No X.org developer called anything a bloody mess. I used some choice terms (and I actually *edited* my post shortly after making it...), but I'm not an X.org developer. Phoronix introduced the 'bloody mess' term specifically, and they're not X.org developers either.
No, I'm telling them to use the existing architecture. To put it bluntly, to do it in X.org. I suspect the only reason it wasn't done this way in the first place is that X.org won't take a driver that depends on a binary blob.
Gwenole, thanks for the info about IEGD, but you're way off base about other stuff.
One, I know the drivers are not open source. That was a large part of what the rant was about.
Two, how can you not understand what my complaint is? It's not that there's no driver. And it's not that the driver is not open source, except that that is related to the real problem. My problem is very simply that the drivers don't fucking work. The post is simply a long explanation of the series of crappy decisions which caused that. I would like for the drivers to be open source and Intel has a public commitment to *make* open source drivers (that they're not keeping), but if they could get their shit together to provide a sane closed source driver that gave me functioning native resolution accelerated 2D, that would be absolutely freaking dandy for me (although it obviously wouldn't be satisfactory for Fedora). All I want is a driver that is available in a logical place, that installs sensibly, and that provides basic 2D functionality, on perfectly normal modern distributions (Ubuntu 8.10, Mandriva 2009, Fedora 10, OpenSUSE 11, etc). This should not be a problem.
Finally, I hadn't actually started working for Red Hat when I wrote that post, though it was incorrectly reported as being from 'Red Hat's Adam Williamson' in a couple of places. Regardless, I'm not in driver development, I'm not involved in talking directly to Intel. When I write about stuff like that on my *personal* blog - not on any official site for any distribution - I am writing as a user just like anyone else, quite simply saying that for a perfectly normal person who went out and bought one of these chipsets and tried to run Linux on it, the situation is massively crappy. I'm not going to let who I work for stop me from doing that, neither when it was Mandriva nor now it's Red Hat. If I want to take an issue up through official channels, I will do that, but if you see it on my blog, that is just me.
You seem to think there's no real problem with this situation, which is odd because you seem quite informed about at least parts of it. Everyone else I've spoken to with knowledge of it (including high-level X developers and folks directly involved with Intel on this issue) has quite clearly said that it is, basically, a giant clusterfuck. There doesn't appear to be any disagreement that it's a big mess.
Intel *didn't* make an open source driver. If you read my post, you'll note that there's three different closed-source components to the driver, without which significant features break.
Aside from that, what's needed for meaningful open source development is not "here's some code, have fun". There needs to be a proper development process hosted in an accessible fashion, and proper documentation. The 'intel' driver for previous Intel chipsets satisfies all of these goals. It's 100% open-source, it's developed within X.org and so easily accessible to external contributions in a widely-understood fashion, and the hardware is properly documented.
This 'psb' driver satisfies none of the goals. It was previously hosted within Moblin (which doesn't really have much of an external development community), and even that version of the code is now not being used. It now only shows up in obscure Ubuntu Netbook Remix repositories, with no independent source that anyone can find. So there's no sane development process to which external people can sensibly contribute. It contains large closed-source chunks. And there's no public hardware documentation, which makes it very hard for anyone else to work on it in the first place.
This is what I (and anyone else stuck with one of these chips) am complaining about.
That's the reputation of the power of the hardware. Yes, as far as playing your 3D games goes, you're not going to get very far with Intel.
However, up till now Intel has had a very good reputation for open source friendliness with regards to supporting the hardware, disregarding the actual power of the hardware. Intel are actively involved in maintaining the (100% open source) driver for all other Intel graphics chipsets, and they also contribute to general X.org development and the development of new technologies within X. Intel graphics hardware is generally the least powerful of the big three, but until this mess, it's been by far the best (and most openly) supported hardware in Linux.
My new desktop boots current Mandriva Cooker x86-64, from lilo to gdm, in 16 seconds. And a full working GNOME desktop with Firefox running takes about 25.
Of course, it's a heavily overclocked quad-core system with a pretty fast disk drive and 4GB of RAM.
Point being, random raw numbers (21.4 seconds!!!) don't mean an awful lot. Hardware is rather important. The original story's a good one - it explains the hardware used to test, and carefully compares ext3 and ext4, which is what they were actually trying to do. But Slashdot's abstraction of it is crap.
Well, I just built a new system and looked all this crap up the other day, so I feel qualified to comment.
Yes, the new chips finally just about match equally priced Intel chips - at stock.
The problem is they only go down to $250 or so. Not down to $200 or under, where the lowest-priced Core 2 Quad lives: the Q6600. And the thing about the Q6600 is, it can be trivially overclocked well past 3GHz on stock cooling with no enthusiast messing about. Just bump up the FSB speed and you're good. So for $200 or less you can get a year-old chip that will comfortably outperform AMD's latest and greatest. That's not such great news for AMD, is it?
And, don't forget, i7 is just Intel bleeding the 'enthusiast' market dry. Essentially they're getting an overpriced version of the new platform out first to capitalize on the market that'll spend any kind of money to have the fastest system on the block. Once they're done screwing those guys, they'll be making essentially the same chips available at much lower prices. Which will in turn drive down the price of the Core 2 generation, making them even more of a no-brainer purchase over the new AMD chips. AMD is basically a whole generation behind, and not looking like it'll catch up any time soon. I just don't see who - except someone who happens to have an existing AM2 motherboard that'll take one of these new chips and is looking for a modest upgrade - is going to buy one of these.
It obliquely credits my own article on the same topic - written a whole one day previously - while describing it as a 'review' of Barry. No, it isn't. It's a guide to the exact same topic covered in this article, only it's a hell of a lot more useful, because my article tells you exactly what packages to download for what distributions, and how to actually use them. Rather than just saying "oh, look, this application lets you do this, isn't it cool?"
Would've been a lot more useful to link to me. But then, my server probably can't take the load, so I should be thankful.:)
""If the persons were real, such depictions could never be permitted,"Justice Adams said in his judgement." "the mere fact that they were not realistic representations of human beings did not mean that they could not be considered people."
In other news, the director of 300 has been found guilty of mass murder, and sentenced to execution....and the world's gone mad.
"Now, I'm bad at understanding why one person can't have the same idea after someone else has had that same idea"
That's not what Harmonix did. They were perfectly well aware of GF/DM when making Guitar Hero. The debt has been explicitly acknowledged in interviews with top-level Harmonix people.
I saw a post on Engadget this morning proclaiming how great a new, specially tuned pre-beta of Windows 7 was for netbooks. After all, it only takes 450MB of RAM to boot to a desktop!
Frankie70: the highlights? hiring relatives of government ministers, and offering $400,000 in "marketing funds" to a reseller in order to persuade them to replace Mandriva ($3 per machine) with Windows ($45 per machine) on a large order of Intel Classmate PCs.
Later on, the article covers the Mandriva / Microsoft in Nigeria battle that was covered here before:
"TSC approached Mandriva SA, a French company that sells a Linux version. Believing Microsoft had offered its $3 package, Mandriva proposed a $3 price for a Linux operating system, plus about $2 for other software, say people familiar with the situation. In August 2007, TSC issued a purchase order for Mandriva Linux, and the laptop's Taiwanese manufacturer began loading it.
Microsoft continued to push Windows. It offered its XP and Office software for about $45 per machine, says Nyimbi Odero, then TSC's chief executive.
Mr. Odero says Microsoft wanted TSC to delete Linux from the initial shipments of Classmates. He says Microsoft proposed a way to "make it worth your while" through a joint-marketing agreement. According to a draft agreement Microsoft sent to TSC last Sept. 13, Microsoft would pay TSC to fund "certain marketing activities to encourage the sale and distribution" of Microsoft products. Mr. Odero says Microsoft made it clear that TSC wouldn't really be expected to market the products, but could keep the money as an incentive to replace Linux with Windows."
(for anyone who doesn't know, I work for Mandriva).
Good lord, does that article really say gray eminence? And does someone on the New York Times really think "well, Joe Sixpack isn't going to understand 'eminence grise', but he'll pick up 'gray eminence' like a hanging curveball"? Jeez.
Er, no. It's because One is only capable of installing the kernel it uses: One's 'installer' is really just a dd command (okay, I'm over-simplifying, but you get the idea) - it doesn't understand the concept of packages, it's just dumping a ton of data from one place (itself) to another (the hard disk).
So, it can only install the kernel it uses, no matter the hardware. Why does it use desktop586? Simple: because it supports the widest possible range of hardware. The default kernel - desktop - is built with i686 optimizations for performance reasons, so it doesn't support i586 CPUs, which there's still a few people out there using. So to make sure One runs on those CPUs, One uses desktop586.
This hasn't actually changed for 2009, but what has changed is that desktop586 is now built to support up to 4GB of RAM. Previously we only built it to support up to 1GB on the theory that no-one's likely to have an i586 system with >1GB of RAM. This is sensible, but misses the One case, where desktop586 is being used on everything. So we changed it for 2009.
I've not actually checked that (I think the KDE engineering team has), but I suspect it wouldn't work because it would rely on KDE 3-style IPC which isn't present in KDE 4.
It's rather because Slashdot didn't put the story on the front page (only as a between-stories bar, which is easy to miss). When they make a Mandriva story a full front page story, it still gets hundreds of comments.
Mandriva has always been a commercial distribution, ever since 1998. It's only got *more* open over time, never less.
We already have this (and have since 2008.1) - the XF86Audio keysyms. The problem is not at that level. That level is there and working fine in 2009 - pretty much all keyboards with multimedia keys pass the appropriate keysyms. The issue is specific to KDE 4 and it's just that it has nothing to *handle* those keysyms. KDE 3 has an app called KMilo which handles them, but it's not been ported to KDE 4 yet.
If you run KDE 3, GNOME, or Xfce in Mandriva 2009, the multimedia keys will work fine.
"I thought KDE4 was supposed to use less resources?"
I don't think they've ever actually said that, and it's certainly not the case:). There was a random comparison run by that ever dependable figure Some Guy On His Personal Blog which claimed 39% lower memory usage for KDE 4 or something, but it was basically utterly wrong. I did test this for my own curiosity - there's nothing insanely broken about our KDE 4 build, but on a clean 2009 install, the memory usage figures booted to a default desktop with a terminal and htop running are:
KDE 4 - 250MB GNOME 2.24 - 150MB Xfce 4.4.2 - 160MB(!!, although quite a lot of that is its hideously memory-hungry Terminal, you can reduce it by using rxvt) LXDE - 120MB
About 100MB of that is actually the base system plus X, so it's even worse if you only look at the numbers of the desktops themselves:
I didn't actually test KDE 3, but I think it's between 50 and 100MB by the second measure. There's probably some efficiencies to be had in KDE 4 somewhere, but it sure isn't less resource-intensive in a stock out of the box setup. I think this would be replicated in other distros, if you checked.
Slashdot Mirror
No X.org developer called anything a bloody mess. I used some choice terms (and I actually *edited* my post shortly after making it...), but I'm not an X.org developer. Phoronix introduced the 'bloody mess' term specifically, and they're not X.org developers either.
No, I'm telling them to use the existing architecture. To put it bluntly, to do it in X.org. I suspect the only reason it wasn't done this way in the first place is that X.org won't take a driver that depends on a binary blob.
Gwenole, thanks for the info about IEGD, but you're way off base about other stuff.
One, I know the drivers are not open source. That was a large part of what the rant was about.
Two, how can you not understand what my complaint is? It's not that there's no driver. And it's not that the driver is not open source, except that that is related to the real problem. My problem is very simply that the drivers don't fucking work. The post is simply a long explanation of the series of crappy decisions which caused that. I would like for the drivers to be open source and Intel has a public commitment to *make* open source drivers (that they're not keeping), but if they could get their shit together to provide a sane closed source driver that gave me functioning native resolution accelerated 2D, that would be absolutely freaking dandy for me (although it obviously wouldn't be satisfactory for Fedora). All I want is a driver that is available in a logical place, that installs sensibly, and that provides basic 2D functionality, on perfectly normal modern distributions (Ubuntu 8.10, Mandriva 2009, Fedora 10, OpenSUSE 11, etc). This should not be a problem.
Finally, I hadn't actually started working for Red Hat when I wrote that post, though it was incorrectly reported as being from 'Red Hat's Adam Williamson' in a couple of places. Regardless, I'm not in driver development, I'm not involved in talking directly to Intel. When I write about stuff like that on my *personal* blog - not on any official site for any distribution - I am writing as a user just like anyone else, quite simply saying that for a perfectly normal person who went out and bought one of these chipsets and tried to run Linux on it, the situation is massively crappy. I'm not going to let who I work for stop me from doing that, neither when it was Mandriva nor now it's Red Hat. If I want to take an issue up through official channels, I will do that, but if you see it on my blog, that is just me.
You seem to think there's no real problem with this situation, which is odd because you seem quite informed about at least parts of it. Everyone else I've spoken to with knowledge of it (including high-level X developers and folks directly involved with Intel on this issue) has quite clearly said that it is, basically, a giant clusterfuck. There doesn't appear to be any disagreement that it's a big mess.
Intel *didn't* make an open source driver. If you read my post, you'll note that there's three different closed-source components to the driver, without which significant features break.
Aside from that, what's needed for meaningful open source development is not "here's some code, have fun". There needs to be a proper development process hosted in an accessible fashion, and proper documentation. The 'intel' driver for previous Intel chipsets satisfies all of these goals. It's 100% open-source, it's developed within X.org and so easily accessible to external contributions in a widely-understood fashion, and the hardware is properly documented.
This 'psb' driver satisfies none of the goals. It was previously hosted within Moblin (which doesn't really have much of an external development community), and even that version of the code is now not being used. It now only shows up in obscure Ubuntu Netbook Remix repositories, with no independent source that anyone can find. So there's no sane development process to which external people can sensibly contribute. It contains large closed-source chunks. And there's no public hardware documentation, which makes it very hard for anyone else to work on it in the first place.
This is what I (and anyone else stuck with one of these chips) am complaining about.
Coincidence, given that the documents discussed in that post are about 8 years old.
That's the reputation of the power of the hardware. Yes, as far as playing your 3D games goes, you're not going to get very far with Intel.
However, up till now Intel has had a very good reputation for open source friendliness with regards to supporting the hardware, disregarding the actual power of the hardware. Intel are actively involved in maintaining the (100% open source) driver for all other Intel graphics chipsets, and they also contribute to general X.org development and the development of new technologies within X. Intel graphics hardware is generally the least powerful of the big three, but until this mess, it's been by far the best (and most openly) supported hardware in Linux.
My new desktop boots current Mandriva Cooker x86-64, from lilo to gdm, in 16 seconds. And a full working GNOME desktop with Firefox running takes about 25.
Of course, it's a heavily overclocked quad-core system with a pretty fast disk drive and 4GB of RAM.
Point being, random raw numbers (21.4 seconds!!!) don't mean an awful lot. Hardware is rather important. The original story's a good one - it explains the hardware used to test, and carefully compares ext3 and ext4, which is what they were actually trying to do. But Slashdot's abstraction of it is crap.
Well, I just built a new system and looked all this crap up the other day, so I feel qualified to comment.
Yes, the new chips finally just about match equally priced Intel chips - at stock.
The problem is they only go down to $250 or so. Not down to $200 or under, where the lowest-priced Core 2 Quad lives: the Q6600. And the thing about the Q6600 is, it can be trivially overclocked well past 3GHz on stock cooling with no enthusiast messing about. Just bump up the FSB speed and you're good. So for $200 or less you can get a year-old chip that will comfortably outperform AMD's latest and greatest. That's not such great news for AMD, is it?
And, don't forget, i7 is just Intel bleeding the 'enthusiast' market dry. Essentially they're getting an overpriced version of the new platform out first to capitalize on the market that'll spend any kind of money to have the fastest system on the block. Once they're done screwing those guys, they'll be making essentially the same chips available at much lower prices. Which will in turn drive down the price of the Core 2 generation, making them even more of a no-brainer purchase over the new AMD chips. AMD is basically a whole generation behind, and not looking like it'll catch up any time soon. I just don't see who - except someone who happens to have an existing AM2 motherboard that'll take one of these new chips and is looking for a modest upgrade - is going to buy one of these.
Well, this was a nice article.
It obliquely credits my own article on the same topic - written a whole one day previously - while describing it as a 'review' of Barry. No, it isn't. It's a guide to the exact same topic covered in this article, only it's a hell of a lot more useful, because my article tells you exactly what packages to download for what distributions, and how to actually use them. Rather than just saying "oh, look, this application lets you do this, isn't it cool?"
Would've been a lot more useful to link to me. But then, my server probably can't take the load, so I should be thankful. :)
http://www.happyassassin.net/blackberry-tethering-and-more-on-linux/
Boo freaking hoo.
""If the persons were real, such depictions could never be permitted,"Justice Adams said in his judgement."
"the mere fact that they were not realistic representations of human beings did not mean that they could not be considered people."
In other news, the director of 300 has been found guilty of mass murder, and sentenced to execution. ...and the world's gone mad.
Exactly. To me, telling people they should pay you not to assert your patents sounds exactly like...well...asserting your patents.
"Now, I'm bad at understanding why one person can't have the same idea after someone else has had that same idea"
That's not what Harmonix did. They were perfectly well aware of GF/DM when making Guitar Hero. The debt has been explicitly acknowledged in interviews with top-level Harmonix people.
I saw a post on Engadget this morning proclaiming how great a new, specially tuned pre-beta of Windows 7 was for netbooks. After all, it only takes 450MB of RAM to boot to a desktop!
*sigh*
Frankie70: the highlights? hiring relatives of government ministers, and offering $400,000 in "marketing funds" to a reseller in order to persuade them to replace Mandriva ($3 per machine) with Windows ($45 per machine) on a large order of Intel Classmate PCs.
Later on, the article covers the Mandriva / Microsoft in Nigeria battle that was covered here before:
"TSC approached Mandriva SA, a French company that sells a Linux version. Believing Microsoft had offered its $3 package, Mandriva proposed a $3 price for a Linux operating system, plus about $2 for other software, say people familiar with the situation. In August 2007, TSC issued a purchase order for Mandriva Linux, and the laptop's Taiwanese manufacturer began loading it.
Microsoft continued to push Windows. It offered its XP and Office software for about $45 per machine, says Nyimbi Odero, then TSC's chief executive.
Mr. Odero says Microsoft wanted TSC to delete Linux from the initial shipments of Classmates. He says Microsoft proposed a way to "make it worth your while" through a joint-marketing agreement. According to a draft agreement Microsoft sent to TSC last Sept. 13, Microsoft would pay TSC to fund "certain marketing activities to encourage the sale and distribution" of Microsoft products. Mr. Odero says Microsoft made it clear that TSC wouldn't really be expected to market the products, but could keep the money as an incentive to replace Linux with Windows."
(for anyone who doesn't know, I work for Mandriva).
Good lord, does that article really say gray eminence? And does someone on the New York Times really think "well, Joe Sixpack isn't going to understand 'eminence grise', but he'll pick up 'gray eminence' like a hanging curveball"? Jeez.
Reading only the title - as is traditional around here - why would I want an adhesive which only adheres to geckos?
http://bugzilla.mandriva.com/ There ya go. :)
"It was supposedly done to improve usability."
Er, no. It's because One is only capable of installing the kernel it uses: One's 'installer' is really just a dd command (okay, I'm over-simplifying, but you get the idea) - it doesn't understand the concept of packages, it's just dumping a ton of data from one place (itself) to another (the hard disk).
So, it can only install the kernel it uses, no matter the hardware. Why does it use desktop586? Simple: because it supports the widest possible range of hardware. The default kernel - desktop - is built with i686 optimizations for performance reasons, so it doesn't support i586 CPUs, which there's still a few people out there using. So to make sure One runs on those CPUs, One uses desktop586.
This hasn't actually changed for 2009, but what has changed is that desktop586 is now built to support up to 4GB of RAM. Previously we only built it to support up to 1GB on the theory that no-one's likely to have an i586 system with >1GB of RAM. This is sensible, but misses the One case, where desktop586 is being used on everything. So we changed it for 2009.
Yes, because there's too many damn Ones already. And we can't consolidate into larger ISOs because the point of One is to be one CD in size :)
I've not actually checked that (I think the KDE engineering team has), but I suspect it wouldn't work because it would rely on KDE 3-style IPC which isn't present in KDE 4.
It's rather because Slashdot didn't put the story on the front page (only as a between-stories bar, which is easy to miss). When they make a Mandriva story a full front page story, it still gets hundreds of comments.
Mandriva has always been a commercial distribution, ever since 1998. It's only got *more* open over time, never less.
We already have this (and have since 2008.1) - the XF86Audio keysyms. The problem is not at that level. That level is there and working fine in 2009 - pretty much all keyboards with multimedia keys pass the appropriate keysyms. The issue is specific to KDE 4 and it's just that it has nothing to *handle* those keysyms. KDE 3 has an app called KMilo which handles them, but it's not been ported to KDE 4 yet.
If you run KDE 3, GNOME, or Xfce in Mandriva 2009, the multimedia keys will work fine.
"I thought KDE4 was supposed to use less resources?"
I don't think they've ever actually said that, and it's certainly not the case :). There was a random comparison run by that ever dependable figure Some Guy On His Personal Blog which claimed 39% lower memory usage for KDE 4 or something, but it was basically utterly wrong. I did test this for my own curiosity - there's nothing insanely broken about our KDE 4 build, but on a clean 2009 install, the memory usage figures booted to a default desktop with a terminal and htop running are:
KDE 4 - 250MB
GNOME 2.24 - 150MB
Xfce 4.4.2 - 160MB(!!, although quite a lot of that is its hideously memory-hungry Terminal, you can reduce it by using rxvt)
LXDE - 120MB
About 100MB of that is actually the base system plus X, so it's even worse if you only look at the numbers of the desktops themselves:
KDE 4 - 150MB
GNOME 2.24 - 50MB
Xfce 4.4.2 - 60MB
LXDE - 20MB
I didn't actually test KDE 3, but I think it's between 50 and 100MB by the second measure. There's probably some efficiencies to be had in KDE 4 somewhere, but it sure isn't less resource-intensive in a stock out of the box setup. I think this would be replicated in other distros, if you checked.