Go Back   nV News Forums > Linux Support Forums > NVIDIA Linux

Newegg Daily Deals

View Poll Results: What driver do you use ?
I use the binary driver and use 3d acceleration. 90 67.67%
I use the binary driver but don't really need 3d acceleration. 14 10.53%
I use the free driver and actually "need" acceleration. 26 19.55%
I use the free driver and don't really need 3d acceleration. 3 2.26%
Voters: 133. You may not vote on this poll

Reply
 
Thread Tools
Old 08-01-06, 02:54 PM   #37
russofris
Registered User
 
Join Date: Jul 2003
Posts: 125
Default Re: Opinions on driver license

I use the Nv driver. I wish someone with the time and resources would port the nvidia UtahGLX driver to Mesa/DRI so that work could begin on adding 3D features to the OSS driver for newer Nvidia cards. The specs would be extremely helpful though. I'd also like to see nvidia develop an HTX compatible GPU to get around some of the Intel IP issues. The only other IP problems are the Microsoft vertex programming and fragment shading patents, and I dbout that we would see them act on that.

I'll see if I can get someone at Novel to act on the the Mesa/DRI driver (since they will no longer distribute the illegal binary modules).

Thank you for your time,
Frank Russo
russofris is offline   Reply With Quote
Old 08-01-06, 04:16 PM   #38
pierrec
Registered User
 
Join Date: Jul 2006
Posts: 9
Default Re: Opinions on driver license

Quote:
Originally Posted by pe1chl
I don't think so. You buy a card, and at that time you investigate the market. You have some OS and you look if the card supports it and to what level.
Then your requirements change. The newest OS release has features that your card does not yet support. I don't think you can request action from the card manufacturer when you have already bought something and your requirements change.
You misunderstood what I said. if I get a card today, of course I don't expect it to run tomorrow's software. Vista is a pig, but if Vista expect some super-duper video card and my Geforce MX200 doesn't support it, then that's fine.

However, when I get a card today, I would like very much to get today's features on Linux (or FreeBSD or whatever) as well as Windows, *and* I expect it to keep that same set of features as Linux evolves. With closed-source drivers, nothing is less certain: each time you get a new video card, it's a toss of the dice whether it'll work well or not under Linux. And each time you upgrade the kernel, there's a fair chance that the driver won't build anymore. And if you upgrade the driver, there's a good chance it'll break something that used to work too.

In Windows, the nVidia driver almost always works, and keeps working when Microsoft puts out updates. For some reason or other, unlike Linux users, Windows users are considered *users*, and users never put up with the kind of crap we do. nVidia and ATI/AMD know that, so they put the man/hours in the Windows driver and let the F/OSS folks marinade. They do just enough non-windows support so they can claim they do Linux too, but no more. And since they've got the 3D-graphics-under-Linux market pretty much cornered, they have no incentive to do more. That's why people whine.
pierrec is offline   Reply With Quote
Old 08-01-06, 04:29 PM   #39
pe1chl
Registered User
 
Join Date: Aug 2003
Posts: 1,026
Default Re: Opinions on driver license

You should not forget to mention that Microsoft goes to every possible effort to keep things compatible (sometimes making it more complex than it needs to be), thus Windows drivers tend to work for quite a long time. Only in a major release you need to update drivers, and even then not always.

On the other hand, open source developers (Linux, XF86/Xorg) find it rewarding to break the entire driver interface at each and every patchlevel. The more proprietary drivers fail for a new release, the better. They think.
This means the compatability problem is much more pronounced, and also costs the closed driver developer more time.

They probably think they can force manufacturers to give up and release the source to be done with it, but in cases where this does not happen they only hit the users (of Linux) with it. Contrary to developer's belief, there actually *are* people who use Linux, and are not kernel or X developers.

Even when you are a developer, there must be some base of things that works. When working on an application, it is no fun to be bothered by the effects of kernel updates all the time. This has driven me from compiling kernels from source into using a distribution long ago. And even now it irritates me that I have to re-install nvidia and vmware drivers even when completely unrelated security patches have been released by SuSE.
pe1chl is offline   Reply With Quote
Old 08-01-06, 05:04 PM   #40
acreal
Registered User
 
acreal's Avatar
 
Join Date: Nov 2003
Posts: 24
Default Re: Opinions on driver license

Quote:
Originally Posted by pe1chl
On the other hand, open source developers (Linux, XF86/Xorg) find it rewarding to break the entire driver interface at each and every patchlevel. The more proprietary drivers fail for a new release, the better. They think. This means the compatability problem is much more pronounced, and also costs the closed driver developer more time.

They probably think they can force manufacturers to give up and release the source to be done with it, but in cases where this does not happen they only hit the users (of Linux) with it. Contrary to developer's belief, there actually *are* people who use Linux, and are not kernel or X developers.
The general consensus agreeing not to enforce API/ABI compatibility between release (though many project do offer this to some extent) is actually a well thought choice, and by no means an attack directed towards closed source software. Compatibility requires a lot of work, and does not fit really well in an open source world, neither is it much wanted (still in this same FOSS world). So when the question "shall we ensure compatibility, requiring more work and a much smaller latitude for the code to constantly evolve ?" has virtually been asked, the answer was obviously no, because this is only needed by closed source software.
acreal is offline   Reply With Quote
Old 08-01-06, 05:22 PM   #41
pierrec
Registered User
 
Join Date: Jul 2006
Posts: 9
Default Re: Opinions on driver license

Quote:
Originally Posted by pe1chl
You should not forget to mention that Microsoft goes to every possible effort to keep things compatible (sometimes making it more complex than it needs to be), thus Windows drivers tend to work for quite a long time. Only in a major release you need to update drivers, and even then not always.

On the other hand, open source developers (Linux, XF86/Xorg) find it rewarding to break the entire driver interface at each and every patchlevel. [...]

They probably think they can force manufacturers to give up and release the source to be done with it, but in cases where this does not happen they only hit the users (of Linux) with it.
Well not quite. The Linux folks have made the conscious decision long ago that if something wasn't made Just Right[tm], they'll redo it, and compatibility be damned. In a sense, it's rather good, as things are clean, easy to work on, and lead to a lot less security-related issues. On the other hand, you're right, Microsoft goes out of their way to make things compatible with previous OSes.

Long ago, I maintained the DR-DOS kernel. Let me tell you, the code was a horror story. It had special cases all over the place, assembly blocks to reproduce faulty behavious previous versions had, so this-or-that program didn't break, it needed several assemblers and several C compilers (16 or 17 total iirc) to build, because some compilers didn't place undocumented data structures where popular programs that didn't follow the API were looking for them, etc etc... The last versions of this thing was just gross and impossible to maintain. If Linux people decided to go this way, I'd be worried. Even Microsoft is breaking the habit I think.

But the point is, if nVidia (or any other manufacturer that puts out closed source drivers) want to do Linux, then they need to bite the bullet and accept that they'll have 10x more maintenance work than on the Windows platform. imho, nVidia isn't doing too badly, I mean they follow the 2.6 serie close enough that the driver almost always builds on the latest kernels, but they don't seem to address bugs that have been outstanding for months or years, and they don't talk much to the community. I'm inclined to believe that the guys who do the Linux work at nVidia are understaffed and they honestly do their best, but it's still a shame that nVidia doesn't put more resources on F/OSS platforms.

I guess what I'm saying is, if people buy nVidia cards because they've read it's supported under Linux, then nVidia should do all it takes to fully support the driver under Linux, and not just do a half-assed effort, otherwise they shouldn't do it at all. These Linux folks are also customers who paid for their hardware, and customers don't like to be left out in the cold.
pierrec is offline   Reply With Quote
Old 08-01-06, 09:19 PM   #42
Jecos
Registered User
 
Join Date: Aug 2006
Posts: 5
Lightbulb Re: Opinions on driver license

"Closed Source kernel modules are illegal!!"
They kill progress on development and possibly violate the GPL.

The sooner a company ATI or Nvidia releases their source the sooner they have a crowd of people buying their card, b/c they won't have to wait for a binary driver that holds them back on newer and funner things .
Jecos is offline   Reply With Quote
Old 08-02-06, 03:04 AM   #43
pe1chl
Registered User
 
Join Date: Aug 2003
Posts: 1,026
Default Re: Opinions on driver license

The decision about what is more work, maintaining a compatible interface or having it change at every version cannot reasonably be made by the developers. It is obvious that having some compatability means some extra work for them, but it is questionable of the total amount of work spent (by the developers, the external manufacturers and the end-users) will indeed be more. I think it will be less.

In the early days I often tried to follow developments and contribute to it, but it quickly became impossible to do because everything is wired together. For every new kernel release there was a list of 20 other packages and the minimum release level, and invariably there would be 3 or 4 that you needed to update before installing this kernel. And those packages again depended on other packages that you needed to update.
So, like many others, I became a distribution user. SuSE in my case. Maybe once a year I update the entire system to a new distribution release and I run the kernel and X that comes with it.
What I don't get, is why an incremental update within such a distribution, e.g. a security fix in some networking protocol, needs to invalidate the compatability. The kernel version number changes so all modules are incompatible.
I know why this is happening. Long ago I discussed this on a mailinglist and a solution I proposed was even implemented (not sure if this was coincidence or because of this discussion). The idea was to provide every interface function with an automatically generated signature that validates that the interface did not change. This is the MODVERSIONS option. But as it cannot describe every aspect, it is apparently not trusted, and we still need to recompile all modules.

Imagine the hell that would break loose when you would need to update your nvidia video driver on Windows for every visit to Windows Update. Nobody would accept that. But in Linux it is normal practice.

About opensource of the driver: I think most advocates of this are wildly underestimating the whole issue. The binary module is 4.5MB in size. The sourcefiles probably are something like 50MB.
It would be very naive to assume that this piece of coude would suddenly "maintain itself" after being released to the public. Every issue would still need to be tracked and fixed, and while some issues may be simple, there still will remain a large number of issues that are not so easy to fix.
Look at projects that have been in open source for years, like Mozilla (now Firefox and Thunderbird). It has thousands of open bugs, and some very irritating problems have not been fixed even while they have been in bugzilla since 2002. So availability of the source certainly isn't the end of all problems.
Besides, even when opensourcing may smooth the integration with new developments in X and the solution of problems that the manufacturer considers low priority, at the same time it will slow the development for new hardware. Contributions for new hardware drivers will certainly not be released before the new hardware is in the shops. Now, nvidia can develop drivers for hardware that is still unreleased and have them released at the same time as the cards.
pe1chl is offline   Reply With Quote
Old 08-02-06, 08:50 AM   #44
acreal
Registered User
 
acreal's Avatar
 
Join Date: Nov 2003
Posts: 24
Default Re: Opinions on driver license

Quote:
The decision about what is more work, maintaining a compatible interface or having it change at every version cannot reasonably be made by the developers.
Considering the developers are the ones doing everything, there really isn't anybody else to make the call. Maybe this decision does not suits closed source software well, but you really have to consider the sheer number of people involved in FOSS: in open source environments closed source software are a rather small minority.

Quote:
It is obvious that having some compatability means some extra work for them, but it is questionable of the total amount of work spent (by the developers, the external manufacturers and the end-users) will indeed be more. I think it will be less.
Well for one thing, fixing internal bugs and/or bugs caused by 3rd parties wrongly using the API without breaking it can be difficult. Another factor is that most open source software is written by volunteers, and, as you said, breaking compatibility is so much fun.

Quote:
In the early days I often tried to follow developments and contribute to it, but it quickly became impossible to do because everything is wired together. For every new kernel release there was a list of 20 other packages and the minimum release level, and invariably there would be 3 or 4 that you needed to update before installing this kernel. And those packages again depended on other packages that you needed to update.
So, like many others, I became a distribution user. SuSE in my case. Maybe once a year I update the entire system to a new distribution release and I run the kernel and X that comes with it.
I don't really see this as a problem. As for development, *some* people must succeed in following it. Clearly it demands dedication, and few people have the time to do it, but in the same time it show that this bazaar development thing isn't totally unrealistic, even on a large scale.

Quote:
What I don't get, is why an incremental update within such a distribution, e.g. a security fix in some networking protocol, needs to invalidate the compatability.
Come on, every security or bug fix release does not invalidate compatibility.

Quote:
Imagine the hell that would break loose when you would need to update your nvidia video driver on Windows for every visit to Windows Update. Nobody would accept that. But in Linux it is normal practice.
Well, i'm sorry to say this but the actual breaking is not caused by the Linux kernel, but by the Nvidia kernel-side driver (just a different pov).

Quote:
About opensource of the driver: I think most advocates of this are wildly underestimating the whole issue. The binary module is 4.5MB in size. The sourcefiles probably are something like 50MB.
It would be very naive to assume that this piece of coude would suddenly "maintain itself" after being released to the public. Every issue would still need to be tracked and fixed, and while some issues may be simple, there still will remain a large number of issues that are not so easy to fix.
Of course not, but if the drivers were still supported/developed by NVidia, it would let the time for X and kernel developers to grasp and adapt the code to what best suits them. Once fully merged, NVidia could still support support the drivers if they want to, and provide specs for their next generation cards. Why not even hire one or two of the kernel/xorg devs to participate a little ? Besides, open sourcing would allow NVidia to accept code contribution, fixing bugs and adding features without having them to write said code. Fair trade really. (again, this thread was not intended to talk about open sourcing NVidia's drivers )

Quote:
Look at projects that have been in open source for years, like Mozilla (now Firefox and Thunderbird). It has thousands of open bugs, and some very irritating problems have not been fixed even while they have been in bugzilla since 2002. So availability of the source certainly isn't the end of all problems.
It is. Open source is paradise, bugs flee away (to closed source software, where they belong). Of course not. And open source does not mean success or quality either.

Quote:
Besides, even when opensourcing may smooth the integration with new developments in X and the solution of problems that the manufacturer considers low priority, at the same time it will slow the development for new hardware.
You must be joking. This document lists the following as linux primers:
  • USB 2.0
  • Bluetooth
  • PCI Hotplug
  • CPU Hotplug
  • memory Hotplug (ok, some of the older Unixes did support CPU and memory hotplug in the past, but no desktop OS still supports this.)
  • wireless USB
  • ExpressCard
Besides, i think i remember the Linux kernel being amd64 ready before anyone else too (but i'm not sure though). Of course if you talk about the closed spec hardware, this isn't really a fair comparaison.

Quote:
Contributions for new hardware drivers will certainly not be released before the new hardware is in the shops. Now, nvidia can develop drivers for hardware that is still unreleased and have them released at the same time as the cards.
Why not ? Many kernel developers are willing to sign DNAs, just send in a prototype card and the specs for it, and they may just be happy to get you linux support (or else, this is not a linux world after all) for free. This is just something companies are not used to (yet).


We finally got down to a nice open source vs closed source thread, how unpredictable. Though i'm pretty happy with the thread. It kinda shows what i was thinking: on both extremes you and me (i'm not that extreme actually, and i guess you're not either ), and in between most people don't really care, or just think it would be good to have a free driver. Which just shows how much our computing experience/environment influence our point of view.
acreal is offline   Reply With Quote

Old 08-02-06, 12:43 PM   #45
kwizart
Registered User
 
Join Date: Feb 2005
Location: Paris, France
Posts: 129
Default Re: Opinions on driver license

I do not have read all the topic but few things.
What i understand of that probleme is that inserting non-GPL kernel-module inside the linux kernel is simply illegal.

The poll seems to say that is a user choice. But it is not. Since Novell do not provide anymore closed module (As Fedora/Redhat that i use) they only move the legality problem to the responsability of the users... This is not a acceptable!

The linux kernel is GPL, then all module for that kernel need to be at least GPL. Since we use linux and with Nvidia 3d acceleration or not we need to have fully functionnal hardware. This implies to provide a linux driver to the users and this driver must be GPL...

The main problem behind this, is that the developpment of the whole architecture of the kernel is made difficult if driver are not open sources! That is why it is different from non-GPL sofware under linux that do not need to be open-source.
kwizart is offline   Reply With Quote
Old 08-02-06, 01:35 PM   #46
Thunderbird
 
Join Date: Jul 2002
Location: Netherlands, Europe
Posts: 2,105
Default Re: Opinions on driver license

The linux kernel module thing is a very gray area. The nvidia drivers don't actually 'link' to the nvidia drivers and that's what makes it unclear. I forgot where I read the way it is done but it is done a bit in the way of the following analogy. If the linux kernel is a webserver then the nvidia linux module can be seen as a webbrowser. It doesn't really use kernel functionality it only uses some information from the kernel.

The whole GPL issue came up recently with the Kororaa distribution which was shipping prebuild nvidia/ati drivers with their kernel. The following article explains the gray area quite clearly: http://kororaa.org/index.php?entry=entry060521-200059
Thunderbird is offline   Reply With Quote
Old 08-02-06, 04:28 PM   #47
Robster
Registered User
 
Join Date: May 2004
Location: Christchurch, New Zealand
Posts: 40
Default Re: Opinions on driver license

Quote:
Originally Posted by Jecos
"Closed Source kernel modules are illegal!!"
They kill progress on development and possibly violate the GPL.
I've been enjoying this thread because of the considered opinions, and the lack of hyperbole and posturing. Please don't post unsupported opinions like the above, it will just reduce my enjoyment

The binary module as nVidia do it, does not "kill progress on development". nVidia has not to my knowledge tried to reduce innovation in the Linux kernel or X, and they have shown themselves to be responsive in (eventually!) making their module compatible with the changes that have been happening in those areas.

As for "possibly violating the GPL", Thunderbird's link above is a nice discussion of the issue. I don't think nVidia violates the GPL at all, because they provide a binary module which doesn't use GPL code, and a source code component which must be compiled to produce an actual kernel module. It is by compiling the source code together with the binary module that an arguably derivative work is produced, and nVidia doesn't do that - the users or distributors do it. So while Kororaa may be violating the GPL by distributing a derivative work, nVidia certainly isn't. And in my case, because I compile the module myself and don't distribute it, there is no GPL violation either.

Quote:
Originally Posted by Jecos
The sooner a company ATI or Nvidia releases their source the sooner they have a crowd of people buying their card, b/c they won't have to wait for a binary driver that holds them back on newer and funner things.
OK, well I'll be a little bit inflammatory here because my opinion is that even now, with GNU/Linux use really starting to make inroads into the Windows monopoly, Linux needs nVidia more than nVidia needs Linux. Right now, Linux support probably brings nVidia a few thousand card sales per annum. If nVidia stopped supporting Linux right now, I don't think their bottom line would suffer in any significant way.

To achieve "Total World Domination" for GNU/Linux and the open-source Way, Linux not only needs to be demonstrably better than the alternatives, but mindsets and business methods also need to change. And changing people's minds always takes quite a bit of time and effort

Linux is better than the alternatives (IMHO), but part of the reason for this is because of the effort that has gone into making it conform to the old Windows/proprietary model. As well as kind-of allowing binary modules, I'd put Wine and ndiswrapper in the same category. Linux has not yet succeeded to the point where third-party application suppliers and hardware manufacturers always provide support for Linux, and that point may yet be some way off. In the meantime, to provide a satisfying experience for users, there has to be a way for people to use their existing and familiar applications and hardware.

So Wine and ndiswrapper especially, provide a sort of "half-way house" for new adopters of Linux. Once people are using Linux, then in my experience they really appreciate the variety and quality of the Free alternatives to their existing solutions and will migrate away from their proprietary ones. It's a very powerful effect! But it is vital that that initial experience is positive.

As for changing business methods, you can understand the unwillingness of a company that makes all their money from software, to make that software Free . Most application providers can continue to make binary-only applications that will work on most Linux distributions, thanks to LSB/LFS etc. Also, many bits of software that are "drivers" in the windows world can operate in user space on Linux, which means that they can be binary-only as well. The grey area (forgive my British spelling, I can't bring myself to say "gray"!) is for software that must exist in kernel space, and nVidia is in that category.

What is "an nVidia product"? I don't think you can say it's just the card, chips etc. The driver is an integral part of the product because without it you can't make the hardware work to its potential. I'm sure there is technology in the kernel driver that could be used to make an ATI card, for example, work better- maybe things like the code paths for different CPU capabilities, or the pre-rendering calculations that have to be done. It's perfectly valid for nVidia to want to keep their work from enhancing their competitor's products, which in turn justifies keeping the driver closed-source.

For this to change it would be necessary for the software to become truly a low-value companion to the hardware, as opposed to an integral part of the package, which it is at the moment. I can actually see a future where this is the case: while DRI is just a mechanism for getting fast access to the hardware, if there was another generic layer on top of DRI for decomposing 2D and 3D drawing commands into a lower, more hardware-compatible format, then perhaps it would minimise the need for vendor-supplied software. Is this what Mesa does? I don't know enough about the technology to comment.

It seems at the moment that Microsoft drives the hardware developers to some extent, by publishing its DirectX specs. I'm sure that nVidia and ATI optimise their hardware designs to give good performance when rendering the output generated from DirectX. Maybe in the future, X and OpenGL will (once again?) be the most important references for the hardware developers to design to
Robster is offline   Reply With Quote
Old 08-02-06, 06:37 PM   #48
energyman76b
Registered User
 
Join Date: Dec 2002
Location: Clausthal/Germany
Posts: 1,104
Default Re: Opinions on driver license

Quote:
Originally Posted by kwizart
I do not have read all the topic but few things.
What i understand of that probleme is that inserting non-GPL kernel-module inside the linux kernel is simply illegal.
no, distributing a closed source module together with the kernel is illegal. Maybe programming one.
Inserting is a totally different area.

Quote:
Originally Posted by kwizart
The poll seems to say that is a user choice. But it is not. Since Novell do not provide anymore closed module (As Fedora/Redhat that i use) they only move the legality problem to the responsability of the users... This is not a acceptable!
and here you are totally wrong. What the user does is outside of the scope of the GPL. You can do with the software whatever you want, as long as you do not distribute it. So loading a closed source kernel module AS USER is absolutly LEGAL. It may be imoral (from a kernel dev point of view), but there is nothing that makes it illegal. The user is allowed to load such modules.

Quote:
Originally Posted by kwizart
The linux kernel is GPL, then all module for that kernel need to be at least GPL. Since we use linux and with Nvidia 3d acceleration or not we need to have fully functionnal hardware. This implies to provide a linux driver to the users and this driver must be GPL...
well, yes and no. There is some discussion about it. If the module uses kernel interna, than yes, it has to be under the GPL as soon as it is distributed. But if it does not use kernel interna and uses a GPL'ed glue layer to the kernel (like the nvidia) module, they might be on the safe side.
Besides of that, Linus B. Torvalds said something about the nvidia module a long time ago, that makes NVIDIA sit on the safe side anyway . . .

A lot of people would be happier, if nvidia would provide opensource drivers or collaborate with devs to create one (or even pays a different corporation to create that drivers, like ATI did a long time ago), but as long as nvidia thinks that keeping the drivers closed is better for it, that will not happen.
energyman76b is offline   Reply With Quote
Reply


Thread Tools

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Forum Jump


Similar Threads
Thread Thread Starter Forum Replies Last Post
Redhat 8.0 NVIDIA works - INSTRUCTIONS STEEL1 NVIDIA Linux 267 04-15-03 06:48 PM
Getting the proprietary nvidia driver to run with Debian 3.0 r0 (woody) Katchina404 NVIDIA Linux 9 01-12-03 08:49 AM
failing to initialize NV driver PsyShook NVIDIA Linux 10 09-19-02 12:41 PM
Nvidia Driver causes crash on Standby/Suspend dasher NVIDIA Linux 5 09-16-02 05:57 PM
new nforce driver!! 1.12!! frenchy2k1 CPUs, Motherboards And Memory 9 08-27-02 01:09 AM

All times are GMT -5. The time now is 12:44 PM.


Powered by vBulletin® Version 3.7.1
Copyright ©2000 - 2014, Jelsoft Enterprises Ltd.
Copyright 1998 - 2014, nV News.