that isn't really a concern -- with a current glibc and such, the upgrade to 2.4.0-testX has been painless for me on my redhat 6.2 system. I would expect that 7.0 is ready for 2.4 so that when a stable version is released, you'll be able to just download the source or (ick) an rpm and install it no sweat.
Don't forget that the voodoo3 was limited to 256x256 textures while the tnt2 supported 2048x2048 pixel textures. While this didn't come into play nearly as much as the 16 bit color and 16 megs of memory, it was still a stupid move by 3dfx.
I was always entertained by 3dfx's late notion of including support for 22.5 bit colour, as a halfway point between 16 and 32. I don't remember how they were going to get half a bit, so if someone else knows the details, I'd be most interested.
Not too long ago, to legally use office, you had to buy a license for every machine in your office/school/home/whatever, regardless of whether it would be, or even was able to run office. For instance, a school's computer lab, with 20 dell pc's and 20 macs, and 10 linux boxes (just to make sure you know this is hypothetical:) would have to pay for 50 licenses from microsoft to legally use office.
Happily, microsoft responded to their angry customers and changed the license agreement a bit. if enough people are pissed off about this, they might respond. or maybe not.
you're very correct -- having some 3d acceleration in your video is much better than having none, but the processor still plays a very important role. that voodoo2 can't do any geometry work, for instance. relative sizes of everything due to perspective is still done by your cpu. or, in quake3 especially, bot AI can drag down a quick machine and the video card won't help much there.
i just moved from a tnt2 ultra to a geforce2 paired up with that celeron 567, and didn't notice nearly the difference i'd hoped. right now, it's still quick, but i'm sure it'll be noticeably faster with a thunderbird 800 or so powering the parts of the game that aren't helped by the video card.
so while you're very right that a ppro 200 is going to get more help from the addition of a 3d accelerator than overclocking, the same doesn't hold true for high end gaming systems.
most overclockers do it for gaming, where the cpu is a large bottleneck. it's also a matter of saving money. approximately one year ago, i bought a celeron 366, and overclocked it to 567 (still running like that now without problems). at the time i bought it, the absolute fastest chip on the market was a p3 550. i used a $100 chip to match the performance of a $600 chip. i'd say that's too much of a price/performance difference to overlook.
similarly, right now, you can buy a 600 MHz celeron/duron/thunderbird and often get it over 900 MHz. for a college student, who abuses the fast internet connection to play first person shooters, overclocking becomes a very tempting proposition.
the official ssh packages at ssh.com are free for non-commercial use. we have a site license here at u of i, and i don't think we payed anything for it.
that's not true. as of registration for the summer semesters, all students were required to use a kerberos client to connect to the registration server.
students also have a separate password for that than the rest of their accounts, an effort to keep it more secure than other systems. if someone drops all your classes for you here, you're likely fucked.
actually, part of the reason that ssh wasn't required to connect to students.uiuc.edu to register was that at the peak of registration, when a few hundred students would be using each machine in the cluster, the encryption overhead would become pretty nasty. the suns they use do a lot, and the extra work for hundreds of ssh connection isn't something they need.
i believe this scenario actually happened a while back when housing installed ssh as the default telnet client in the dorm labs.
I may be biased, but I work on housing network stuff at the University of Illinois (UC) and I don't think this is an issue. Our campus-wide network is comprised mostly of switches, making packet sniffing tough. And the dorm networks, which are likely the most dangerous place to have people sniffing, were set up with hubs that scramble data for anyone besides the recipient of that packet (that was the beginning of switching technology, 8 years ago). They're being replaced with full-fledged switches as i type this.
That being said, I would hope that most other campuses have taken similar precautions against packet sniffing when they designed their networks. There's nothing really radical here, mostly using switches instead of hubs.
On a well designed network, choice of protocol should matter a lot less.
speaking as an employee housing at the university of illinois, most schools do ban wireless networking because it allows for ridiculously easy packet sniffing. wireless ethernet traffic is unencrypted as it is now (we're awaiting an new ieee spec to fix that), and until there's an encryption scheme built in, it will not be allowed for use in our buildings. i suspect other schools have similar policies.
the review docked linux because samba can only be configured by editing "a cryptic text file." in truth, samba is supported by linuxconf, and it ships with SWAT, which IMO is a pretty nice web-based config tool.
Slashdot Mirror
on our local (uiuc) mirror, there's no read access to the new directory until everything's been uploaded (including iso's). just wait a few hours.
chris
that isn't really a concern -- with a current glibc and such, the upgrade to 2.4.0-testX has been painless for me on my redhat 6.2 system. I would expect that 7.0 is ready for 2.4 so that when a stable version is released, you'll be able to just download the source or (ick) an rpm and install it no sweat.
chris
Don't forget that the voodoo3 was limited to 256x256 textures while the tnt2 supported 2048x2048 pixel textures. While this didn't come into play nearly as much as the 16 bit color and 16 megs of memory, it was still a stupid move by 3dfx.
I was always entertained by 3dfx's late notion of including support for 22.5 bit colour, as a halfway point between 16 and 32. I don't remember how they were going to get half a bit, so if someone else knows the details, I'd be most interested.
It's even still on firehose:
:)
http://www.firehose.net/free/aimazing/
and i must say, it works very well
chris
Not too long ago, to legally use office, you had to buy a license for every machine in your office/school/home/whatever, regardless of whether it would be, or even was able to run office. For instance, a school's computer lab, with 20 dell pc's and 20 macs, and 10 linux boxes (just to make sure you know this is hypothetical :) would have to pay for 50 licenses from microsoft to legally use office.
Happily, microsoft responded to their angry customers and changed the license agreement a bit. if enough people are pissed off about this, they might respond. or maybe not.
chris
you're very correct -- having some 3d acceleration in your video is much better than having none, but the processor still plays a very important role. that voodoo2 can't do any geometry work, for instance. relative sizes of everything due to perspective is still done by your cpu. or, in quake3 especially, bot AI can drag down a quick machine and the video card won't help much there.
i just moved from a tnt2 ultra to a geforce2 paired up with that celeron 567, and didn't notice nearly the difference i'd hoped. right now, it's still quick, but i'm sure it'll be noticeably faster with a thunderbird 800 or so powering the parts of the game that aren't helped by the video card.
so while you're very right that a ppro 200 is going to get more help from the addition of a 3d accelerator than overclocking, the same doesn't hold true for high end gaming systems.
most overclockers do it for gaming, where the cpu is a large bottleneck. it's also a matter of saving money. approximately one year ago, i bought a celeron 366, and overclocked it to 567 (still running like that now without problems). at the time i bought it, the absolute fastest chip on the market was a p3 550. i used a $100 chip to match the performance of a $600 chip. i'd say that's too much of a price/performance difference to overlook.
similarly, right now, you can buy a 600 MHz celeron/duron/thunderbird and often get it over 900 MHz. for a college student, who abuses the fast internet connection to play first person shooters, overclocking becomes a very tempting proposition.
chris
ssh.com has a very nice gui ssh client, with graphical scp support. try it out sometime.
it's free, but it only supports ssh2.
chris
the official ssh packages at ssh.com are free for non-commercial use. we have a site license here at u of i, and i don't think we payed anything for it.
chris
that's not true. as of registration for the summer semesters, all students were required to use a kerberos client to connect to the registration server.
students also have a separate password for that than the rest of their accounts, an effort to keep it more secure than other systems. if someone drops all your classes for you here, you're likely fucked.
chris
actually, part of the reason that ssh wasn't required to connect to students.uiuc.edu to register was that at the peak of registration, when a few hundred students would be using each machine in the cluster, the encryption overhead would become pretty nasty. the suns they use do a lot, and the extra work for hundreds of ssh connection isn't something they need.
i believe this scenario actually happened a while back when housing installed ssh as the default telnet client in the dorm labs.
chris
if you're willing to use ssh.com's software, scp works rather well between their windows client and a unix ssh2 server.
that's the only one i know of, but it works well enough for me to replace ftp with it whenever i'm going over an unsecured network.
chris
I may be biased, but I work on housing network stuff at the University of Illinois (UC) and I don't think this is an issue. Our campus-wide network is comprised mostly of switches, making packet sniffing tough. And the dorm networks, which are likely the most dangerous place to have people sniffing, were set up with hubs that scramble data for anyone besides the recipient of that packet (that was the beginning of switching technology, 8 years ago). They're being replaced with full-fledged switches as i type this.
That being said, I would hope that most other campuses have taken similar precautions against packet sniffing when they designed their networks. There's nothing really radical here, mostly using switches instead of hubs.
On a well designed network, choice of protocol should matter a lot less.
chris
speaking as an employee housing at the university of illinois, most schools do ban wireless networking because it allows for ridiculously easy packet sniffing. wireless ethernet traffic is unencrypted as it is now (we're awaiting an new ieee spec to fix that), and until there's an encryption scheme built in, it will not be allowed for use in our buildings. i suspect other schools have similar policies.
the review docked linux because samba can only be configured by editing "a cryptic text file." in truth, samba is supported by linuxconf, and it ships with SWAT, which IMO is a pretty nice web-based config tool.
just picking nits..
chris
millenium != y2k