Yeah but deskpros are a real pain to use as headless servers since they won't boot without a keyboard attached. Some BIOS revisions have the super-secret handshake to disable it, if you can remember what it is. They also want to reboot twice in a row for things like changing the memory size or hard disk!
Our company got a shipment of 18 or so GX270s about 2 years ago. In the first year of operation over half of their power supply units failed and had to be replaced.
Google suggests that evdev is a Linux specific thing. Sounds kind of like what sysmouse does on FreeBSD -- all the mouse drivers use a common protocol so X.org only has to understand one protocol.
What X11 really needs, IMO, is real wheel/multiaxis support in the X protocol rather than the hack of mapping to buttons 4 and 5. Also, support for more mouse buttons would be nice.
Yeah, it seems like all the worst legislation as far back as I can remember has been introduced via riders like this. I'd love to see the practise done away with, but it'd take a massive petition drive to pull it off...
Or a president with enough backbone to veto anything (and I mean ANYTHING) with extra crap attached to it. Yeah, they could still use the 2/3 override for anything that's really well supported, but it would at least send a message to Congress to cut it out.
Of course since everyone in DC is a shill for their policial party that will never happen.
Yeah, but despite being incredibly common, the general consensus is that Silicon Image SATA controllers, to put it simply, suck. We're talking hardware bugs here.
Probably the reason they're integrated into so many motherboards is because they're so cheap.
Yes, I know, I tried using the AIM (and MSN, and Yahoo) transports for Jabber on my server, but they just don't work as well as a native client.
For one, the server has to keep your unencrypted password saved. For me that wasn't a problem since it was my server, but for other users they have to trust some unknown third party.
Also, things like buddy icons don't work (which is another rant I have about Jabber and incompatible proposals for that, but anyway). Little things like typing notification and profile information don't work quite right.
Some Jabber clients also don't work correctly with transports. Gaim for one only sort of worked. In the end I'd rather use Oscar (or TOC when I'm using naim, but I wish naim had oscar support since TOC is horrible and often broken...).
I'm hoping that any XMPP to AIM google put in place would work better than transports. For one thing it should at least prevent you from having to have an AIM account to proxy through.
Though, I'm not too keen on Google's federation
I agree with you there, hence the whole "big if" comment. They'd have to open up site-to-site for everyone, because I doubt my little server with a few friends on it would be eligibale for their "federation". That's the main reason I've avoided google talk altogether so far.
ISTM if there's a spam problem they could just selectively block certain users, or entire domains if whoever's operating the server is in on it. Unlike email I think Jabber is fairly spoof-resistent.
If Google Talk gets connected with the AIM network, and Google eventually allows Jabber server-to-server (big if, I know), I might possibly be able to talk to my friends on AIM without having to use Oscar...
Wait, so instead of asking the user to provide his/her identity, they want to embed a chip into the hardware, and trust the machine rather than the individual user?
And here I thought trust based on the identity of the computer was a bad idea, and that UNIX had learned this long ago with the shift away from rlogin/rhosts based authentication.
Why? So that we could have skill based enforcement/graduated limits.
I'm sorry but that's a horrible idea. Different vehicles going different speeds on the same road just creates chaos and makes it more dangerous. We used to have truck speed limits in Texas but they have all but disappeared in the last few years.
I visited Houston in 2002 when they had reduced the limit to 55 on all highways for "environmental" reasons. Everyone ignored the new limits and continued driving 70-80. Attempting to drive 55, in ANY lane, was extremely dangerous. Needless to say, the 55 limit didn't last very long before it was repealed.
e.g. if you go into othe city in rush hour, you pay more than if you go in at 3 am, or if you go on a trip on a rural backroad.
Geez, talk about adding insult to injury. Not only do you have to put up with getting stuck in traffic, but you have to pay more for the pleasure of it?
A lot of functional languages can (theoretically) parallelize automatically as well, especially "pure" functional languages that prohibit or restrict side effects. IIRC Haskell has done some work in this area.
Wow, I think somebody was pulling someone's leg there. I'm not aware of any 3-CPU motherboards ever being manufactured, so they were probably quad boards with the fourth socket empty (or terminator card, whatever). I'm actually kind of surprised that a BIOS of that era wouldn't reject that as an invalid configuration.
What's even funnier, is that for NT 3.51 and 4.0 at least, the MP HAL was used by the setup program when you first booted off the CD/floppies (you could tell because of the MultiProcessor Kernel line), even for single-CPU machines. So it definitely wouldn't fail to boot with only 1 left.
Only other thing I can think of is if they were REALLY weird machines. I have vague memories of heading about a few custom HALs being written for certain manufacturers, but I don't remember who.
Um, what? The Multi-Processor HAL works fine with single CPU machines, so long as the motherboard has an I/O APIC and MP Table and all that good stuff.
I used to install dual-capable servers that only had a single CPU in them with the MP HAL all the time. Either F5 or F6 at the very beginning of the text mode setup lets you override the auto-detected HAL. That way I could add a second CPU without having to reinstall.
There were also various tricks for changing the HAL without having to do a complete reinstall, but that's another discussion.
Just be careful to click End Process rather than End Task...
"End Task" will send the app a request to terminate, which for annoying installers that give you only an "OK" at the end, will often cause them to reboot the system anyway.
"End Process" doesn't notify it first, just kill -9s it, so it won't be able to pull any crap like that.
Or just don't do anything when one is inserted. Click on the damn CD icon in "My computer" if you want to run it -- popping crap up just annoys the user.
Yeah, but you own your computer. You're not renting it from the music/movie/software publishers.
If you rent a washer/dryer, that doesn't give Maytag the right to enter your home and inspect it (or inspect your other appliances -- the Sony rootkit inserts itself between the driver and the OS so it can see ALL discs, not just Sony's).
Media companies are even worse off because you buy something from them ONCE. It's not a rental agreement.
Slashdot Mirror
Each of those hours passively glued to the TV is an hour of your life. Are you sure you want to spend it there?
That's 44 minutes, thank you very much.
Yeah but deskpros are a real pain to use as headless servers since they won't boot without a keyboard attached. Some BIOS revisions have the super-secret handshake to disable it, if you can remember what it is. They also want to reboot twice in a row for things like changing the memory size or hard disk!
Might not work either -- that early in the setup it would surprise me if enough drivers were loaded for USB to work at all.
Our company got a shipment of 18 or so GX270s about 2 years ago. In the first year of operation over half of their power supply units failed and had to be replaced.
Can't speak for the others, but while FreeBSD does have a procfs, it's considered deprecated and its use is discouraged.
The cat is obviously on fire.
You'll want to check out the new evdev driver
Google suggests that evdev is a Linux specific thing. Sounds kind of like what sysmouse does on FreeBSD -- all the mouse drivers use a common protocol so X.org only has to understand one protocol.
What X11 really needs, IMO, is real wheel/multiaxis support in the X protocol rather than the hack of mapping to buttons 4 and 5. Also, support for more mouse buttons would be nice.
BSD make has had this for ages, see mkdep(1).
The Nvidia implementation of xrandr appears to run rotated screens with full acceleration on low end cards.
Full XRender support maybe, but it kills hardware support for Xv, so mplayer chomps up 10x more CPU cycles.
Yeah, it seems like all the worst legislation as far back as I can remember has been introduced via riders like this. I'd love to see the practise done away with, but it'd take a massive petition drive to pull it off...
Or a president with enough backbone to veto anything (and I mean ANYTHING) with extra crap attached to it. Yeah, they could still use the 2/3 override for anything that's really well supported, but it would at least send a message to Congress to cut it out.
Of course since everyone in DC is a shill for their policial party that will never happen.
Yeah, but despite being incredibly common, the general consensus is that Silicon Image SATA controllers, to put it simply, suck. We're talking hardware bugs here.
Probably the reason they're integrated into so many motherboards is because they're so cheap.
Yes, I know, I tried using the AIM (and MSN, and Yahoo) transports for Jabber on my server, but they just don't work as well as a native client.
For one, the server has to keep your unencrypted password saved. For me that wasn't a problem since it was my server, but for other users they have to trust some unknown third party.
Also, things like buddy icons don't work (which is another rant I have about Jabber and incompatible proposals for that, but anyway). Little things like typing notification and profile information don't work quite right.
Some Jabber clients also don't work correctly with transports. Gaim for one only sort of worked. In the end I'd rather use Oscar (or TOC when I'm using naim, but I wish naim had oscar support since TOC is horrible and often broken...).
I'm hoping that any XMPP to AIM google put in place would work better than transports. For one thing it should at least prevent you from having to have an AIM account to proxy through.
Though, I'm not too keen on Google's federation
I agree with you there, hence the whole "big if" comment. They'd have to open up site-to-site for everyone, because I doubt my little server with a few friends on it would be eligibale for their "federation". That's the main reason I've avoided google talk altogether so far.
ISTM if there's a spam problem they could just selectively block certain users, or entire domains if whoever's operating the server is in on it. Unlike email I think Jabber is fairly spoof-resistent.
If Google Talk gets connected with the AIM network, and Google eventually allows Jabber server-to-server (big if, I know), I might possibly be able to talk to my friends on AIM without having to use Oscar...
(wipes coffee off monitor)
Mod parent up.
Wait, so instead of asking the user to provide his/her identity, they want to embed a chip into the hardware, and trust the machine rather than the individual user?
And here I thought trust based on the identity of the computer was a bad idea, and that UNIX had learned this long ago with the shift away from rlogin/rhosts based authentication.
Don't bother. You'll pass enough scanners that will pick up your government-mandated RFID "health" implant anyway.
Why? So that we could have skill based enforcement/graduated limits.
I'm sorry but that's a horrible idea. Different vehicles going different speeds on the same road just creates chaos and makes it more dangerous. We used to have truck speed limits in Texas but they have all but disappeared in the last few years.
I visited Houston in 2002 when they had reduced the limit to 55 on all highways for "environmental" reasons. Everyone ignored the new limits and continued driving 70-80. Attempting to drive 55, in ANY lane, was extremely dangerous. Needless to say, the 55 limit didn't last very long before it was repealed.
e.g. if you go into othe city in rush hour, you pay more than if you go in at 3 am, or if you go on a trip on a rural backroad.
Geez, talk about adding insult to injury. Not only do you have to put up with getting stuck in traffic, but you have to pay more for the pleasure of it?
A lot of functional languages can (theoretically) parallelize automatically as well, especially "pure" functional languages that prohibit or restrict side effects. IIRC Haskell has done some work in this area.
Wow, I think somebody was pulling someone's leg there. I'm not aware of any 3-CPU motherboards ever being manufactured, so they were probably quad boards with the fourth socket empty (or terminator card, whatever). I'm actually kind of surprised that a BIOS of that era wouldn't reject that as an invalid configuration.
What's even funnier, is that for NT 3.51 and 4.0 at least, the MP HAL was used by the setup program when you first booted off the CD/floppies (you could tell because of the MultiProcessor Kernel line), even for single-CPU machines. So it definitely wouldn't fail to boot with only 1 left.
Only other thing I can think of is if they were REALLY weird machines. I have vague memories of heading about a few custom HALs being written for certain manufacturers, but I don't remember who.
Um, what? The Multi-Processor HAL works fine with single CPU machines, so long as the motherboard has an I/O APIC and MP Table and all that good stuff.
I used to install dual-capable servers that only had a single CPU in them with the MP HAL all the time. Either F5 or F6 at the very beginning of the text mode setup lets you override the auto-detected HAL. That way I could add a second CPU without having to reinstall.
There were also various tricks for changing the HAL without having to do a complete reinstall, but that's another discussion.
Two-choice paper-rock-scissors is so boring.
Just be careful to click End Process rather than End Task...
"End Task" will send the app a request to terminate, which for annoying installers that give you only an "OK" at the end, will often cause them to reboot the system anyway.
"End Process" doesn't notify it first, just kill -9s it, so it won't be able to pull any crap like that.
Or just don't do anything when one is inserted. Click on the damn CD icon in "My computer" if you want to run it -- popping crap up just annoys the user.
Yeah, but you own your computer. You're not renting it from the music/movie/software publishers.
If you rent a washer/dryer, that doesn't give Maytag the right to enter your home and inspect it (or inspect your other appliances -- the Sony rootkit inserts itself between the driver and the OS so it can see ALL discs, not just Sony's).
Media companies are even worse off because you buy something from them ONCE. It's not a rental agreement.