You know what's ironic - I went to read that article and the first thing I looked for was the "Print" view so that I could read it without all the crap around it. So print isn't dead!
The operating system needs to be able to control caching behaviour of areas of memory.
Consider if you map the registers of a device into some memory range, you don't want writes there to be cached, e.g. you write to the memory mapped to the device register that syncs your disk or sends a network packet or something and the underlying device doesn't actually see it until the CPU gets around to flushing the cache some random time later.
Userspace programmers should not (and can not) modify this state. The exploit requires root, and from there it's just handy that Linux lets you access these settings via/proc, rather than having to go to the trouble of writing and loading your own kernel module, etc.
"Presumably this will include sources such as YouTube, Hulu, Comedy Central, MSNBC, etc."
I'm pretty sure this would be a licensing issue, not at technical issue about streaming formats.
It's one thing to have to sit at your computer, or faff about with a media PC to watch content. But I imagine the networks would be very scared of a simple, cheap, no subscription, plug-in box. I'd be (pleasantly) surprised if they let Roku get to that content.
Once Hulu comes to my Tivo (I hate having a million boxes for different things), I'd reconsider why I even pay for even the cheapest tier cable.
If anyone wonders what the Comcast 250G caps are about, they have nothing to do with bandwidth contention and everything to do with them realising soon they will loose half (more?) of their business when cable TV dies.
This is not true. Standard ESX runs Linux for about 2 seconds; it carves out 256MiB for itself, boots and the first thing that happens in the ESX initrd is the vmkernel is loaded, which takes over.
From then on, the Redhat OS running is more like a special VM (has some hooks into the vmkernel, etc) which you use to manage the system.
I used to work help-desk, and late at night there would only be two people in the quite large building - me and one of the operators. Anyone who as worked with "ops" knows they generally turn a bit strange due to them working nights with nobody around and only DAT tapes for company.
So anyway, there is this big fire alarm panel with tons of buttons that we never really thought about, until one night when it started beeping constantly. The ops guy found a key to it, and then we both stood there looking at the probably 60 buttons and flashing lights, etc. Personally, I would have chosen one of the black buttons marked "mute", but the ops guy went straight for the biggest red one on the board.
The result was more beeping, lots of red lights and about 5 fire-engines.
With all this fantastic clock technology, where can I get an alarm clock that has technology that wasn't cutting edge in 1969?
I'd like
Ability to set different alarms for Monday-Friday and Sat-Sun
Multiple alarms, so I can get up early and my parter can sleep in until the second alarm for her goes off
Digital tuning (AM/FM) and volume control
Ability to match a station/volume to a function: i.e. go to sleep with quiet AM radio and wake up to loud FM radio
Clock radios haven't changed at all since I first got one when I was about 5! Someone out there must be able to package up a glorified palm pilot with some big buttons and red led's and make a killing. These days you could put 802.11 in it and get weather/traffic reports on a led ticker... I'm sure there is a market!
Good C code only gets you so far, and isn't the point of Ulrich's rant.
In the areas Ulrich is active in, like glibc, the kernel, binutils and gcc, the quality of the code really isn't the problem. Each architecture does things so differently there isn't one solution that can fit all. If you change one thing, you're sure to break another. You abstract things into generic and architecture code, but whenever you change that generic code it generally breaks all the architectures. His point is that you decide what you support (i.e. check doesn't break when you make a change), and anyone else has to keep up. If they don't, too bad!
But even for userspace apps, writing anything more than a trivial application also really becomes hard when you want to start making it perform. For example, for a high performance web-server you want to use epoll on linux, kqueue on BSD, poll on Solaris, etc etc. There are of course nice ways to abstract all that stuff, but it all takes developer time and maintaince effort. Is it worth it?
What are you talking about? You can run any x86 application just fine on any Itanium box. In fact I am writing this on a browser running in a x86 chroot environment on an IA64 box so I can use flash and other similar binary crap.
Especially for jazz albums, where the personel and circumstances surrounding the album help understand the music. Others have written on the importance of keeping this part of musical history.
Not to mention the cover art which is often a casulty of being squashed into 200x200 pixels, if you're lucky.
Reminds me of that often quoted story about Thomas Watson, head of IBM, when some executive made a bad decision that ended up costing $10 million. The guy comes in and says "I suppose you'll want my resignation now" and Watson replies something like "Are you crazy! I just spent $10 million educating you!"
-X or --no-init
Disables sending the termcap initialization and deinitialization strings to the terminal. This is sometimes desirable if the deinitialization string does something unnecessary, like clearing the screen.
Only this community of emacs users at GUI-less workstations users would think it better to not use CSS.
No, this is exactly who wants people to use CSS. If people separated content from layout, imagine how much of a better place the world would be. Personally, this is why I have gravitated to DocBook so much recently. On the extreme end, you have Microsoft Word, where layout and presentation is completely embeded with your actual information. LaTeX is better, but you still end up with \parskips and things around. DocBook may be verbose, but you have all your content separate and then mark it up the way you like for presentation in a completely separate style sheet.
That's why you put your drivers in userspace! The only disadvantage of doing this is that you are making it exponentially eaiser for distributers to ship binary drivers, but the advantages you get include isolation, language independance (imagine an IDE driver in python), and simple forward compatability.
See this lwn artice for one approach that is currently working quite well (self interest note; I'm involved in the project).
My boss gave a paper at linux.conf.au this year, and you can see more information on our WiKi. One of the guys here has even written an IDE driver in Python.
We also want to push microstate accounting, a way to get really fine grained statistics on what your kernel is doing (i.e. how long is spent waiting on a futex, etc?).
The emusic download manager has clients for windows/mac/linux and they all seem to at least minimally work.
I've found emusic great for older jazz artists, finding albums I wouldn't have otherwise been able to get my hands on. It almost seems a bit cheap in fact, but then I remember they have essentially zero costs once they've ripped the CD.
One thing I would really really like is access to the cover artwork and linear notes... none of the services seem to provide this.
Then he wizened up... he went twenty over the speed limit still, he just parked at a rest stop before exiting, or so she claims.
That's a pretty funny definition of 'wizening up'... he took all the extra risks of extra speed but received none of the benefits (i.e. getting there earlier).
Todd Proebsting has a "law" like Moore's law... from his webpage
Moore's Law" asserts that advances in hardware double computing power every 18 months.
I (immodestly) assert "Proebsting's Law", which says that advances in compiler optimizations double computing power every 18 years.
There is more information on his webpage.
It's not strictly true -- compilers have moving targets (different architectures and hardware optimisations over time, greater complexity in languages, etc etc), and for important things like scientific applications compilers can really optimise code; but in general R&D towards compiler development seems to be sorely lacking compared to other areas.
I work on IA64, which is fundamentally designed to take advantage of smarter compilers. While there are some interesting projects (ORC for example) nothing is really close to ready to take over gcc/intel compiler. We really want something that compiles once, can profile code execution and then recompile to optimise based on the execution profile; something along the lines of what this article talks about but built right into the compiler.
For those of you who skipped structural engineering lots of information on damping systems is available here especially the two "technotes" papers 1 and 2.
I think the article said it was an 800 ton damping system; it doesn't suggest how much the building will weigh but that second paper shows that even a 2% mass ratio can signficiantly reduce vibrations in the building.
When showing some people around your very impressive computer room say: see this! It's a hot plug RAID array for one of our production file servers with a couple of hundred gigabytes of storage. I could just pull any one of these drives out right now and no one would even notice. In fact, let me demonstrate...
Unfortunatley it wasn't as redundant as he expected:(
Re:Map of UPS battery exhaustions
on
Network Blackout
·
· Score: 1
On the topic of emergency generator stories. I used to work at a helpdesk, and arrived very early one Saturday morning to about 5 fire trucks and the HAZMAT (hazardous materials) team. Turns out a blackout the night before had triggered the generator, which had split a fuel line. Of course the fuel was above the split so pushed out a couple of thousand litres of diesel into the surrounding environment before the operator 3 floors up said "what's that smell"...
Slashdot Mirror
You know what's ironic - I went to read that article and the first thing I looked for was the "Print" view so that I could read it without all the crap around it. So print isn't dead!
To beat a bunch of UNIX engineers? I guess he got in at about 10.30am
The operating system needs to be able to control caching behaviour of areas of memory.
Consider if you map the registers of a device into some memory range, you don't want writes there to be cached, e.g. you write to the memory mapped to the device register that syncs your disk or sends a network packet or something and the underlying device doesn't actually see it until the CPU gets around to flushing the cache some random time later.
Userspace programmers should not (and can not) modify this state. The exploit requires root, and from there it's just handy that Linux lets you access these settings via /proc, rather than having to go to the trouble of writing and loading your own kernel module, etc.
The article says
I'm pretty sure this would be a licensing issue, not at technical issue about streaming formats.
It's one thing to have to sit at your computer, or faff about with a media PC to watch content. But I imagine the networks would be very scared of a simple, cheap, no subscription, plug-in box. I'd be (pleasantly) surprised if they let Roku get to that content.
Once Hulu comes to my Tivo (I hate having a million boxes for different things), I'd reconsider why I even pay for even the cheapest tier cable.
If anyone wonders what the Comcast 250G caps are about, they have nothing to do with bandwidth contention and everything to do with them realising soon they will loose half (more?) of their business when cable TV dies.
Standard ESX runs on linux.
This is not true. Standard ESX runs Linux for about 2 seconds; it carves out 256MiB for itself, boots and the first thing that happens in the ESX initrd is the vmkernel is loaded, which takes over.
From then on, the Redhat OS running is more like a special VM (has some hooks into the vmkernel, etc) which you use to manage the system.
Just the usual "colour of the bikeshed" problem which hits every collaborative project ever built.
I used to work help-desk, and late at night there would only be two people in the quite large building - me and one of the operators. Anyone who as worked with "ops" knows they generally turn a bit strange due to them working nights with nobody around and only DAT tapes for company.
So anyway, there is this big fire alarm panel with tons of buttons that we never really thought about, until one night when it started beeping constantly. The ops guy found a key to it, and then we both stood there looking at the probably 60 buttons and flashing lights, etc. Personally, I would have chosen one of the black buttons marked "mute", but the ops guy went straight for the biggest red one on the board.
The result was more beeping, lots of red lights and about 5 fire-engines.
Do you have any idea what was used to make those presentations?
I'd like
Clock radios haven't changed at all since I first got one when I was about 5! Someone out there must be able to package up a glorified palm pilot with some big buttons and red led's and make a killing. These days you could put 802.11 in it and get weather/traffic reports on a led ticker
Good C code only gets you so far, and isn't the point of Ulrich's rant.
In the areas Ulrich is active in, like glibc, the kernel, binutils and gcc, the quality of the code really isn't the problem. Each architecture does things so differently there isn't one solution that can fit all. If you change one thing, you're sure to break another. You abstract things into generic and architecture code, but whenever you change that generic code it generally breaks all the architectures. His point is that you decide what you support (i.e. check doesn't break when you make a change), and anyone else has to keep up. If they don't, too bad!
But even for userspace apps, writing anything more than a trivial application also really becomes hard when you want to start making it perform. For example, for a high performance web-server you want to use epoll on linux, kqueue on BSD, poll on Solaris, etc etc. There are of course nice ways to abstract all that stuff, but it all takes developer time and maintaince effort. Is it worth it?
Nope, Debian, where the minimum one needs to do to run their x86 applications on an Itanium box is apt-get install ia32-libs.
What are you talking about? You can run any x86 application just fine on any Itanium box. In fact I am writing this on a browser running in a x86 chroot environment on an IA64 box so I can use flash and other similar binary crap.
Liner notes please!
Especially for jazz albums, where the personel and circumstances surrounding the album help understand the music. Others have written on the importance of keeping this part of musical history.
Not to mention the cover art which is often a casulty of being squashed into 200x200 pixels, if you're lucky.
Reminds me of that often quoted story about Thomas Watson, head of IBM, when some executive made a bad decision that ended up costing $10 million. The guy comes in and says "I suppose you'll want my resignation now" and Watson replies something like "Are you crazy! I just spent $10 million educating you!"
you can even set it with LESS=X ; export LESS
units knows all
ianw@mingus:~$ units
2084 units, 71 prefixes, 32 nonlinear units
You have: 40mph
You want: knot
* 34.75905
/ 0.028769486
So about 35 knots
See this lwn artice for one approach that is currently working quite well (self interest note; I'm involved in the project).
Exactly what we are working on.
My boss gave a paper at linux.conf.au this year, and you can see more information on our WiKi. One of the guys here has even written an IDE driver in Python.
We also want to push microstate accounting, a way to get really fine grained statistics on what your kernel is doing (i.e. how long is spent waiting on a futex, etc?).
The emusic download manager has clients for windows/mac/linux and they all seem to at least minimally work.
... none of the services seem to provide this.
I've found emusic great for older jazz artists, finding albums I wouldn't have otherwise been able to get my hands on. It almost seems a bit cheap in fact, but then I remember they have essentially zero costs once they've ripped the CD.
One thing I would really really like is access to the cover artwork and linear notes
That's a pretty funny definition of 'wizening up'
There is more information on his webpage. It's not strictly true -- compilers have moving targets (different architectures and hardware optimisations over time, greater complexity in languages, etc etc), and for important things like scientific applications compilers can really optimise code; but in general R&D towards compiler development seems to be sorely lacking compared to other areas.
I work on IA64, which is fundamentally designed to take advantage of smarter compilers. While there are some interesting projects (ORC for example) nothing is really close to ready to take over gcc/intel compiler. We really want something that compiles once, can profile code execution and then recompile to optimise based on the execution profile; something along the lines of what this article talks about but built right into the compiler.
For those of you who skipped structural engineering lots of information on damping systems is available here especially the two "technotes" papers 1 and 2.
I think the article said it was an 800 ton damping system; it doesn't suggest how much the building will weigh but that second paper shows that even a 2% mass ratio can signficiantly reduce vibrations in the building.
When showing some people around your very impressive computer room say: see this! It's a hot plug RAID array for one of our production file servers with a couple of hundred gigabytes of storage. I could just pull any one of these drives out right now and no one would even notice. In fact, let me demonstrate ...
:(
Unfortunatley it wasn't as redundant as he expected
On the topic of emergency generator stories. I used to work at a helpdesk, and arrived very early one Saturday morning to about 5 fire trucks and the HAZMAT (hazardous materials) team. Turns out a blackout the night before had triggered the generator, which had split a fuel line. Of course the fuel was above the split so pushed out a couple of thousand litres of diesel into the surrounding environment before the operator 3 floors up said "what's that smell" ...