curious about code inclusion in the kernel...

nightfrost · 2005-12-09 13:04:34

Say there's a driver needed for something with closed specifications. For the sake of example, let's consider ntfs. The only thing to do for the kernel devs, is to get it backward engineered, and that's precisely what they're doing. Let's say they finally manage to get a read/write code for ntfs, and have that included in the kernel. This code will be under the GPL, and there's nothing illegal about it; it's most definitely not code taken from the windows drivers. This is all good and dandy.

Now, consider e.g. the ZFS filesystem which is open source. It is, however, under another license and as such it's not compatible with GPL and can never be included in the mainline kernel. What opportunities are there if the kernel devs would want to have it included? I mean, it would be very strange indeed if they started backward engineering drivers for ZFS, if the source code is open for all to see. But they can't use the source, because it's not GPL.

Does this make any attempt of inclusion of non-GPL opensource drivers in the kernel fruitless?

I'm interested in this as a general problem, not particularly related to ntfs and/or zfs.

stavrosg · 2005-12-09 14:38:11

The licenses problem is valid indeed;
In a few words: It can't be done.
At least neither from the kernel developers, nor from any distributor. I don't know the license zfs is released under, but theoritically, you could mix and match software under various open source licenses, but in your own pc only. You will never be able to distribute the resulted sources and/or binary from that proccess, at least legally.

nightfrost · 2005-12-09 14:50:14

Hmm... this is very problematic indeed. It could mean that although there is a fair chance that there will be a fully functional ntfs-driver in the kernel soon enough, there will never be a zfs-driver. The paradox is that Microsoft's hiding of specifications actually implies the possibility of having the relevant driver in the kernel, whereas sun's readily opening up of their specifications prevents this. Very interesting.

Btw, zfs is licenced under CDDL. There are some interesting discussions over at lkml on the topic. Someone claimed that the incompatibility was a design goal on sun's part.

stavrosg · 2005-12-09 14:55:58

nightfrost wrote:

Btw, zfs is licenced under CDDL. There are some interesting discussions over at lkml on the topic. Someone claimed that the incompatibility was a design goal on sun's part.

I heard that, too. But I don't know anyhting more.
The fact is that GPL is incompatible to almost any other license out there (by design)

syamajala · 2005-12-09 15:09:26

so maybe its a design goal on the fsf's part ;-p

Dusty · 2005-12-09 16:45:11

Hmmm.... what about the BSD style licenses? IIRC, they say you can redistribute modified versions under whatever license you want. So that means you could modify a BSD app and release it as GPL?

Dusty

nightfrost · 2005-12-09 17:01:22

I think that's correct. It's nor surprising really that the GPL is so strict. The whole idea is to strictly preserve freedom, so compatible licenses must be compatible on the terms of GPL.

However, there must be some reason for sun not choosing to use an already existing license model. The sun guys must have been aware of the fact that their license is incompatible with GPL, which is after all the most commonly used open license. The question is why.

Dusty · 2005-12-09 17:48:37

Sun is still scared of open source, I think.... They seem to think if they release GPL Microsoft or somebody might steal everything on them and they'll lose control. They are getting better though.

Dusty

nightfrost · 2005-12-09 18:18:30

Fully agree. But I can't stop thinking that this doesn't make sense. Why can we have ntfs-drivers in the kernel but not zfs-drivers?

Assume a bunch of people would swear to never take a look at the zfs-source, and instead start hacking on a driver by means of backward engineering and publish their results under the GPL. How would that be? (I'm no programmer - how much resemblance would that code bear to sun's official code?).

I'm thinking; what if MS opens up alot their source (ntfs, .doc, whatever) under a very incompatible license. Would that effectively prevent GPL-apps to implement their stuff? Wouldn't that be very strategical on MS's side?

Euphoric Nightmare · 2005-12-13 14:39:26

I think the Sun license also restricts reverse engineering. I took a look at the license for OpenSolaris, and it had section that stated you can not reverse engineer it.

I think Sun is currently on their, 'We hate GPL' kick. Which is in improvement over their 'We hate Linux' view.

I also think they are planning to focus more on developing for linux, but they don't ever seem to make up their mind. I was very surprised that they released the source for some of their software.

nightfrost · 2005-12-13 14:55:05

bu reverse engineering can't be forbidden, can it? That doesn't make sense.

Gullible Jones · 2005-12-13 20:52:20

Yes... Couldn't be that hard to make a different ZFS driver for the Linux kernel.

(Also, isn't there already a project to incorporate ZFS into Linux?)

Euphoric Nightmare · 2005-12-15 00:04:58

nightfrost wrote:

bu reverse engineering can't be forbidden, can it? That doesn't make sense.

I agree, but thats in the license for a good amount of different software. You can always read over the license for OpenSolaris and see for yourself.

But then again licenses can't always be enforced, because of that reason.

I remember when microsoft said that they were going to release some source and then say if you modify the code, you have to put in the code that microsoft wrote it originally.

Euphoric Nightmare · 2005-12-15 00:06:22

Gullible Jones wrote:

Yes... Couldn't be that hard to make a different ZFS driver for the Linux kernel.
(Also, isn't there already a project to incorporate ZFS into Linux?)

Yes, but I don't think that is violating any licensing as long as you do not distribute it in such a way.

Gullible Jones · 2005-12-15 00:10:54

And what's to keep them from making their own GPLed driver, as they did with NTFS?

Euphoric Nightmare · 2005-12-15 02:08:04

I think its the fact that the source has already been released, and its not as 'necessary' as NTFS for some people.

But I do believe the Sun is releasing their DRI stuff as GPL, and I think that they will loosen up on licensing after a while

Arch Linux

#1 2005-12-09 13:04:34

curious about code inclusion in the kernel...

#2 2005-12-09 14:38:11

Re: curious about code inclusion in the kernel...

#3 2005-12-09 14:50:14

Re: curious about code inclusion in the kernel...

#4 2005-12-09 14:55:58

Re: curious about code inclusion in the kernel...

#5 2005-12-09 15:09:26

Re: curious about code inclusion in the kernel...

#6 2005-12-09 16:45:11

Re: curious about code inclusion in the kernel...

#7 2005-12-09 17:01:22

Re: curious about code inclusion in the kernel...

#8 2005-12-09 17:48:37

Re: curious about code inclusion in the kernel...

#9 2005-12-09 18:18:30

Re: curious about code inclusion in the kernel...

#10 2005-12-13 14:39:26

Re: curious about code inclusion in the kernel...

#11 2005-12-13 14:55:05

Re: curious about code inclusion in the kernel...

#12 2005-12-13 20:52:20

Re: curious about code inclusion in the kernel...

#13 2005-12-15 00:04:58

Re: curious about code inclusion in the kernel...

#14 2005-12-15 00:06:22

Re: curious about code inclusion in the kernel...

#15 2005-12-15 00:10:54

Re: curious about code inclusion in the kernel...

#16 2005-12-15 02:08:04

Re: curious about code inclusion in the kernel...

Board footer