Most IRCds can be configured to do this. The issue is not automatic rerouting, it's that while rerouting takes place the network is disrupted. IRC, as specified, is a tree-like structure; there can only ever be one path from node A to node B, regardless of how many nodes are in the network.
A more mesh-like protocol would use more bandwidth overall (some duplicate messages) and more memory (to ensure that the same message recieved twice by one server is handled correctly) but be much, much more resistant to netsplit-based attacks and general network flakiness.
It hasnt made it on slashdot yet, but netcraft is reporting that future versions of IE will no longer be supporting user information in HTTP or HTTPS URLs.
Well, the ".root-servers.net" servers only tell you where to find.com,.net,.cx,.us, and so on. A good chunk of those (.com and.net at the very least) reside on servers in ".gtld-servers.net", which could also be considered root.
While the.root-servers.net servers only get requeried by any given client DNS server fairly rarely, the.gtld-servers.net servers get queried very often. If you go to www.example.com and you've never been there before, you'll end up asking a.gtld-servers.net server for example.com's name servers. The answer will come back with a TTL (time to live), which will probably be rather shorter than the TTL on com. Hosts within that domain, like www, might have an even shorter TTL, depending on how frequently that IP might change.
I'm tempted to mark this off as firewall cargo-cultism. Barring bugs in the kernel, a port that is not listening cannot accept packets. Given that iptables runs in the same kernel, I'm inclined to believe that it's equally-likely to have a bug that causes it to accept packets it should drop/reject.
With a read-only system such as knoppix, it's extremely hard for a trojan to set up a rogue service on a high port for long enough to be useful to an attacker, and next to impossible to arrange for that to come back next time the system is rebooted.
Without copyright, nobody would have a legal right to prevent others from copying music
B does not follow from A, here: removing copyright protection doesn't necessarily make copy-protection illegal. Without copyright, 'rights management' schemes (DRM, copy protection, etc) would become the only way to 'control' your work's distribution. I think we'd see a lot more cockamamie DRM schemes, not less.
I don't get this. I am free to make as many copies of some text or music as I want, so long as I keep them all and destroy/transfer them if I transfer the original work. So why can't I freely copy my software into my RAM? It's certainly not distribution.
And if a machine dies (think "power failure" or "kernel panic") before decrementing the mounted count file? The grandparent's solution doesn't have that limitation, in exchange for a little processor time and a little bandwidth.
i mean, after all, we all turn off ping before we put our servers up... don't we?
No, as a matter of fact I don't turn off ECHO responses on boxes I manage. I prefer to be able to tell if an operating system or tcp/ip stack has fallen over without having to go over and hook up a console. I'm actually rather annoyed at certain ISPs for continuing to block ping even after Welchia and Slammer have mostly abated.
Which is not to say you can't turn off pings on your boxes, but neither your preference nor mine is everyone's preference.
No, as the HTTP error is handled in an entirely unrelated connection prior to the one to the destination site. Some user agents may set a Referer: header but this behaviour is not mandatory.
I don't find it hypocritical at all. The broadcast flag by itself does not make it illegal to record the 'protected' material. It only advises the recieving hardware not to record the stream. The fact that a GPLed piece of software is capable of ignoring the flag doesn't make itself responsible for the user violating copyright by distributing copyrighted material recorded with that tool.
I've read somewhere ( don't remember the place, sorry ) that some old versions of some browsers had issues about ISO char tables.
That'd be Netscape 4.x, which you can and should ignore as much as possible. It doesn't even get &#(number) character references right unless the document is in UTF-8 in the first place, in which case you can mostly get away with using the original character.
One possible (and likely) source of character mangling is an incorrect Content-Type HTTP header. The <meta http-equiv="Content-Type: text/html; charset=ISO-8859-1"> (or whatever character encoding you're using) hack works, most of the time, but you're almost always better off simply configuring your web server to send the right information in the first place.
I'd suggest nosing around groups.google in comp.infosystems.www.authoring.html -- there's a lot of useful information in the archives, and the people who post there have a fairly large amount of Clue.
Furthermore, writing accuented text in plain HTML is such a pain in the ass it's not even funny. You have to type stuff like "é" instead of a sole key on a French keyboard ( I'm French-speaking ), and since most languages have non-standard - according to English, that is... - characters and that these are very common in text for some languages, I think such a feature is essential to a top notch international HTML editor.
Or you could write your page in the ISO-8859-1 character encoding rather than US-ASCII (you should be anyways, even if you're only using english), which has accented characters at U+00C0 through U+00FF (with interspersed non-accent characters). Not knowing how to use the tool is not the same thing as having a bad tool.
Aside: Slashdot's comment box only allows characters in the printable range of US-ASCII, so I can't include any examples here. Try Everything2's LATIN-1 writeup.
With this many IP addresses, there's no reason why every connection can't be given 255 (or more) IPs. For example, I connect with my cable modem. Where's the hurt in giving me 255 IPs to use?
While that's certainly the sensible point of view, who says that ISPs, especially large commercial providers, are going to break with the one-connection, one-machine business model they've held so far? While they currently allow NAT because there's really no technical way to prevent it, connecting more than one computer is still against most ISPs' terms of service.
Ultimately, you're coming at it from the wrong end, asking why they shouldn't give you more than one. I suspect the thought processes are closer to "so why on earth should we give out more than one IPv6 address?"
You mean Hocus Pocus? That game was passing fun, but it had some issues where you could get yourself stuck, because special effect potions would run out before you got to where you needed to be unless you ran there.
I specifically didn't mention OpenBeOS because I do not think of it as an OS without cruft. Their project right now is to rebuild BeOS, but what happens when they reach that point? I rather suspect they'll do the obvious thing: extend it.
Tell me, does Linux have any cruft left over from extending UNIX? Does Windows have any cruft left over from extending Windows?
I know from here it's hard to believe, but even within the tech world there are a lot of people who actually believe, for one reason or another, that what Microsoft is doing and has created is the best thing since sliced bread.
Not to mention that anyone who wanted BeOS for free could just download it...
Actually, anyone who already knew they wanted BeOS could just download it. The OEM installs OSes so that the end user doesn't have to worry about it. How many Windows users do you think specifically sat down and thought "I want Windows?" By the same token, the Walmart Lindows boxes have actually sold -- when I was working for Road Runner I took support calls for a couple of them (sadly I couldn't help them due to corporate policy) and the people I've talked to (not just the support cases) seem pretty happy with it.
Not to say that Microsoft is solely responsible for the current situation, and on a level you are correct. There is not much business in alternate operating systems for OEMs right now.
Slashdot Mirror
Most IRCds can be configured to do this. The issue is not automatic rerouting, it's that while rerouting takes place the network is disrupted. IRC, as specified, is a tree-like structure; there can only ever be one path from node A to node B, regardless of how many nodes are in the network.
A more mesh-like protocol would use more bandwidth overall (some duplicate messages) and more memory (to ensure that the same message recieved twice by one server is handled correctly) but be much, much more resistant to netsplit-based attacks and general network flakiness.
Down with spanning tree!
Judge Naut, lest ye be Judged?
Yes. Don't.
(For reference port 113 is the 'ident' identification protocol. Anyone using this for serious authentication should be shot.)
You misspelled "we".
Beg your pardon?
Well, the ".root-servers.net" servers only tell you where to find .com, .net, .cx, .us, and so on. A good chunk of those (.com and .net at the very least) reside on servers in ".gtld-servers.net", which could also be considered root.
While the .root-servers.net servers only get requeried by any given client DNS server fairly rarely, the .gtld-servers.net servers get queried very often. If you go to www.example.com and you've never been there before, you'll end up asking a .gtld-servers.net server for example.com's name servers. The answer will come back with a TTL (time to live), which will probably be rather shorter than the TTL on com. Hosts within that domain, like www, might have an even shorter TTL, depending on how frequently that IP might change.
I'm tempted to mark this off as firewall cargo-cultism. Barring bugs in the kernel, a port that is not listening cannot accept packets. Given that iptables runs in the same kernel, I'm inclined to believe that it's equally-likely to have a bug that causes it to accept packets it should drop/reject.
With a read-only system such as knoppix, it's extremely hard for a trojan to set up a rogue service on a high port for long enough to be useful to an attacker, and next to impossible to arrange for that to come back next time the system is rebooted.
Nah. In C++ main() should be either "int main ()" or "int main (int, char**)" or something equivalent (typedefs, char*[], etc).
Start here.
IHBT.
B does not follow from A, here: removing copyright protection doesn't necessarily make copy-protection illegal. Without copyright, 'rights management' schemes (DRM, copy protection, etc) would become the only way to 'control' your work's distribution. I think we'd see a lot more cockamamie DRM schemes, not less.
I don't get this. I am free to make as many copies of some text or music as I want, so long as I keep them all and destroy/transfer them if I transfer the original work. So why can't I freely copy my software into my RAM? It's certainly not distribution.
And if a machine dies (think "power failure" or "kernel panic") before decrementing the mounted count file? The grandparent's solution doesn't have that limitation, in exchange for a little processor time and a little bandwidth.
No, as a matter of fact I don't turn off ECHO responses on boxes I manage. I prefer to be able to tell if an operating system or tcp/ip stack has fallen over without having to go over and hook up a console. I'm actually rather annoyed at certain ISPs for continuing to block ping even after Welchia and Slammer have mostly abated.
Which is not to say you can't turn off pings on your boxes, but neither your preference nor mine is everyone's preference.
No, as the HTTP error is handled in an entirely unrelated connection prior to the one to the destination site. Some user agents may set a Referer: header but this behaviour is not mandatory.
I don't find it hypocritical at all. The broadcast flag by itself does not make it illegal to record the 'protected' material. It only advises the recieving hardware not to record the stream. The fact that a GPLed piece of software is capable of ignoring the flag doesn't make itself responsible for the user violating copyright by distributing copyrighted material recorded with that tool.
That'd be Netscape 4.x, which you can and should ignore as much as possible. It doesn't even get &#(number) character references right unless the document is in UTF-8 in the first place, in which case you can mostly get away with using the original character.
One possible (and likely) source of character mangling is an incorrect Content-Type HTTP header. The <meta http-equiv="Content-Type: text/html; charset=ISO-8859-1"> (or whatever character encoding you're using) hack works, most of the time, but you're almost always better off simply configuring your web server to send the right information in the first place.
I'd suggest nosing around groups.google in comp.infosystems.www.authoring.html -- there's a lot of useful information in the archives, and the people who post there have a fairly large amount of Clue.
Or you could write your page in the ISO-8859-1 character encoding rather than US-ASCII (you should be anyways, even if you're only using english), which has accented characters at U+00C0 through U+00FF (with interspersed non-accent characters). Not knowing how to use the tool is not the same thing as having a bad tool.
Aside: Slashdot's comment box only allows characters in the printable range of US-ASCII, so I can't include any examples here. Try Everything2's LATIN-1 writeup.
Or, if you're using Windows 2000/XP, use the built-in NTP client.
/setsntp:your.ntp.server.here
> net time
> net start w32time
Or go into the Services manager and set it to start automatically.
While that's certainly the sensible point of view, who says that ISPs, especially large commercial providers, are going to break with the one-connection, one-machine business model they've held so far? While they currently allow NAT because there's really no technical way to prevent it, connecting more than one computer is still against most ISPs' terms of service.
Ultimately, you're coming at it from the wrong end, asking why they shouldn't give you more than one. I suspect the thought processes are closer to "so why on earth should we give out more than one IPv6 address?"
You mean Hocus Pocus? That game was passing fun, but it had some issues where you could get yourself stuck, because special effect potions would run out before you got to where you needed to be unless you ran there.
Did you miss the citation on that particular definition?
I specifically didn't mention OpenBeOS because I do not think of it as an OS without cruft. Their project right now is to rebuild BeOS, but what happens when they reach that point? I rather suspect they'll do the obvious thing: extend it.
Tell me, does Linux have any cruft left over from extending UNIX? Does Windows have any cruft left over from extending Windows?
I know from here it's hard to believe, but even within the tech world there are a lot of people who actually believe, for one reason or another, that what Microsoft is doing and has created is the best thing since sliced bread.
Creepy, isn't it?
Actually, anyone who already knew they wanted BeOS could just download it. The OEM installs OSes so that the end user doesn't have to worry about it. How many Windows users do you think specifically sat down and thought "I want Windows?" By the same token, the Walmart Lindows boxes have actually sold -- when I was working for Road Runner I took support calls for a couple of them (sadly I couldn't help them due to corporate policy) and the people I've talked to (not just the support cases) seem pretty happy with it.
Not to say that Microsoft is solely responsible for the current situation, and on a level you are correct. There is not much business in alternate operating systems for OEMs right now.