Some Linux/Open-Source Letdowns Last For Years

[geearf]

there's only one next-gen linux fs - btrfs

btrfsf has been mainlined over a decade ago, how is that next-gen? Is it not current-gen at best?

[oiaohm]

The messed up part is the CDDL has a technicality that allows the source and the compiled binary to be under different licenses so upstream Linux could accept OpenZFS if the kernel added some sort of "if module then CDDL else if builtin then GPLv2" or "if source then CDDL else if compiled then GPLv2" catch-all parameter for CDDL code.

That does not legally work. GPLv2 binary requires you to be able to release all of it source code under a GPLv2 compatible license. Yes CDDL allows you to license the binary under a different license but it does not allow you to license the source code under a different license at all. So GPLv2 does not allowt binary as GPLv2 and have source as not GPLv2 compatible and CDDL does not legally allow release binary as GPLv2 and source as GPLv2 compatible because it mandates source be CDDL.

https://www.mozilla.org/en-US/MPL/2.0/FAQ/#mpl-and-lgpl

https://www.mozilla.org/en-US/MPL/2....g-mpl-and-gpl/

These are really good reads for anyone wanting to combined licenses. Com-binding with GPL is not straight forwards

CDDL is technically MPL 1.1 based that is absolutely GPL incompatible proven in court and CDDL contains the same clauses tested in court. Only way around the MPL 1.1 problem is dual license problem is the patent grant with ZFS is dependant on CDDL. Lets say you dual license OpenZFS as GPLv2 and CDDL the GPLv2 build does not have the patent grant from SUN that is now grandfathered by Oracle so now Oracle can sue you.

There is also great fun here with CDDL you can build your binary as not CDDL one big nasty loop hole that binary also does not have the patent grant in the CDDL license because its no longer CDDL. Yes to dual license OpenZFS as CDDL/GPLv2 would require going to Oracle and getting a new patent grant even your idea hey we can just build a GPLv2 binary because CDDL allows a different license binary to source also requires we need a new patent grant.

Its really fun to say CDDL allows more freedom due to allowing binary to be under a different license so yes this is more freedom problem this freedom is double sided this also allows you ass to be sued for patent infringement as well because the binary is no longer covered by CDDL and its patent grant.

You really don't want to be able to make a binary as a different license to the source code if you can avoid it as there as big pitfalls like disappearing patent grants.

Remember something like apache license stuff can be used in closed source binaries having to give attribution to the apache code in their means you keep the patent grant intact.

Legally in most cases you don't want the right to release the binary as different license to source code. Anyone the OpenZFS case releasing under a non CDDL even a binary is not legally workable anyone doing this is setting themselves up for a whipping as soon as Oracle decide they need more money from their patent list. CDDL option to license the binary under not CDDL only comes workable once the patents over CDDL work have expired.

[elatllat]

Waland/gnome/Ubuntu crashes every few weeks for me.

btrfs is faster than zfs and stratis for me, similar features though.

Wish brands or governments would force OEMs to open drivers at least after 3 years (right to secure/repair).

[oiaohm]

Wish brands or governments would force OEMs to open drivers at least after 3 years (right to secure/repair).

That not going to work. OEM could have been liquidated inside 3 years so ceased to exist so no one to release the information. You really need the right to repair/secure from day 1 the device is released.

Please note it would be possible to build in a physical kill switch/wire to signed firmware that a person would have to cut a particular wire or provided power to a particular pin to disable signing of course if that is done the OEM could have the right to say you modified the device its your problem and you better be sure what you are doing is legal if you do that(a development mode). Or put boot signing keys in a replaceable rom outside the cpu if vendors wanted to.

With firmware vendor releasing signing keys is not going to be that possible.

There are things governments could mandate:
1) OEM have to include means to disable firmware signing or replace signing key if they ship device demanding signed firmware of course if consumer does that OEM is also legally not required to back warranty on device if this feature has been used. This is fairly much a fair trade. Yes being a physical modification it can nicely leave behind physical evidence like blowing a fuse inside the silicon and leaving software detectable item that its been done.
2) Items that serial bind to each other there need to be a documented from day 1 method to repair these devices. With all tools required released at least to government on day 1 to be released the day the OEM no longer provided warranty.
3) same with driver source code and hardware documentation. It need to be released day 1 to government at least so company cannot go bankrupt between the time they release the device and when they need to release this information.

Notice something 3 years down the track heck even one week after the device is released is could still be company gone.

Right to repair and right to secure is really a day 1 problem not years down the track. Lot of ways you need to look at when you buy a device can it be repaired or not if there is a product that can be repaired buy it instead.

[N0va]

But as it stands at the start of 2020, the GeForce GTX 900 series and newer on the open-source NVIDIA drivr is basically garbage

You misspelled 'driver'

[f0rmat]

Michael

Hopefully Xfce 4.18 strikes at the en of 2021 or at least to a better and more predictable cycle than the long and drawn out Xfce 4.14 happenings from a few years ago.

Possible typo - should that be end?

[elatllat]

...Right to repair and right to secure is really a day 1 problem...

Yes source code could be sent to government regulators like Federal Communications Commission, Patent and Trademark Office, etc for later release.

[skeevy420]

That does not legally work. GPLv2 binary requires you to be able to release all of it source code under a GPLv2 compatible license. Yes CDDL allows you to license the binary under a different license but it does not allow you to license the source code under a different license at all. So GPLv2 does not allowt binary as GPLv2 and have source as not GPLv2 compatible and CDDL does not legally allow release binary as GPLv2 and source as GPLv2 compatible because it mandates source be CDDL.

https://www.mozilla.org/en-US/MPL/2.0/FAQ/#mpl-and-lgpl

https://www.mozilla.org/en-US/MPL/2....g-mpl-and-gpl/

These are really good reads for anyone wanting to combined licenses. Com-binding with GPL is not straight forwards

CDDL is technically MPL 1.1 based that is absolutely GPL incompatible proven in court and CDDL contains the same clauses tested in court. Only way around the MPL 1.1 problem is dual license problem is the patent grant with ZFS is dependant on CDDL. Lets say you dual license OpenZFS as GPLv2 and CDDL the GPLv2 build does not have the patent grant from SUN that is now grandfathered by Oracle so now Oracle can sue you.

There is also great fun here with CDDL you can build your binary as not CDDL one big nasty loop hole that binary also does not have the patent grant in the CDDL license because its no longer CDDL. Yes to dual license OpenZFS as CDDL/GPLv2 would require going to Oracle and getting a new patent grant even your idea hey we can just build a GPLv2 binary because CDDL allows a different license binary to source also requires we need a new patent grant.

Its really fun to say CDDL allows more freedom due to allowing binary to be under a different license so yes this is more freedom problem this freedom is double sided this also allows you ass to be sued for patent infringement as well because the binary is no longer covered by CDDL and its patent grant.

You really don't want to be able to make a binary as a different license to the source code if you can avoid it as there as big pitfalls like disappearing patent grants.

Remember something like apache license stuff can be used in closed source binaries having to give attribution to the apache code in their means you keep the patent grant intact.

Legally in most cases you don't want the right to release the binary as different license to source code. Anyone the OpenZFS case releasing under a non CDDL even a binary is not legally workable anyone doing this is setting themselves up for a whipping as soon as Oracle decide they need more money from their patent list. CDDL option to license the binary under not CDDL only comes workable once the patents over CDDL work have expired.

And that is what this really comes down to. The IP and Patent rights, Copyleft, and Derived Works. They both handle that differently. That's it.

I did a lot of reading last night. What's kind of funny is that, technically (especially when reading Stallman and using his POV), all that BSD and MIT code in the kernel should be GPL'd due to the GPLv2's viral nature. They're just granted exceptions so it's all good. Exceptions so it's all good. Heh.

Here's the thing though -- if OpenZFS or any other module is considered a derived work of the kernel, how come Alien Isolation and every Linux native game on Steam isn't a derived work of Glibc? System library exceptions; that's why. It wouldn't be possible to run a modern Linux system without system library exceptions. The GPL is simply too viral.

I love how "system library exceptions" are given with Glibc in regards to linking. But how is that any different than a kernel module?

Serious question. Time for thought experiments.

How is a kernel module any different? In the simplest of terms, a bit Layman perhaps (but also how you'd explain it to a jury), a kernel module is a kernel's library. A module is to the kernel what a library is to libc. Like a library is to a system, it isn't necessary for the kernel to function and is only there if the user puts it there and wants it ran. But if libraries are allowed special linking exceptions, how come kernel modules aren't granted one either?

Does that mean modules should be granted a de-facto special exception?

Are Nvidia and AMD not being sued ever for a closed source kernel modules the case studies we need to allow a special module exception in an updated GPLv2? AMD had Catalyst and still has AMDGPU-Pro. Nvidia has their driver. They both clearly violate the spirit and meaning of open source by using closed source modules linked against a GPLv2'd kernel....but they've never once been challenged as far as I know.

If closed source, obvious GPL violators aren't being sued over their kernel modules, then open source and trying to play ball OpenZFS is the obvious winner in a court case. 20+ years of unchallenged GPL violations is a pretty damn good defense. Basically, if OpenZFS as a module is violating the GPL, that means that AMD and Nvidia are violators too. There's no way around that. Suing over OpenZFS means suing AMD and Nvidia combined.

If OpenZFS is derived, so are Catalyst, AMDGPU-Pro, and Nvidia's driver. If OpenZFS is violating the license, so are...well...y'all know who...

I think that the GPLv2 needs to be updated to address libraries, modules, and to clearly define what special exceptions mean and where they're given. Either that or I've shown that closed sourced and/or GPL-incompatible licensed modules are acceptable in Linux by Linus and the LFS.

[f0rmat]

Yes source code could be sent to government regulators like Federal Communications Commission, Patent and Trademark Office, etc for later release.

Given that the FCC has allowed DRM, they are probably the last people I would trust. You cannot repair a John Deere tractor console because you might use it to play "stolen" music? Give me a break.

And as far as the US is concerned, patents are somewhat of a joke. There are so many holes and walls in the US patent laws that the only people who benefit are the ones who sell their patents to large corporations or those that are large corporations. Research the difference between the "weed-eater" original patent and the patent for the push button ratchet. The former did not get anything for it - the major corporations "cracked" the patent legally and in accordance with US patent laws. The second sold his patent (and made money) to Sears and Roebuck - and when the patent expired, everybody copied it. Just like everybody copied Fein with the Fein multi-tool. Only Fein had it before, now every serious tool manufacturer makes one. (The multitool is an electrical power tool that you can change the heads on to be a hammer, a drill/driver, a jig-saw, etc. - I have one. They are very useful.)

Ok...I am showing my redneck here.

[Vistaus]

I am on Debian. How is the Deepin-on-Wayland experience like?

It's okay. It more or less works, but there's still a lot of work to be done (so basically in the same ballpark as KDE). But as I said: they're working really hard on it, so expect to see many improvements this year

Currently I can use Plasma on Wayland with Nouveau if I disable threaded GL, but there is a chance of a random lockup. Is there any way to disable threaded GL in Deepin Wayland?

My setup is Intel-only, so I don't know anything about Nouveau. Sorry.