you could i suppose use some system of variable length numbers but the overhead of parsing and routing that would be huge
as for the dotted quad/ipv6 hex blocks thats simply a way of expressing the numbers for humans to easilly read it has nothing to do with ip itself (btw there is at least one protocol in common use that does send an ipv4 ip as one large decimal number)
the main reason ipv6 ips are so much larger was to actually SIMPLIFY routing/network admininstration by avoiding the need to worry about wasting addresses in the allocation plans.
i'd need to do some tests to be sure but i think linux iptables generally doesn't change the port unless it has to so it will generally work with such methods (and be identified as a restricted port cone).
i'm not entirely sure how it behaves when multiple lan machines use the same source port to contact different hosts though.
yes distro package managers with a sufficiantly anal design can mitigate a lot of the problems of bad installers but at the end of the day it only helps if your distro package manager supports the app you wan't.
iirc the original release of win95 didn't ship with IE either (i think osr2 did though).
whats also a pita is if you install IE 4 with windows desktop update on windows 95 (i've only tested this with original release dunno about osr2) then do an over the top reinstall of windows 95 then you CAN'T install IE4 on the resulting system (i found this one out the hard way).
if you plan to do an over the top reinstall on such a system you MUST remove IE first!
as you found you can accelerate the web by a HUGE amount if you control both ends of a high latency link. by using a protocol designed for the job.
tcp connection setup is pretty expensive in terms of round trips so if you keep a persistant connection (either tcp or something propietry) over the high latency link to handle resends etc (on the rare occasion they are needed) possiblly even with forward error correction to reduce resends further then you can reduce the overall delay in getting a page to just about 1 round trip over the link.
you CAN transfer data directly between two users behind most nats without needing to set up port forwarding etc its just that app designers have been slow to adopt the methods reqired.
it depends on exactly what you mean by the internet. and on how critical the system is.
buying two links from the same provider or from providers you know have good peering arrangements and making sure there is good infrasturcture in between is very different buying two links from the cheapest commodity isps and hoping they keep routing to each other properly with no comeback if they don't
similarly you may rely on grid power at home or in an office where nothing too critical is done but if you really need power all the time you will put backup systems in place.
before relying on anything you need to ask yourself whats the likely downtime,how pridictable that downtime is,how much that downtime will cost you, what you can do to reduce downtime and at what cost.
sorry from reading further down the comments i think i confused pipes used to condense water out of the air with pourus watering pipes (which do save a lot of water compared to say sprinklers)
sadly i can't go back and put the correction in the original post
a torrent-server (by which i mean something that can offer a whole tree of files by bittorrent) would be nice for one thing though and that is sites that host a lot of files which occasionally gets certain files hit big time (thing getting/.ed etc)
ie when one user is downloading a file at a time the torrent server just sends to them if a few people download the same file they help each other a bit but the server still does most of the work.
if a file gets/.ed the server restricts the bandwidth it offers for that one file and the torrent swarm takes over most of the load.
google for bnetd and you get a british site offering it
it doesn't work with the latest patch versions of the games though iirc 1.07 is the most recent version that works with it (it shouldn't be too hard to find the 1.07 patch on the net).
there is no security archive for them and upgrading everything from the debian testing and unstable upstream respositries on an install as big as knoppix is a pain in the arse.
i guess you could upgrade individual packages for security issues from the main debian testing/unstable archives but this would still be quite a bit of work keeping track of security issues.
imo best bet right now is to do a clean install of sarge edit the sources.list so it stays on sarge when sarge goes stable and set up a script for automatic security updates.
firstly have you never heared of a TCP window? TCP is NOT a send and wait protocol it sends a number of packets before getting a response and that number is dynamically adjusted based on the senders previous experiance with the connection (though admittedly changes did have to be made to tcp to properly allow for long fat networks like satalite).
secondly there was an analysis paper on sypes behaviour recently and it appears to only use TCP when it can't get through using UDP.
there are several different sat phone systems arround some (inmarsat for example) use geostationary others (irridium and globalstar for example) use low earth orbit.
trouble with low earth orbit is you need a huge amount of satalites to do it properly
if the feds can't crack a safe with the tools they have then they will presumaablly get more powerfull tools until they can
i'm sure that almost any safe could be cracked with sufficiant use of explosives etc
one possiblity is to build a self destruct mechanism into the safe but in many countries that in itself may be illigal (depending on how its done if it setriously injures people etc)
this seems to have become quite the in troll recently. some versions seem to have more variation from the original that others but they are generally pretty easy to recognise
ok so you could put dsl hardware on non backed up power but doing so would add a lot of wiring complexity (two seperate power systems would be a lot of installation) may cause complaints from users who do have backup power for thier end and could cause major problems if someone hooked up important kit up to the grid only power system.
most modern worms whilst they may have high damages in agregate are not nasty enough to the infected individuals to really cause people trouble.
i define a virus/worm as being really nasty if it actually destroys or manipulates user data pretty much anything else is annoying but not a major loss.
yeah making a server listen for both smtp and http on the same port would require using some kind of timeout. since smtp requires the server to send 220 to tell the client the connection is established.
having said that many smtp servers do an ident lookup with a fairly significant timeout before sending this greeting so mail clients should be prepared for such a timeout delay.
Slashdot Mirror
wtf do you mean by normal numbers?!
you could i suppose use some system of variable length numbers but the overhead of parsing and routing that would be huge
as for the dotted quad/ipv6 hex blocks thats simply a way of expressing the numbers for humans to easilly read it has nothing to do with ip itself (btw there is at least one protocol in common use that does send an ipv4 ip as one large decimal number)
the main reason ipv6 ips are so much larger was to actually SIMPLIFY routing/network admininstration by avoiding the need to worry about wasting addresses in the allocation plans.
i'd need to do some tests to be sure but i think linux iptables generally doesn't change the port unless it has to so it will generally work with such methods (and be identified as a restricted port cone).
i'm not entirely sure how it behaves when multiple lan machines use the same source port to contact different hosts though.
i said INSTALLERS not distro packages
yes distro package managers with a sufficiantly anal design can mitigate a lot of the problems of bad installers but at the end of the day it only helps if your distro package manager supports the app you wan't.
iirc the original release of win95 didn't ship with IE either (i think osr2 did though).
whats also a pita is if you install IE 4 with windows desktop update on windows 95 (i've only tested this with original release dunno about osr2) then do an over the top reinstall of windows 95 then you CAN'T install IE4 on the resulting system (i found this one out the hard way).
if you plan to do an over the top reinstall on such a system you MUST remove IE first!
oh i'm sure you could do it on *nix without too much trouble assuming you are root (which most app installers are going to need).
fact is if an installer runs amock on any platform you can end up with a pretty fucked up system.
as you found you can accelerate the web by a HUGE amount if you control both ends of a high latency link. by using a protocol designed for the job.
tcp connection setup is pretty expensive in terms of round trips so if you keep a persistant connection (either tcp or something propietry) over the high latency link to handle resends etc (on the rare occasion they are needed) possiblly even with forward error correction to reduce resends further then you can reduce the overall delay in getting a page to just about 1 round trip over the link.
if you have a public ipv4 ip then just use 6to4 and use the ipv6 /48 it gives you for basically as many machines as you wan't
you CAN transfer data directly between two users behind most nats without needing to set up port forwarding etc its just that app designers have been slow to adopt the methods reqired.
the teqniques needed are in RFC 3489
the main issue is that to use it you have to do everything over udp and that requires far more programmer effort to do well than tcp.
it depends on exactly what you mean by the internet. and on how critical the system is.
buying two links from the same provider or from providers you know have good peering arrangements and making sure there is good infrasturcture in between is very different buying two links from the cheapest commodity isps and hoping they keep routing to each other properly with no comeback if they don't
similarly you may rely on grid power at home or in an office where nothing too critical is done but if you really need power all the time you will put backup systems in place.
before relying on anything you need to ask yourself whats the likely downtime,how pridictable that downtime is,how much that downtime will cost you, what you can do to reduce downtime and at what cost.
sorry from reading further down the comments i think i confused pipes used to condense water out of the air with pourus watering pipes (which do save a lot of water compared to say sprinklers)
sadly i can't go back and put the correction in the original post
a torrent-server (by which i mean something that can offer a whole tree of files by bittorrent) would be nice for one thing though and that is sites that host a lot of files which occasionally gets certain files hit big time (thing getting /.ed etc)
/.ed the server restricts the bandwidth it offers for that one file and the torrent swarm takes over most of the load.
ie when one user is downloading a file at a time the torrent server just sends to them if a few people download the same file they help each other a bit but the server still does most of the work.
if a file gets
google for bnetd and you get a british site offering it
it doesn't work with the latest patch versions of the games though iirc 1.07 is the most recent version that works with it (it shouldn't be too hard to find the 1.07 patch on the net).
you either make the pipes out of something water can slowly flow through or you make little holes in them its not rocket science
is they are a pita to update.
there is no security archive for them and upgrading everything from the debian testing and unstable upstream respositries on an install as big as knoppix is a pain in the arse.
i guess you could upgrade individual packages for security issues from the main debian testing/unstable archives but this would still be quite a bit of work keeping track of security issues.
imo best bet right now is to do a clean install of sarge edit the sources.list so it stays on sarge when sarge goes stable and set up a script for automatic security updates.
firstly have you never heared of a TCP window? TCP is NOT a send and wait protocol it sends a number of packets before getting a response and that number is dynamically adjusted based on the senders previous experiance with the connection (though admittedly changes did have to be made to tcp to properly allow for long fat networks like satalite).
secondly there was an analysis paper on sypes behaviour recently and it appears to only use TCP when it can't get through using UDP.
there are several different sat phone systems arround some (inmarsat for example) use geostationary others (irridium and globalstar for example) use low earth orbit.
trouble with low earth orbit is you need a huge amount of satalites to do it properly
doesn't using dsl for the upstream kinda defeat the point of satalite internet?
if the feds can't crack a safe with the tools they have then they will presumaablly get more powerfull tools until they can
i'm sure that almost any safe could be cracked with sufficiant use of explosives etc
one possiblity is to build a self destruct mechanism into the safe but in many countries that in itself may be illigal (depending on how its done if it setriously injures people etc)
what i've never understood is why criminals on bail or sent a notice to come in for trial don't get the f*ck out of the country asap.
surely living in exile is better than going to prison (especially the US fedral pound me in the ass variety) for years.
this seems to have become quite the in troll recently. some versions seem to have more variation from the original that others but they are generally pretty easy to recognise
ok so you could put dsl hardware on non backed up power but doing so would add a lot of wiring complexity (two seperate power systems would be a lot of installation) may cause complaints from users who do have backup power for thier end and could cause major problems if someone hooked up important kit up to the grid only power system.
given all that would it really be worth doing?
most modern worms whilst they may have high damages in agregate are not nasty enough to the infected individuals to really cause people trouble.
i define a virus/worm as being really nasty if it actually destroys or manipulates user data pretty much anything else is annoying but not a major loss.
sure my point is why did this power cut cause major network outages?
do major datacenters in russia not bother with backup power?
are thier backup power systems in such a state of disrepair that a large proportion of them simply didn't work?
or whats going on?
yeah making a server listen for both smtp and http on the same port would require using some kind of timeout. since smtp requires the server to send 220 to tell the client the connection is established.
having said that many smtp servers do an ident lookup with a fairly significant timeout before sending this greeting so mail clients should be prepared for such a timeout delay.
the way i do it (slightly simplified from my real setup as i use this ssh link for other stuff and the sshd is on a nonstandard port):
ssh -L 25:127.0.0.1:25 user@server
then just set up the mail client to send to port 25 on localhost
(yes i know about -D and i do use it for other stuff but i use my mail client for other accounts that i don't wan't to go through an ssh tunnel)