Slashdot Mirror

We need to get a truly working pluggable driver model.
The UDI Project tried something like this. It didn't work, mostly because of political reasons. Platforms that are very popular with driver writers (read: windows, solaris to some extent) have an advantage in NOT having a pluggable architecture like this. So the big guys will never have a pluggable infrastructure. So that means the pluggable infrastructure will be used only (if at all) by fringe operating systems. So, then the hardware manufacturer has a choice of spending extra time and money for coding and testing on a minor architecture. Classic chicken and egg, Catch 22. PLus, a pluggable architecture makes it easier for binary modules, and there are some people very opposed to that.

The only thing Linux has over other operating systems right now, is price.
If you use RedHat Advanced Server, i can guarantee you price is not an advantage. Yet, they have more and more people using it. I think you're grossly oversimplifying things.

Why do people insist on reusing names for unrelated things? Project UDI is a technology allowing device drivers to be portable across different operating systems and platforms. Project UDI doesn't address display technologies, much less DRM. This "Unified Display Interface" seems to be something entirely different, and it's unfortunate that they're trying to re-coin the "UDI" acronym. The UDI link in the summary is simply wrong.

On the other hand, Project UDI is a very cool technology that people should be supporting, so I guess the extra exposure could help, as long as people don't confuse UDI (Uniform Driver Interface) with UDI (Unified Display Interface)... *sigh*

Why talk about something called Unified Display Interface and then link to the site www.projectudi.org which concerns itself about the Uniform Driver Interface?!?. Slashdot editors at its best I guess...

Not that the Uniform Driver Interface is that great idea either, it's some kind of let's make some cozy wrapper that lets hardware manufacturers cross platform binary only drivers.

And what about Unified Display Interface? The only thing I can find about it is the sensationalist blurb on The Register. Have they just invented a new name for the new DisplayPort standard from VESA maby?

/greger

There was a commercial supported attempt at this, the Uniform Driver Interface project. It was an environment loaded into kernel space which provided APIs for memory management, interrupts, memory mapped I/O, and all of the normal stuff you'd expect for a driver. There were no synchronization primitives, your code segment was to be run until you yielded and would not be interrupted. as such you had responsibility to compartmentalize your code into smaller chunks, with the benefit of having someone else deal with all the synchronization stuff. It was somewhat message passing, a comparison to DragonFly BSD probably wouldn't be inappropriate.

Cool parts: device drivers are source code compatible on all platforms, no code changes at all. Binary compatible for platforms that share an ABI (Can compile adriver on Linux x86 and have it work for UnixWare x86).

But it never really went anywhere outside of Caldera. That website looks pretty much what it looked like in year 2000. Most of the issues we dealt with were non-technical. One big one being the implosion of Caldera (which became NewSCO). Caldera wanted this environment to make it easier for people to move device drivers from OpenServer (which had a lot of device drivers but was essentially a dead codebase) to UnixWare (which was newer, SVR4.2, then became SVR5).

There was/is a Linux environment, but RMS hates the fact that it's easy to have binary only drivers, making it less likely to release source. Obviously Microsoft wouldn't use it since everyone makes drivers for windows, and that's their competitive advantage. I forgot why the BSDs never used it. I don't think it was licensing, we used a BSD style license.

What with Solaris's recent still-mysterious "open sourcing", the large amount of cash infused into SCO by Sun Microsystems

There is only one mystery about the money Sun paid to SCO, and that is, "How long will we keep having to read this innuendo when what they bought is well known?" Sun bought UDI drivers and driver technology so Solaris X86 wouldn't suck so bad. Or maybe I'm reading this wrong, and the reason this keeps coming up is that some members of the Linux community feel the need to engage in FUD of their own?

One wonders if Sun Microsystems might be hoping that the SCO suit will drive people from Linux to Solaris the same way that the USL suit drove people from BSD to Linux.

They can't be hoping that. People interested in free software would just go the BSD route. The BSDs are every bit as good as Linux. GNU software runs very nicely on the BSDs. You can forget that one.

increasingly common yet always vague claims by Sun executives that "intellectual property issues" will become of increasing importance in software development in the near future

What is much more likely is that the Sun execs know what they have in Solaris, and the technology that they have been developing for it. Some of the best of that technology will only be in Solaris land. You want that technology? You run on Solaris. Linux NFS still makes me clench my teeth from time to time.

Sun paid for Uniform Driver Interface driver technology and drivers from SCO for Solaris X86. They were trying to revive Solaris X86 after almost killing it and creating an uproar among their customers. Paying for that driver technology gave them quick and easy access to an important technology to make Solaris X86 more competitive in the future. They saved considerable time by buying a working System V Unix implementation of those drivers to incorporate into their System V Unix implementation. Using UDI drivers lets them easily leverage drivers written to the UDI spec for other oerating systems for their own operating system. Incorporating UDI technology into Solaris X86 made it a better product that will be more competitive in the future, will cut costs, and make more hardware available for it. I don't see any mystery in why they would be interested in it.

Sun paid 9.3 million for what?

Sun paid for Uniform Driver Interface drivers that SCO had developed. Since SCO and Sun both had System V Unix it would have been trivial for Sun to port them from Unixware to Solaris X86.

Sun badly needed this since they were trying to reinvigorate Solaris X86 after almost killing it. Their customers wanted Solaris X86.

Uniform Driver Interface support means that Sun can easily leverage drivers written for other operating systems. It helps them support more current hardware with much less effort. It makes Solaris X86 suck much less in terms of hardware support.

No conspiracy, just buying IP to cut development time and costs while making their product more competitive. Nothing more to see. Just move along.

Not everything is about trying to trip up Linux. Sometimes a cigar is just a cigar.

Maybe you missed the news flash, but Sun also happens to have been a major contributor to SCO's legal war chest, and continues to be an antagonistic purveyor of anti-FOSS FUD.


You seem to have missed a few news flashes yourself. SCO is a business, not a charity. Sun didn't "contribute to SCO's legal war chest", they bought IP to incorporate into Solaris for X86, namely Uniform Driver Interface based drivers. Don't let the fact that they paid a lot of money fool you... that is the way that capitalism works. Sun and SCO are competitors, they both sell Unix System V operating systems for X86. Why should SCO help them? It will only make their competitor stronger. Well, maybe if Sun wanted it bad enough to pay real money SCO would listen to them. After all, Sun was being slammed by its customers for appearing to abandon Solaris X86 so they might be motivated to pay some real money from the BILLIONS it had in the bank. I guess you will just have to damn Sun for listening to its customers and trying to make Solaris X86 better. And, if they really were just propping up SCO to attack Linux, you would think that they would have spent more since this was going to be a one time deal and they had billions in the bank. If their purpose was to crush Linux they were being damn cheap for something you apparently believe they thought would make or break them. Or maybe this has nothing to do with crushing Linux and everything to do with making Solaris better. Wow. It is almost inconceivable, a company trying to make its products better to make them more appealing to their customers. It is crazy, but it just might work.

You will also have to excuse them for not folding at the announcement by some in the Linux community that they are doomed, that Linux will crush them. Sun has had to listen to announcements of its coming doom for almost 20 years. Microsoft will crush them any day now. Or IBM. Or HP. Or DEC, er Compa, er HP. Riiiight.

Yes, Sun is hostile to FOSS. Just compare their record to IBM and HP:

Sun IBM HP
Yes(1) Yes(2) No Contribute / Pay for Linux file system
Yes(3) No No Buy & open source entire office suite
Yes No No Support development of Xemacs
Yes(4) No No Buy and open source network queue sys
Yes(5) No No Contribute important code for X11
Yes(6) No No Incorporate Gnome into their Unix
Yes Yes Yes Sells Linux on desktops & servers
Yes No No Lives or dies by success of *nix
Yes(7) No No Contract to sell 500,000 - 1,000,000 Linux desktops / yr

Wow! Now that is a damning indictment of Sun!

On second thought, you just seem to be a part of the anti-Sun FUD brigade. Or maybe you really are just clueless. Whatever.

1) NFS v4
2) JFS
3) Star Office -> Open Office
4) Codine queue -> SunGrid
5) Internationalization system code for X11
6) Gnome 2 part of OS. Others make it separate install as add-on.
7) Java Desktop systems to China

You mean something like the Uniform Driver Interface?

Then perhaps someone will explain being why Sun paid SCO $8M for a worthless SCO licence (along with Microsoft, themselves no friend of OS).

Sun didn't buy a worthless license. SCO has an implementation of driver technology that Sun needed in their bid to revitalize Solaris on X86 processors. In fact, they picked up hundreds of drivers. Now, if you have any suggestions as to other System V Unixes on X86 with a large base of relatively recent drivers and driver technology that Sun could have used instead, I'm sure that Sun would love to hear about it. Until then, Sun has a real business to run, and buying IP from SCO no doubt saved them an enormous amount of time, and maybe even money.

Bottom line: this has nothing to do with SCO attacking Linux. Sun was just taking care of business.

(By the way, you do know that Sun has deals to sell something on the order of 500,000 - 1,000,000 of the Linux based Java Desktop systems a year, don't you? You do know that Sun sells servers with Linux, don't you?)

I was aware of the Uniform Driver Interface (UDI) project, but again, I don't think their goal was exactly what I had in mind. They specify a "UDI environment", which is an API to be provided by any OS that wants to support UDI drivers. What I'm talking about is making the proprietary driver code present an API much more like a device BIOS interface.

The "UDI environment" is just their way of saying that the OS has to implement UDI's API to run UDI drivers. (Well, duh!) UDI drivers can be proprietary code -- the UDI specification is not under the GPL, and is free for anyone to implement. There is also a reference implementation of a UDI environment (for Linux) which is open source.

I'm not a lawyer, but I can't fathom any argument for claiming an independently-developed UDI driver to be a derivative work of Linux, even if you choose to load it dynamically into a Linux kernel. (Obviously, static linking creates a binary which is a derived work of both, and therefore couldn't be distributed.) UDI effectively neutralizes the GPL at the device-driver boundary...

First, M$ might not make very secure software, but they sure got a load of talented people working there, and even if their software is crap they know how to do some things right, the NT kernel is not all that bad(after all, it's just VMS ;)), and they did put a *load* of effort in making drivers compatible, and they had the src for all the OSes they wanted to make compatible, and they still failed miserably, other efforts by a bunch of amateur kids aren't going to work much better(and if you do some research, you will find that some other people have tried, and failed too).

I was aware of the Uniform Driver Interface (UDI) project, but again, I don't think their goal was exactly what I had in mind. They specify a "UDI environment", which is an API to be provided by any OS that wants to support UDI drivers. What I'm talking about is making the proprietary driver code present an API much more like a device BIOS interface.

And yes, I *really* need to know how to interface with my hardware, and no, adding a layer of proprietary software on top of proprietary hardware is *not* OK, for *many* reasons, for example security, stability, CPU and OS portability, etc...

Yes, you do need to know how to interface with your hardware, and so does everyone else who wants develop an OS. Issues of security and stability apply to hardware as well as software. Although I am addressing OS portability, CPU portability would remain an issue. It could be addressed by releasing proprietary code in some low-level intermediate code, or by the device manufacturer simply responding to reasonable requests from developers.

In all cases specs are withheld to screw the user, and nothing else, which is *exactly* what Open Source tries to avoid.

Chill. It's not that black and white, as several other posters have pointed out.

You have to remember that in Open Source, we are all the developers, and M$ got access to the src of all their drivers(their license for hardware vendors force them to give the src to M$), and we at the very least need the same.

M$ no doubt also agreed not to disclose the vendors' source to the whole world. Open Source can't make that same agreement by its very nature. Nevertheless, I won't claim that even that concession was necessary, given M$'s monopoly power. I certainly would encourage you to go back to the hardware vendors and try again as soon as Open Source takes over the world (which would be fine with me).

Perhaps a "Driver Adapter" could be built that would allow drivers written for it to run on any OS? The basic concept is that the adapter itself would be a driver for the OS, then the "Cross Platform Drivers" would deal directly with the adapter.

Check out the UDI Project. Its a project by some of the big name computer and software companies to create a Uniform Driver Interface. They released a reference implementation under the BSD license and a Linux implementation has been made. The problem though is getting hardware companies to code a UDI driver.

It helps when you pay attention, and you obviously aren't.

There is a difference between a Unix source license, which Sun has, and a license for new drivers / driver technology which SCO had and Sun didn't. (Developed from this maybe?) That's not too subtle of a distinction for you I trust. Instead you keep trying to spin this into something else. So, were you simply confused, or lying?

If Sun bought real IP that added value to their product then I don't see any big deal about it. Solaris X86 has been criticized as being driver poor, so why shouldn't they try to improve it if that's what their customers want? I suppose they could have bought it from any of the other System V Unix on X86 vendors, and maybe they should have. There is just one problem with that. Who are they? Are there any left? Ten years ago, maybe, but not anymore. Or is it your view that Sun should simply ignore their customers and shareholders, replace Solaris everywhere with Linux and damn the consequences?

Your distinction between buying product from SCO and getting warrants as part of a large IP deal is largely meaningless. Its all cash in SCO's pocket. If anything it shows Sun to have shrewd negotiators since it offers the potential to leverage Sun's money. The Linux business community could use a few more shrewd negotiators given the number of Linux oriented companies that have gone down the tubes the last three years, harming both Linux technical development and the people involved.

You shouldn't be confused about a one-time expense that is paid over time since its not a hard concept to grasp. I'm sure you could find an accountant or business teacher to help you understand if you are still confused.

As to the rest of your statements, you are apparently speaking to a generic "Sun apologist" strawman, not to any point I made.

You are apparently passionate for Linux and against Sun. To me Linux is just a useful technology, and Sun is a vendor of a number of useful technologies. Unfortunately your passion combined with an apparently tenuous grasp of facts and events is causing you to make statements containing blatant nonsense.

They have been telling everyone for years that they have "full rights" to their UNIX. Except when they took a closer look, they really didn't. Oops

Thats not right. Sun paid for drivers, as in getting more drivers, not for stuff they already had but didn't have clear rights for. It was probably the UDI project which could provide benefits for Sun for years to come. That should help make Solaris X86 easier to support.

You need to make some adjustments to your conspiracy theory.

Schwartz: We took a license from AT&T initially for $100 million as we didn't own the IP. The license we took also made clear that we had rights equivalent to ownership. When we did the deal with SCO earlier this year we bou

ght a bunch of drivers and when we give money to a company oftentimes we get warrants, which is part of the negotiations. I have warrants in 100 different companies, we have a huge venture portfolio. I can't do anything about the perception that's out there and to be blunt, I don't care as those people aren't going to drive our future?customers are.

Sun has paid Sco a quarter million. Why on Earth is it helping bankroll an attempt to destroy Free Software?

It's time to boycott Sun. It's about time it learned you can't bite the hand that feeeds you.

Visit this driver project web site and get a clue.

Here is another clue: Its not all about Linux. The world does not evolve around Linux. Companies developed, sold, and licensed IP before Linux, now, and will do so long after Linux is either everywhere or nowhere.

Just because you don't understand it doesn't make it a conspiracy. It just may mean that you are ignorant about the details or the market. It might also mean that you are just paranoid and viewing everything through Linux tinted glasses.

Come up for some fresh air once in a while.

ms and sun "licensed unix from sco" - i think they gave sco some money to play with without getting their own hands dirty.

Of course its a conspiracy, what else could it be? Sun doesn't have any products that lack anything, like, say, drivers that Cald^H^H^H^H SCO would be in a strong position to provide by license. And everyone knows that Microsoft can't have more than, what, one or two products directly related to Unix that having source code from SCO would be useful to help improve, or for which a license would help remove any doubt about IP rights.

They are just pretending to improve and sell their products to meet Wallstreet expectations when everyone knows its just a plot to undermine Linux. Linux will still crush them and rule all. Go Tux!!

You are misinformed. Check this link

I see SCO and Intel mentioned. No Sun, although they are part of the Consortium. (Oh, and notice who hosts the site.) This entire project started after 1994. I think that it is pretty clear that Sun was paying SCO for access to SCO's work on a new standard driver model. More portable, less work, more drivers, nuff said.

No conspiracy, just better support for Solaris.

http://www.projectudi.org/ Or existed, it seems like it hasn't been updated in a while.

I did some work in this, interesting environment. All drivers are source compatible across all conforming environments. If the environments have the same C ABI (say, they're all x86) then the drivers are binary compatible. Caldera actually made this their native DDI for OpenUNIX 8. There are environments for OU8, Linux, and FreeBSD, and some drivers out there. One cool thing is that the environment handles MP. All your functions are guaranteed never to be interrupted. Downside is, it's a very different environment from what you're used to, and it takes a while to get your head around.

As an aside, RMS doesn't like the environment because it makes it easier to release binary only drivers. Not only does it insulate youfrom DDI differences between platforms, but also between Linux kernel updates.

FYI there's Project UDI (Uniform Driver Interface).

From their FAQ:

Like other device driver interfaces used in Operating Systems today, UDI defines an architecture and a set of APIs for use between the driver and the surrounding system. This allows drivers and OSes to be developed independently. UDI goes a step further and provides APIs that are OS-neutral and platform-neutral, allowing multiple OSes and platforms to use the exact same driver.

Don't need one OS just to get drivers working everywhere.

Check out Project UDI, device drivers source compatible everywhere, binary compatible on platforms with the same ABI. This is what OpenUNIX 8 uses under the hood. Microsoft will never join, because everyone targets them anyway, and its to their advantage for stuff not to work on other systems.

Being a hobby OS writter myself I've looked into device driver portablity. The best effort going (which isn't saying much) is The UDI Project. I have to wonder why more device driver writers don't try to use UDI....

Obviously, having to support the vast amount of devices out there has been one of the most major contributors to the downfalls in operating systems.

There is a project, UDI (Uniform Driver Interface), which is working to rectify the situation, though. It's intent is to even the playing field, by providing source (and in many cases binary) compatibility for device drivers across os and hardware platforms. All the major players are involved - IBM, Sun, Intel, HP, ... (Microsoft, most notably missing from the list).

They already have reference drivers working for several platforms... It will be interesting to see how this plays out.

Is a standard driver interface for x86. Preferably managed by a non-profit group who could then permit the manufacturer to display a logo on the box once the driver has been tested

You can't handle a driver failure. By definition a driver is part of the kernel, and if the kernel screws up, you're toast.

This is an unsupported proposition. While it may be typical for a driver to be implemented in kernel mode, it's not "by definition" a requirement. Take a look at Project UDI for a device driver API that could be implemented in user-mode. This would allow the system to be equally protected against untrusted drivers as it can be from untrusted applications. Now, I'm not saying that the Reference Implementation or any other UDI implementation actually supports user-mode drivers today. However, the API standard is written in such a way as to allow for that possibility.

Of course, user-mode drivers would execute more slowly than kernel-mode drivers, but if an OS supported both types, paying the performance price for stability on untrusted drivers would be a worthwhile tradeoff...

I'm no kernel hacker, but I guess this has a chance of a snowball in hell to get adapted by the kernel folks.
First, the idea seems old, look at the UDI project.
Looks even cleaner IMO.
And now watch the opinion of several high grade kernel hackers about it.
Two citations:
Dave Miller: "No thanks, IMHO OS neutral driver interfaces are a nice idea but they can only lead to mediocrity. (Yes I have read and understand how your stuff works, the problem will still be there)."
Alan Cox "Not sure why anyone thinks this is Linux relevant 8) - other than it will help to make our drivers even faster than the competition if they adopt it. Have a read, but keep a bucket handy". And after smashing the idea of OS independant drivers a bit more, he really gets funny: "So what are you going to do with it. Joysticks?"

On a quick glance, this thing has even more facts going against it's usage:
The worst:
-It seems to demand applications (not only drivers) to link special headers in order to use their infrastructure.
- It's name (WinDriver)
- This sentence on their webpage:
"Use the powerful graphical Wizard (available in Windows only), to create your driver source code. The Wizard will automatically generate make files for both Linux and Windows."

The Uniform Driver Initiative has been working on this problem for a while. Their homepage is here:http://www.projectudi.org/