Forciblly splitting up a few class A's won't buy us much (maybe an extra year or so). What will free up space for more profitable uses is making home users either go behind ISP level NAT or pay extra.
but it's my understanding that IPv6 addresses, unlike IPv4 addresses, include information about the backbone provider Normally it's the ISP (which may be anything from a small colo provider to a huge backbone providers) who gets a block and then they suballocate that block to thier customers. The same way a lot of IPV4 blocks are allocated.
The original intention was that multihomed sites would be given multiple IP blocks (one from each ISP) and would run multiple prefixes in paralell on every host. A frightfully complex addition to DNS known as A6 records was designed to support this but the system still didn't prove very practical.
So afaict they have gone back to doing things the old way and giving multihomed sites thier own prefixes.
I think the key point is that a security fix to visual studio changed the deployment needs of the apps built with it. At best that to mean a lot of head scratching as to why what used to work no longer does. At worst it could mean expensive missed deadlines or inadvertent deployment of broken code.
And while windows devs usually use installers for initial deployment consider things like auto-update mechanisms that are going to have to be changed to deal with this issue.
I think that's the wrong way around - the reason we haven't seen more widespread piracy of books is because they're difficult to pirate. You have to scan them in. That's a huge pain. It's a bit of a pain and the quickest method is destructive. But it's not unreasonablly hard and it only has to be done once to get the book onto the pirate network. Afaict (i've never looked myself) there are pirate networks out there with very good selections of books.
The annoyance of scanning will probablly reduce casual copying a lot but I don't think it's a problem for organised efforts.
I have written a huge estimation application (web-based). The calculations are something like O(n^n) or worse... point is, processors aren't getting faster. We are however getting a lot more cores. The algorithm parallelizes very well, but perl does not do efficient multi-threading. So I will have to re-write the estimation engine in something else. OTOH I strongly suspect that rewriting it in a faster programming language by itself is likely to bring a more significant speedup than going multithreaded. According to http://shootout.alioth.debian.org/u64q/benchmark.php?test=all&lang=all&box=1 perl perl is 70 times slower than C. Few boxes have more than 8 cores.
It makes the JVM look efficient The JVM is really VERY good by VM standards. Yeah the class libraries are a bit bloated leading to long startup times but other than that it is one of the few VMs to come anywhere close to (and in some benchmarks even exceed) the performance of tradtional compiled languages.
Palm is NOT within their rights to use Apple's Vendor ID to "pretend" their device is an iPod made by Apple. Only Apple is authorized to use their own Vendor ID, under the terms of the USB policy board. Palm, if they go that route, will be violating their terms of contract with the USB-IF and may suffer penalties for it. Of course the real question is what are the penalties for violating said contract and are they more or less than the gains from violating it.
I guess it depends on the attitude of your institution. During my time as an electronic systems engineering undergraduate at manchester in the UK I bought exactly one book (for a bloody enterprise module) for study reasons (I also purchased a book on windows drivers during my industrial year but the company I was working for paid me back for that and kept it at the end).
The general attitude here is that the lectures and their associated handouts should be enough to teach you the stuff you really need to know. If you are struggling and want a different perspective or you want more background there is usually a list of recommended books but they are in no way required and if you do want to look at them there are usually plenty of copies available in the library.
I think the HOSTS is parsed before DNS or any proxy is queried. Hosts is queried before DNS (other things may be queried too depending on the OS and configuration, e.g. WINS and netbios broadcast name resoloution) whenever the application asks the OS to look up a name.
So whether the clients hosts file is queried in a proxy setup depends on whether the application does name resouloution itself or leaves that up to the proxy. That in turn depends on the type of proxy and sometimes the configuration of the app.
BTW you are probably better off using one of the modern windows ports of duke3D (which have had the CD check removed) than a cracked version of the old dos executable.
IIRC early arcade models had no onboard storage but were supplied with a memory card in the box. Later arcade models added onboard flash. So not no storage but yes much less storage than the higher xbox models (of which all except the elite seem to have dissapeared now:( ) which come with hard drives.
Also MS were doing a promotion at one point where you could get a free hard drive for your arcade, no idea if it's still going or not.
The Japanese games are written for NTSC and have to be converted to PAL's dimensions This would have been a good excuse years ago but given that pals resoloution comes between NTSC and the lower of the HD resoloutions I doubt it's a big deal these days.
The games have to include voice-acting for multiple languages, not just one language Translation costs are indeed probablly part of it. Frankly though I suspect it's mostly just a matter of charging what they think the market will bear.
Actually the high cost is associated with VAT (http://en.wikipedia.org/wiki/Value_added_tax). Americans on the other hand pay taxes in other forms. The fact that EU prices are usually quoted inc VAT while US prices are usually quoted exclusive of sales tax does make up some of the difference. But even when you compare prices without VAT/sales tax there is still a significant difference.
One thing keeps me buying NVIDIA and that is the fact that some of my machines run linux and even if a machine doesn't run linux right now I want to keep the options open.
Afaict radeonhd isn't finished yet and i've always been even less impressed with fglrx than with nvidia's binary drivers. Intel GPUs are slow and in my experiance thier linux drivers while open source are buggy peices of shit.
And soon, it looks like that configuration will carry an additional caveat. Hooray. BTW that configuration isn't supported with wddm anyway. IIRC on vista (and probablly win7 too) you can make it work by forcing installation of xpdm drivers for all your graphics adaptors but afaict that will cost you support for DX10 and aero.
When a Gameboy Color starts up with a Super Gameboy boy game is put into a Super Game Boy, it uses the Super Gameboy Palette with the border that would normally be used on a TV omitted. No it doesn't, though it does colorise some known roms.
Pokemon Red/Blue/Yellow The colorisation of those games running in SGB mode (interestingly the way games check if they are running on a SGB is to check if the second controller exists) is pretty different from that which the gameboy color uses. The SGB colorisation changes during the game, the stuff applied by the gameboy color does not.
Alot of people thought that Pokemon games were Gameboy Color games, and some are, like Pokemon Crystal, but alot of the games are actually Super Gameboy Games. R/B/Y were gameboy games with support for the SGB pallete/border system. G/S were dual mode gameboy/gameboy color games, I don't know if they also had support for the SGB stuff. crystal was a gameboy color only game (I'm guessing they switched to color only to free up rom space for the new features they added)
My question is, will the dumping of this Bios lead to a better understanding of how Non-Super Gameboy Games are colorized on the Game Boy Color? It will get more details on how the games are identified and what the exact color values are.
It could also be interesting for flashcart menus if it reveals a way to switch from GBC mode to gameboy mode (this is normally selected by the GBC depending on a byte in the roms header).
And why is not more widespread the use of USB2 as networking port? USB is a master-slave system so you are going to need a special device to connect two hosts you can't use an ordinary USB hub or a direct cable. These special devices were pretty expensive (more than a cheap network hub/switch and a few cards IIRC) and they generally only had two ports.
Furthermore afaict (though I haven't dealt with that many computers from that era) ethernet was integrated on motherboards before USB 2 was.
Furthermore USBs range is terrible for networking purposes. 5M before you need a repeater and there is a limit to how many of those you can chain. Compare that to ethernet which can go 100m down cheap UTP.
Now, back in the real world, it becomes the bottleneck for even low-end, high-capacity storage devices built around traditional spinning media. With us now moving towards solid-state storage, USB 2.0 fails us horribly. We can only manage 30% to 35% read/write capacity utilization under real-world conditions. Agreed for connecting current desktop hard drives and high end solid state storage (note: with solid state storage the performance is largely a function of how much you are prepared to pay and to a lesser extent how much physical space you have to run chips in paralell)
Once you get above a resolution of 2000 pixels in either direction, USB 2.0 just can't handle it. Depends on the update speed and color depth you need. 1024x768 at 30fps and 24 bit color is already too much for USB!
USBNET2, basically IP networking over USB 2.0, never took off because it's just too damn slow. A google search for USBNET2 doesn't seem to find any obvious products. Please provide a link to the product or standard you are reffering to.
Generally I doubt performance was the reason USB networking products failed to take off unless it was really bad even by USB 2 standards. Many networks are still on 100 megabit or even 10 megabit ethernet.
USB is a master-slave protocol so to use it for networking requires a special device. These devices were in my experiance far more expensive than buying cheap network cards and hubs/switches. All the ones I saw also only had two ports. Furthermore most if not all the machines i've seen with onboard USB2 also had onboard ethernet.
There are many applications where we need much, much faster transfer rates than USB 2.0 can support. Agreed
Tell me *exactly* how you determine if someone can repay a loan that doesn't come down to the unscientific method of saying "He seems like an honest guy". You can calculate what someone can afford under the assumption that their circumstances don't change. Take the persons income. Subtract an estimate of what they can afford to live on). Self employed people are trickier to deal with but you can estimate tolerably from their past records. Then subtract some safety margin in case your estimates of thier minimum living costs are off. What you've got left is the maximum payment you can be reasonablly sure they can afford to make on a loan.
Yes there is the risk someone will lose their job, defraud you with false details when taking out the loan or that your estimates will be off. In that regard the bank is basically acting as an insurer. That is they calculate that their losses from defaults will be less than their profit on the difference between the interest rate they charge the borrower and their cost of funds.
Collateral reduces risk for the lender by giving them a way to recover some or maybe all of their money if the borrower defaults. As such it is reasonable for the lender to offer better loan deals to borrowers prepared to put up collateral and particularly collateral that is likely to retain it's value through the term of the loan (which is why mortgages do and should have lower interest rates than almost any other type of loan).
But while collateral reduces the risk it is IMO both evil and stupid to use collateral that appears to be rising in value as a substitute for checking that the borrower can reasonably pay the loan. All that does is drive a bubble and when that bubble pops both the lender and the customers who can't afford thier loans are screwed big time. Even the customers who can afford thier loans are in negative equity which basically means they are stuck in thier current property and unable to move.
I could have purchased a home in 2003 on an interest-only mortgage with all the other idiots, but I knew that I wouldn't be able to afford the payments once they included the principal. What people assumed is they would either get a payrise or be able to refinance onto a better deal since the value of thier property would have gone up. They assumed the worst case would be they have to flip the house. The only risk is if property prices go down which all the buyers assumed wouldn't happen (and the salemen deliberately kept them ignorant of).
Don't forget renting has risks too at least in areas that don't have rent controls. If you see property prices and rents both climbing ever skyward then an interest only mortgage starts to look like a good deal.
The tricky bit with replacing video with a general purpose interface would be to sort out signal routing inside the computer. There still needs to be a GPU/framebuffer and that GPU needs a high bandwidth path (we are talking a couple of PCIe 1.x lanes worth per display) from the framebuffer to the general purpose interface.
Not saying this couldn't be done but it would definately require cooperation between the GPU vendor and the vendor of the general purpose interface in question to allow them to communicate over PCIe without involving the CPU.
The obstacle I see with networking is that the world connects with RJ45, so for wired networking you'd still need an adapter. The other obstacle is security and management. Ethernet networks often contain (relatively) untrusted devices (so anything attatched to them will need to implement security and users will need to deal with setting that up) and are managed by someone other than the computers user (so getting more addresses can be nontrivial). Compare this to something like USB or firewire where it is assumed all devices are trusted so no security is needed.
Yes you can have multiple seperate ethernet networks and we geeks often do (especially those of us who work with embedded systems) but that kind of thing is difficult for the lusers to get right (even we geeks have to be pretty careful to make sure we don't go plugging a DHCP server intended for a private network into the departments main lan and bringing the wrath of IT on ourselves).
recent Macbooks The air never had firewire, probablly because it always designed as a cut down ultraslim machine.
The basic polycarbonate macbook has always had firewire 400.
All apples other current machines have firewire 800 (which is compatible with 400 with a wiring adaptor)
The 13 inch unibody didn't initially have firewire which many people at the time thought was a sign of apple dropping it. However either the pundits were wrong or apple decided the backlash was too much because soon afterwards the 13 inch unibody was redesignated as a macbook pro and had firewire 800 added.
Thanks for the info I did an image search just now which seemed to confirm your statement. The connector seems to have changed totally since the early pictures I saw.
And I mean good support, not just something that works like a stream, but where you can seek and do everything like you can do with actual files. Afaict the only way to read a portion of a file in a zip is to read and decompress the whole file up to the portion wanted so seeking is going to be pretty damn slow.
Slashdot Mirror
Sad but true
Forciblly splitting up a few class A's won't buy us much (maybe an extra year or so). What will free up space for more profitable uses is making home users either go behind ISP level NAT or pay extra.
I may be mistaken
You are a bit out of date.
but it's my understanding that IPv6 addresses, unlike IPv4 addresses, include information about the backbone provider
Normally it's the ISP (which may be anything from a small colo provider to a huge backbone providers) who gets a block and then they suballocate that block to thier customers. The same way a lot of IPV4 blocks are allocated.
The original intention was that multihomed sites would be given multiple IP blocks (one from each ISP) and would run multiple prefixes in paralell on every host. A frightfully complex addition to DNS known as A6 records was designed to support this but the system still didn't prove very practical.
So afaict they have gone back to doing things the old way and giving multihomed sites thier own prefixes.
I think the key point is that a security fix to visual studio changed the deployment needs of the apps built with it. At best that to mean a lot of head scratching as to why what used to work no longer does. At worst it could mean expensive missed deadlines or inadvertent deployment of broken code.
And while windows devs usually use installers for initial deployment consider things like auto-update mechanisms that are going to have to be changed to deal with this issue.
I think that's the wrong way around - the reason we haven't seen more widespread piracy of books is because they're difficult to pirate. You have to scan them in. That's a huge pain.
It's a bit of a pain and the quickest method is destructive. But it's not unreasonablly hard and it only has to be done once to get the book onto the pirate network. Afaict (i've never looked myself) there are pirate networks out there with very good selections of books.
The annoyance of scanning will probablly reduce casual copying a lot but I don't think it's a problem for organised efforts.
I have written a huge estimation application (web-based). The calculations are something like O(n^n) or worse... point is, processors aren't getting faster. We are however getting a lot more cores. The algorithm parallelizes very well, but perl does not do efficient multi-threading. So I will have to re-write the estimation engine in something else.
OTOH I strongly suspect that rewriting it in a faster programming language by itself is likely to bring a more significant speedup than going multithreaded. According to http://shootout.alioth.debian.org/u64q/benchmark.php?test=all&lang=all&box=1 perl perl is 70 times slower than C. Few boxes have more than 8 cores.
It makes the JVM look efficient
The JVM is really VERY good by VM standards. Yeah the class libraries are a bit bloated leading to long startup times but other than that it is one of the few VMs to come anywhere close to (and in some benchmarks even exceed) the performance of tradtional compiled languages.
According to http://shootout.alioth.debian.org/u64q/benchmark.php?test=all&lang=all&d=data&gpp=on&java=on&ghc=on&csharp=on&sbcl=on&hipe=on&mzscheme=on&lua=on&php=on&vw=on&python=on&perl=on&ruby=on&calc=calculate&box=1 java is behind C,C++ and some language i've never heard of but ahead of everything else. Hell even java in interpreter mode is not that far down the charts.
Palm is NOT within their rights to use Apple's Vendor ID to "pretend" their device is an iPod made by Apple. Only Apple is authorized to use their own Vendor ID, under the terms of the USB policy board. Palm, if they go that route, will be violating their terms of contract with the USB-IF and may suffer penalties for it.
Of course the real question is what are the penalties for violating said contract and are they more or less than the gains from violating it.
I guess it depends on the attitude of your institution. During my time as an electronic systems engineering undergraduate at manchester in the UK I bought exactly one book (for a bloody enterprise module) for study reasons (I also purchased a book on windows drivers during my industrial year but the company I was working for paid me back for that and kept it at the end).
The general attitude here is that the lectures and their associated handouts should be enough to teach you the stuff you really need to know. If you are struggling and want a different perspective or you want more background there is usually a list of recommended books but they are in no way required and if you do want to look at them there are usually plenty of copies available in the library.
I think the HOSTS is parsed before DNS or any proxy is queried.
Hosts is queried before DNS (other things may be queried too depending on the OS and configuration, e.g. WINS and netbios broadcast name resoloution) whenever the application asks the OS to look up a name.
So whether the clients hosts file is queried in a proxy setup depends on whether the application does name resouloution itself or leaves that up to the proxy. That in turn depends on the type of proxy and sometimes the configuration of the app.
BTW you are probably better off using one of the modern windows ports of duke3D (which have had the CD check removed) than a cracked version of the old dos executable.
IIRC early arcade models had no onboard storage but were supplied with a memory card in the box. Later arcade models added onboard flash. So not no storage but yes much less storage than the higher xbox models (of which all except the elite seem to have dissapeared now :( ) which come with hard drives.
Also MS were doing a promotion at one point where you could get a free hard drive for your arcade, no idea if it's still going or not.
The Japanese games are written for NTSC and have to be converted to PAL's dimensions
This would have been a good excuse years ago but given that pals resoloution comes between NTSC and the lower of the HD resoloutions I doubt it's a big deal these days.
The games have to include voice-acting for multiple languages, not just one language
Translation costs are indeed probablly part of it. Frankly though I suspect it's mostly just a matter of charging what they think the market will bear.
Actually the high cost is associated with VAT (http://en.wikipedia.org/wiki/Value_added_tax). Americans on the other hand pay taxes in other forms.
The fact that EU prices are usually quoted inc VAT while US prices are usually quoted exclusive of sales tax does make up some of the difference. But even when you compare prices without VAT/sales tax there is still a significant difference.
One thing keeps me buying NVIDIA and that is the fact that some of my machines run linux and even if a machine doesn't run linux right now I want to keep the options open.
Afaict radeonhd isn't finished yet and i've always been even less impressed with fglrx than with nvidia's binary drivers. Intel GPUs are slow and in my experiance thier linux drivers while open source are buggy peices of shit.
And soon, it looks like that configuration will carry an additional caveat. Hooray.
BTW that configuration isn't supported with wddm anyway. IIRC on vista (and probablly win7 too) you can make it work by forcing installation of xpdm drivers for all your graphics adaptors but afaict that will cost you support for DX10 and aero.
When a Gameboy Color starts up with a Super Gameboy boy game is put into a Super Game Boy, it uses the Super Gameboy Palette with the border that would normally be used on a TV omitted.
No it doesn't, though it does colorise some known roms.
Pokemon Red/Blue/Yellow
The colorisation of those games running in SGB mode (interestingly the way games check if they are running on a SGB is to check if the second controller exists) is pretty different from that which the gameboy color uses. The SGB colorisation changes during the game, the stuff applied by the gameboy color does not.
Alot of people thought that Pokemon games were Gameboy Color games, and some are, like Pokemon Crystal, but alot of the games are actually Super Gameboy Games.
R/B/Y were gameboy games with support for the SGB pallete/border system.
G/S were dual mode gameboy/gameboy color games, I don't know if they also had support for the SGB stuff.
crystal was a gameboy color only game (I'm guessing they switched to color only to free up rom space for the new features they added)
My question is, will the dumping of this Bios lead to a better understanding of how Non-Super Gameboy Games are colorized on the Game Boy Color?
It will get more details on how the games are identified and what the exact color values are.
It could also be interesting for flashcart menus if it reveals a way to switch from GBC mode to gameboy mode (this is normally selected by the GBC depending on a byte in the roms header).
And why is not more widespread the use of USB2 as networking port?
USB is a master-slave system so you are going to need a special device to connect two hosts you can't use an ordinary USB hub or a direct cable. These special devices were pretty expensive (more than a cheap network hub/switch and a few cards IIRC) and they generally only had two ports.
Furthermore afaict (though I haven't dealt with that many computers from that era) ethernet was integrated on motherboards before USB 2 was.
Furthermore USBs range is terrible for networking purposes. 5M before you need a repeater and there is a limit to how many of those you can chain. Compare that to ethernet which can go 100m down cheap UTP.
Now, back in the real world, it becomes the bottleneck for even low-end, high-capacity storage devices built around traditional spinning media. With us now moving towards solid-state storage, USB 2.0 fails us horribly. We can only manage 30% to 35% read/write capacity utilization under real-world conditions.
Agreed for connecting current desktop hard drives and high end solid state storage (note: with solid state storage the performance is largely a function of how much you are prepared to pay and to a lesser extent how much physical space you have to run chips in paralell)
Once you get above a resolution of 2000 pixels in either direction, USB 2.0 just can't handle it.
Depends on the update speed and color depth you need. 1024x768 at 30fps and 24 bit color is already too much for USB!
USBNET2, basically IP networking over USB 2.0, never took off because it's just too damn slow.
A google search for USBNET2 doesn't seem to find any obvious products. Please provide a link to the product or standard you are reffering to.
Generally I doubt performance was the reason USB networking products failed to take off unless it was really bad even by USB 2 standards. Many networks are still on 100 megabit or even 10 megabit ethernet.
USB is a master-slave protocol so to use it for networking requires a special device. These devices were in my experiance far more expensive than buying cheap network cards and hubs/switches. All the ones I saw also only had two ports. Furthermore most if not all the machines i've seen with onboard USB2 also had onboard ethernet.
There are many applications where we need much, much faster transfer rates than USB 2.0 can support.
Agreed
Tell me *exactly* how you determine if someone can repay a loan that doesn't come down to the unscientific method of saying "He seems like an honest guy".
You can calculate what someone can afford under the assumption that their circumstances don't change. Take the persons income. Subtract an estimate of what they can afford to live on). Self employed people are trickier to deal with but you can estimate tolerably from their past records. Then subtract some safety margin in case your estimates of thier minimum living costs are off. What you've got left is the maximum payment you can be reasonablly sure they can afford to make on a loan.
Yes there is the risk someone will lose their job, defraud you with false details when taking out the loan or that your estimates will be off. In that regard the bank is basically acting as an insurer. That is they calculate that their losses from defaults will be less than their profit on the difference between the interest rate they charge the borrower and their cost of funds.
Collateral reduces risk for the lender by giving them a way to recover some or maybe all of their money if the borrower defaults. As such it is reasonable for the lender to offer better loan deals to borrowers prepared to put up collateral and particularly collateral that is likely to retain it's value through the term of the loan (which is why mortgages do and should have lower interest rates than almost any other type of loan).
But while collateral reduces the risk it is IMO both evil and stupid to use collateral that appears to be rising in value as a substitute for checking that the borrower can reasonably pay the loan. All that does is drive a bubble and when that bubble pops both the lender and the customers who can't afford thier loans are screwed big time. Even the customers who can afford thier loans are in negative equity which basically means they are stuck in thier current property and unable to move.
I could have purchased a home in 2003 on an interest-only mortgage with all the other idiots, but I knew that I wouldn't be able to afford the payments once they included the principal.
What people assumed is they would either get a payrise or be able to refinance onto a better deal since the value of thier property would have gone up. They assumed the worst case would be they have to flip the house. The only risk is if property prices go down which all the buyers assumed wouldn't happen (and the salemen deliberately kept them ignorant of).
Don't forget renting has risks too at least in areas that don't have rent controls. If you see property prices and rents both climbing ever skyward then an interest only mortgage starts to look like a good deal.
Techically the bandwidth would be there.
The tricky bit with replacing video with a general purpose interface would be to sort out signal routing inside the computer. There still needs to be a GPU/framebuffer and that GPU needs a high bandwidth path (we are talking a couple of PCIe 1.x lanes worth per display) from the framebuffer to the general purpose interface.
Not saying this couldn't be done but it would definately require cooperation between the GPU vendor and the vendor of the general purpose interface in question to allow them to communicate over PCIe without involving the CPU.
The obstacle I see with networking is that the world connects with RJ45, so for wired networking you'd still need an adapter.
The other obstacle is security and management. Ethernet networks often contain (relatively) untrusted devices (so anything attatched to them will need to implement security and users will need to deal with setting that up) and are managed by someone other than the computers user (so getting more addresses can be nontrivial). Compare this to something like USB or firewire where it is assumed all devices are trusted so no security is needed.
Yes you can have multiple seperate ethernet networks and we geeks often do (especially those of us who work with embedded systems) but that kind of thing is difficult for the lusers to get right (even we geeks have to be pretty careful to make sure we don't go plugging a DHCP server intended for a private network into the departments main lan and bringing the wrath of IT on ourselves).
see 5G iPod,
Granted
recent Macbooks
The air never had firewire, probablly because it always designed as a cut down ultraslim machine.
The basic polycarbonate macbook has always had firewire 400.
All apples other current machines have firewire 800 (which is compatible with 400 with a wiring adaptor)
The 13 inch unibody didn't initially have firewire which many people at the time thought was a sign of apple dropping it. However either the pundits were wrong or apple decided the backlash was too much because soon afterwards the 13 inch unibody was redesignated as a macbook pro and had firewire 800 added.
Thanks for the info I did an image search just now which seemed to confirm your statement. The connector seems to have changed totally since the early pictures I saw.
And I mean good support, not just something that works like a stream, but where you can seek and do everything like you can do with actual files.
Afaict the only way to read a portion of a file in a zip is to read and decompress the whole file up to the portion wanted so seeking is going to be pretty damn slow.