If someone "needs" 5 isos, it makes *far* more sense to talk to a local administrator ("You know, it would be really nice if we ran a local mirror of ftp.redhat.com" or whatever). That way, *he* sets up a mirror accessable to local users, the files get downloaded *once* at off-hours, and then they're accessable rapidly to any local users.
The answer, at least in my opinion, is via a QoS mechanism.
The problem is that you can't have students sucking down gigs of bandwidth to grab the lastest porn flicks off of the gnutellaNet, because it costs you too much to keep them and your "legit" users happy. So set up a QoS system. I'd probably like to have a quota of bandwidth that each person gets per month...and after they've exhausted that bandwidth, they only get network space if there's free space on the network -- their priority drops. So if 128.2.154.2 is sucking down more than his fair share and exhausts his entire quota in the first day of the month. After that, his priority at the router gets knocked down to "two" and his performance suffers. If the network's already jammed, his packet is the first to get dropped. That way, you let people who want to do P2P do P2P, and keep the people who just want a snappy SSH server keep a snappy SSH server.
Since you don't really need real-time response (calculating used bandwidth once an hour in a perl script or something is more than enough), you can do this offline. If I were using a Linux router:
Set up iptables on each router so that you have a chain that sums the bandwidth used by each host in the network that it routes to. Hourly, poll each of the routers and get the latest usage statistics, and regenerate prioritization rulesets based on these. Send these back out to the routers.
Since you can do this offline at your NOC, you can do fancy stuff like sum all the bandwidth used by all the IPs allocated to a single user and stuff like that. Give each user 2GB/month, and if they want to use 1GB on their laptop and 500MB on each of their two desktops, that's okay too.
There is a few potential problems. Technically advanced students could try setting up VPNs. Shouldn't be a huge issue, just means that a slightly larger body of people get 100% utilization of quota.
IP spoofing is always a potential issue, but no end of problems can be caused by IP spoofing already, and the consequences aren't *disasterous* in this case -- if a massive flood of spoofed data is slipped by the sysadmin, the victim would just get somewhat worse performance.
Now, that assumes that the bottleneck is at the outgoing connection to your installation. If it's the LAN and your box is hooked up to a simple switch or hub...well, not much you can do there.
Finally, it's difficult for students to "find loopholes" in rulesets that detect whether software is P2P or not and take advantage of them. Many suggestions that try to rate-limit P2P traffic and P2P traffic alone are vulnerable to this.
That being said, it's also nice to run a big Web opaque proxy server with a policy of no logging (most people get leery of optional proxy servers if they log what they're doing). Also, if you have a bunch of hard drives sitting around, you can set up a Freenet node and do the same thing -- have a big local cache for users
Unless some of the other (i.e. non Sino-govt-propoganda site) news sites are getting bogus information, you're going to need one hell of a cluster. These Intel-killers are supposed to run with the blistering power of a 486. Some other people have posted saying that "they don't mind if it's a bit slower, as long as it doesn't have Palladium." Well, here's their chance to put their money where their mouth is.
If the article I linked to is right, China plans to go from 486 to Pentium III level performance in one more year, and then to "the then internationally advanced level in 2005" in 2005.
That some hella good espionage, if they can pull that off. Moore's Law? Hell, the Chinese can set a new standard of their own, they'd be going so much faster.
On the up side, if they can ban (or throw heavy tariffs on) imported chips from Intel/AMD, I suspect Linux use will skyrocket. Why? Well, if you've ever tried using Windows XP on a 486, you might have some idea...:-)
So you're saying that you can bypass learning to grip and get to learning to manipulat objects...but learning to grip is kind of important too, you know?
First of all, the registry is a DATABASE. It's not loaded into memory in its entirety at once!
What does being a database have to do with being in RAM or on disk? You can use (and I've seen a number of lightweight projects) using CSV as a database backend storage format, if you want a specific counterexample.
Each branch has permissions defined by the system's (or NT Domain's) ACLs, just like NTFS!
Yup. How much nicer it would be if MS had chosen to simply unify the two and put the data on disk as files.
The registry can be backed up and restored simply in a multitude of ways. Just because you don't know how doesn't mean it can't be done.
I've done some work with registry bits that are used to boot Windows. If you manage to bork them, you can't recover because you can't edit a dead registry from the Recovery Console -- I even had a second copy of Windows on the same machine just so that I could recover. No go. I did back up the registry entire and then overwrite it, but in terms of backing up and doing recover, the registry really sucks compared to an equivalent tree in the filesystem. It also means that all the tools you have to deal with the filesystem can't do anything with the registry...
A lot better than text files. Faster and more organized.
It is nice to have a single unified utility that's capable of presenting all your settings hierarchically.
That being said, I really don't consider it worth the pain the registry inflicts. You can have hierarchies with text files, making them just as organized. Also, while it's probably faster, that also means that programs constantly check the registry (and some annoying ones poll it). Most UNIX programs parse and store the contents of their dotfiles at startup.
When you can prove that that unorganized shithole known as/etc/ (or is it {$HOME}/? Or something else? Face it - there is NO consistency under Unix./etc/ for systemwide settings, $HOME for user-level settings. Not that complicated.
Obviously you've never taken a single reasonably difficult programming class in your life.
Yes, I've noticed how UNIX folks tend to be less programmers than Windows folks. Uh, huh.
I understand that the GIMP has a decidedly different UI from what Windows users are accustomed to...heck, different from GNOME.
However, I simply cannot agree on the MDI modification.
First, GIMP's current UI is very good for multiple viewports, where you can spread it out across multiple desktops. MDI would take that away. Even on a single viewport you can put some palletes in the back if you don't need them.
Second, one common complaint about GIMP is its complete and utter lack of modality. There are no dialog boxes that come up and prevent you from doing things. In the middle of setting some plugin settings? Just flip to another window and do some other work. This can get confusing to people that are used to Photoshop -- but I'm quite certain that while this approach is unfamiliar, it's much better. You're never locked in to doing a particular step.
Finally, it would be nice to have "palette" style windows, but unfortunately X11 doesn't support a palette (or in Mac OS UI terms, a "windoid") window style. It would be incredibly nifty if it did...
Text-based interactive fiction contains some of the most amazing games ever made, and most are free.
There are several different IF environments -- TADS and Inform are the most popular, playable by TADS and Frotz, respectively.
There are many incredible games for both, but two of my favorites are Babel and Toonesia. This type of game loses most of its value if you cheat -- most of the value of the game is in gameplay.
Give it a shot, and rack your brains...and don't get eaten by a grue.
Hey, I thought you only trolled! What's with the good suggestions?
PuTTY is a work of art. It's free, it's flexible, it's self-contained in one executable (no bloody installers!)...it's the single Windows program that I must have if I'm using a Windows machine.
It can even emulate X11's middle-click-to-paste and select-to-copy mode. Absolutely wonderful.
But then they would have had to test two totally different KDEs and support them. That's non-trivial, and even more confusing to users trying to figure out what's being installed. You'd have KDE+GNOME style KDE, *and* vanilla KDE (which would have to be packaged separately and eat more space, as well, and would have to be marked as incompatible with KDE+GNOME style KDE).
If someone really wants KDE, they can get it straight from kde.org -- I get a fair amount of software from the original source. The KDE project can distribute their blessed distribution however they want to do so. If you *really* want vanilla KDE with no GNOME integration done at all, then there are distros like Mandrake that *have* taken this approach...but I like seeing the diversity among distros that characterizes Linux. If someone wants to make a distro that has a totally terminal-based UI through an X server and AAlib and runs Enlightenment...well, they can do so. The users will end up voting by using whichever they prefer -- it could be that Red Hat is doing the wrong thing or the right thing. People will try both, and comment on them. RH will probably polish things up in 8.1, and if people still don't like the approach...well, Mandrake will get them.
Yeah. People tend to give people with a "Professor" in front of their name more credence. Good nick choice.
I've been studying psychoacoustics in my spare time
Heh...in my experimentations.
Heh....concentrates its encoding in the more audible midrange...
Yup. Be a pretty pathetic audio codec if it didn't....because the brain unconsciously uses inaudible overtones as a guide to determinine the tone of the music...
Given 128kbps, I should expend more data on the inaudible portion of the spectrum to hint people as to what the audible tone is that I'm already emphasizing in my encoding? Uh, huh.
The way that Vorbis compresses its audio accelerates speaker degredation
Actually, if I remember correctly, Red Hat *does* have KDE as an install option currently (as opposed to GNOME). The KDE guys didn't like it because GNOME was the default checked option in the installer.
Second, and more importantly, they have replaced KDE apps with equivalent apps, either from GNOME or independent projects. For example, they replaced konqueror with Mozilla, Koffice with OpenOffice, KMail with Evolution.
Yes. So? The alternative would be using all KDE apps in their place. So the KDE people are pissed off because RH isn't excising all competitors to their software from their distro and using only KDE? That's just silly.
Very true. Red Hat is doing something that consumers have wanted (and Linux tweakers have been doing for some time with themes) -- making their KDE and GNOME apps look similar.
KDE people whining about this are going to be ignored. The GNOME people have accepted the loss of their icons without throwing fits, though it certainly changes RH GNOME's "look and feel". I can't figure out why the KDE people can't do the same for the changes that affected their "look and feel."
Right up until the release announcement, news.google.com was live, had a *great* layout consisting of really simple, single column text with no images or tables, and was even easier to navigate. I loved it.
Then Google changed things to look like a traditional news site.:-(
The Mhz != Mhz argument is valid, but Apple's stretched it waaay beyond reality. The PPC chips *do* run cooler and *do* have a more sane instruction set, and for a lot of people that is valuable, but as for raw computational power? Nope...x86 chips are way out in front.
Oh, a better quote from the article:
Dammit, Apple, if you're moving to x86 you simply aren't going to be a competitive systems manufacturer any more. You're going to have to accept that and either sell add-ons or only sell software.
"Yes, so we've accepted the nuclear weapons development contract. We need to ship in a month, so we need the testing cluster up tomorrow. You've got tonight to purchase and set up a 300 node cluster."
Slashdot Mirror
If someone "needs" 5 isos, it makes *far* more sense to talk to a local administrator ("You know, it would be really nice if we ran a local mirror of ftp.redhat.com" or whatever). That way, *he* sets up a mirror accessable to local users, the files get downloaded *once* at off-hours, and then they're accessable rapidly to any local users.
Take a look at femfind
And have their own P2P solutions, I would assume? :-)
The answer, at least in my opinion, is via a QoS mechanism.
The problem is that you can't have students sucking down gigs of bandwidth to grab the lastest porn flicks off of the gnutellaNet, because it costs you too much to keep them and your "legit" users happy. So set up a QoS system. I'd probably like to have a quota of bandwidth that each person gets per month...and after they've exhausted that bandwidth, they only get network space if there's free space on the network -- their priority drops.
So if 128.2.154.2 is sucking down more than his fair share and exhausts his entire quota in the first day of the month. After that, his priority at the router gets knocked down to "two" and his performance suffers. If the network's already jammed, his packet is the first to get dropped. That way, you let people who want to do P2P do P2P, and keep the people who just want a snappy SSH server keep a snappy SSH server.
Since you don't really need real-time response (calculating used bandwidth once an hour in a perl script or something is more than enough), you can do this offline. If I were using a Linux router:
Set up iptables on each router so that you have a chain that sums the bandwidth used by each host in the network that it routes to. Hourly, poll each of the routers and get the latest usage statistics, and regenerate prioritization rulesets based on these. Send these back out to the routers.
Since you can do this offline at your NOC, you can do fancy stuff like sum all the bandwidth used by all the IPs allocated to a single user and stuff like that. Give each user 2GB/month, and if they want to use 1GB on their laptop and 500MB on each of their two desktops, that's okay too.
There is a few potential problems. Technically advanced students could try setting up VPNs. Shouldn't be a huge issue, just means that a slightly larger body of people get 100% utilization of quota.
IP spoofing is always a potential issue, but no end of problems can be caused by IP spoofing already, and the consequences aren't *disasterous* in this case -- if a massive flood of spoofed data is slipped by the sysadmin, the victim would just get somewhat worse performance.
Now, that assumes that the bottleneck is at the outgoing connection to your installation. If it's the LAN and your box is hooked up to a simple switch or hub...well, not much you can do there.
Finally, it's difficult for students to "find loopholes" in rulesets that detect whether software is P2P or not and take advantage of them. Many suggestions that try to rate-limit P2P traffic and P2P traffic alone are vulnerable to this.
That being said, it's also nice to run a big Web opaque proxy server with a policy of no logging (most people get leery of optional proxy servers if they log what they're doing). Also, if you have a bunch of hard drives sitting around, you can set up a Freenet node and do the same thing -- have a big local cache for users
in a beowolf cluster
:-)
Unless some of the other (i.e. non Sino-govt-propoganda site) news sites are getting bogus information, you're going to need one hell of a cluster. These Intel-killers are supposed to run with the blistering power of a 486. Some other people have posted saying that "they don't mind if it's a bit slower, as long as it doesn't have Palladium." Well, here's their chance to put their money where their mouth is.
If the article I linked to is right, China plans to go from 486 to Pentium III level performance in one more year, and then to "the then internationally advanced level in 2005" in 2005.
That some hella good espionage, if they can pull that off. Moore's Law? Hell, the Chinese can set a new standard of their own, they'd be going so much faster.
On the up side, if they can ban (or throw heavy tariffs on) imported chips from Intel/AMD, I suspect Linux use will skyrocket. Why? Well, if you've ever tried using Windows XP on a 486, you might have some idea...
It's not for practical use. The point is that the thing can run Linux pretty solidly -- enough to pull something like this off.
Plus, it's a "technically cool" project, not one intended to replace your desktop machine.
This is the same guy (different account) that posted the BS about Ogg Vorbis yesterday and got shot down by Monty. Moderators should mod accordingly.
So you're saying that you can bypass learning to grip and get to learning to manipulat objects...but learning to grip is kind of important too, you know?
First of all, the registry is a DATABASE. It's not loaded into memory in its entirety at once!
/etc/ (or is it {$HOME}/? Or something else? Face it - there is NO consistency under Unix. /etc/ for systemwide settings, $HOME for user-level settings. Not that complicated.
What does being a database have to do with being in RAM or on disk? You can use (and I've seen a number of lightweight projects) using CSV as a database backend storage format, if you want a specific counterexample.
Each branch has permissions defined by the system's (or NT Domain's) ACLs, just like NTFS!
Yup. How much nicer it would be if MS had chosen to simply unify the two and put the data on disk as files.
The registry can be backed up and restored simply in a multitude of ways. Just because you don't know how doesn't mean it can't be done.
I've done some work with registry bits that are used to boot Windows. If you manage to bork them, you can't recover because you can't edit a dead registry from the Recovery Console -- I even had a second copy of Windows on the same machine just so that I could recover. No go. I did back up the registry entire and then overwrite it, but in terms of backing up and doing recover, the registry really sucks compared to an equivalent tree in the filesystem. It also means that all the tools you have to deal with the filesystem can't do anything with the registry...
A lot better than text files. Faster and more organized.
It is nice to have a single unified utility that's capable of presenting all your settings hierarchically.
That being said, I really don't consider it worth the pain the registry inflicts. You can have hierarchies with text files, making them just as organized. Also, while it's probably faster, that also means that programs constantly check the registry (and some annoying ones poll it). Most UNIX programs parse and store the contents of their dotfiles at startup.
When you can prove that that unorganized shithole known as
Obviously you've never taken a single reasonably difficult programming class in your life.
Yes, I've noticed how UNIX folks tend to be less programmers than Windows folks. Uh, huh.
I understand that the GIMP has a decidedly different UI from what Windows users are accustomed to...heck, different from GNOME.
However, I simply cannot agree on the MDI modification.
First, GIMP's current UI is very good for multiple viewports, where you can spread it out across multiple desktops. MDI would take that away. Even on a single viewport you can put some palletes in the back if you don't need them.
Second, one common complaint about GIMP is its complete and utter lack of modality. There are no dialog boxes that come up and prevent you from doing things. In the middle of setting some plugin settings? Just flip to another window and do some other work. This can get confusing to people that are used to Photoshop -- but I'm quite certain that while this approach is unfamiliar, it's much better. You're never locked in to doing a particular step.
Finally, it would be nice to have "palette" style windows, but unfortunately X11 doesn't support a palette (or in Mac OS UI terms, a "windoid") window style. It would be incredibly nifty if it did...
In the same vein, isn't Limewire open source?
Text-based interactive fiction contains some of the most amazing games ever made, and most are free.
There are several different IF environments -- TADS and Inform are the most popular, playable by TADS and Frotz, respectively.
There are many incredible games for both, but two of my favorites are Babel and Toonesia. This type of game loses most of its value if you cheat -- most of the value of the game is in gameplay.
Give it a shot, and rack your brains...and don't get eaten by a grue.
Hey, I thought you only trolled! What's with the good suggestions?
PuTTY is a work of art. It's free, it's flexible, it's self-contained in one executable (no bloody installers!)...it's the single Windows program that I must have if I'm using a Windows machine.
It can even emulate X11's middle-click-to-paste and select-to-copy mode. Absolutely wonderful.
What the hell is wrong with you? Aqua has no technical value whatsoever. It's not useful from an Open Source standpoint. It's a blinking *theme*.
It's also how Apple designed a unique looking product. They're handing you the important code, so quite complaining about some skin-level deep issues.
But then they would have had to test two totally different KDEs and support them. That's non-trivial, and even more confusing to users trying to figure out what's being installed. You'd have KDE+GNOME style KDE, *and* vanilla KDE (which would have to be packaged separately and eat more space, as well, and would have to be marked as incompatible with KDE+GNOME style KDE).
If someone really wants KDE, they can get it straight from kde.org -- I get a fair amount of software from the original source. The KDE project can distribute their blessed distribution however they want to do so. If you *really* want vanilla KDE with no GNOME integration done at all, then there are distros like Mandrake that *have* taken this approach...but I like seeing the diversity among distros that characterizes Linux. If someone wants to make a distro that has a totally terminal-based UI through an X server and AAlib and runs Enlightenment...well, they can do so. The users will end up voting by using whichever they prefer -- it could be that Red Hat is doing the wrong thing or the right thing. People will try both, and comment on them. RH will probably polish things up in 8.1, and if people still don't like the approach...well, Mandrake will get them.
Monty, love your work. Thanks, and congrats on the 1.0 release of Vorbis!
But hey, I loved that TNG episode where Data gets his dick caught in the food synthesizer, so that goes to show Vorbis is Bad.
This may be making a serious run for being my new signature.
He didn't have any *true* points, much less salient ones.
by Professor Collins
...in my experimentations.
...concentrates its encoding in the more audible midrange...
...because the brain unconsciously uses inaudible overtones as a guide to determinine the tone of the music...
Yeah. People tend to give people with a "Professor" in front of their name more credence. Good nick choice.
I've been studying psychoacoustics in my spare time
Heh
Heh.
Yup. Be a pretty pathetic audio codec if it didn't.
Given 128kbps, I should expend more data on the inaudible portion of the spectrum to hint people as to what the audible tone is that I'm already emphasizing in my encoding? Uh, huh.
The way that Vorbis compresses its audio accelerates speaker degredation
Tech can now destroy speakers! Muahahaha!
Actually, if I remember correctly, Red Hat *does* have KDE as an install option currently (as opposed to GNOME). The KDE guys didn't like it because GNOME was the default checked option in the installer.
Second, and more importantly, they have replaced KDE apps with equivalent apps, either from GNOME or independent projects. For example, they replaced konqueror with Mozilla, Koffice with OpenOffice, KMail with Evolution.
Yes. So? The alternative would be using all KDE apps in their place. So the KDE people are pissed off because RH isn't excising all competitors to their software from their distro and using only KDE? That's just silly.
Very true. Red Hat is doing something that consumers have wanted (and Linux tweakers have been doing for some time with themes) -- making their KDE and GNOME apps look similar.
KDE people whining about this are going to be ignored. The GNOME people have accepted the loss of their icons without throwing fits, though it certainly changes RH GNOME's "look and feel". I can't figure out why the KDE people can't do the same for the changes that affected their "look and feel."
Right up until the release announcement, news.google.com was live, had a *great* layout consisting of really simple, single column text with no images or tables, and was even easier to navigate. I loved it.
:-(
Then Google changed things to look like a traditional news site.
That extra RAM will get used for caching.
The Mhz != Mhz argument is valid, but Apple's stretched it waaay beyond reality. The PPC chips *do* run cooler and *do* have a more sane instruction set, and for a lot of people that is valuable, but as for raw computational power? Nope...x86 chips are way out in front.
Oh, a better quote from the article:
Dammit, Apple, if you're moving to x86 you simply aren't going to be a competitive systems manufacturer any more. You're going to have to accept that and either sell add-ons or only sell software.
"Yes, so we've accepted the nuclear weapons development contract. We need to ship in a month, so we need the testing cluster up tomorrow. You've got tonight to purchase and set up a 300 node cluster."