
It has long been established that you can control cats' movements using nothing but a laser pointer.