Re: [petsc-dev] NVIDIA licensing

2021-09-07 Thread Barry Smith


  There is no way to do do it consistently for all contributors, it would be a 
big mess plus it would miss all the history.

  If NVIDIA requires it to contribute then I see no harm in having those two 
lines in certain source files. It is not tracking provenance in any way, it is 
just a way to allow NVIDIA to check its check marks. 

  Barry



> On Sep 6, 2021, at 11:23 PM, Satish Balay  wrote:
> 
>>> 2) NVIDIA's lawyers have some thoughts that we will need to address 
> 
> When I saw this is, my thought was: ANL lawyers would be involved here [as 
> generally employees couldn't respond to legal stuff]
> 
>>  git grep Copyright locates some files that are copyrighted by other groups 
>> (but fit within the BSD license) so I think it is fine to include the NVIDIA 
>> copyright information in appropriate files. 
> 
> My impression here was we are bundling some external stuff [yaml, 
> abi-compliance-checker, khash, valgrid] - so in some sense they are still 
> external stuff - not petsc native stuff - where we are to retain their 
> copyright - as per terms of use.
> 
> So would 'SPDX' be similar module? It wasn't clear to me if - say any current 
> petsc sources say src/ksp/pc/interface/pcregis.c gets updated as part of this 
> work - would it also need this copyright update? Perhaps thats ok?
> 
> And if this a policy (for current petsc sources) - we should have to do this 
> for all contributions to be consistent?
> 
> Satish
> 
> On Mon, 6 Sep 2021, Barry Smith wrote:
> 
>> 
>>  git grep Copyright locates some files that are copyrighted by other groups 
>> (but fit within the BSD license) so I think it is fine to include the NVIDIA 
>> copyright information in appropriate files. 
>> 
>>  We should also add some text to LICENSE indicating certain files have 
>> additional copyrights indicated at the top of the individual file.
>> 
>>  Barry
>> 
>> 
>>> On Sep 3, 2021, at 7:48 AM, Mark Adams  wrote:
>>> 
>>> All,
>>> 
>>> Matthew (cc'ed) and I are going to start working on an AMGx interface in 
>>> PETSc, PCAMGX, and I suggested that he may want to think about moving the 
>>> core AMG PC code into PETSC as a built-in PC.
>>> 
>>> 1) We would have to decide of we want the maintenance burden of moving the 
>>> whole thing in, but we will probably start with a 3rd party library unless 
>>> this move starts looking like a really good idea. We have not started to 
>>> really dig into this.
>>> 
>>> 2) NVIDIA's lawyers have some thoughts that we will need to address even 
>>> with a 3rd party library (I know Hypre has this license notification and 
>>> LLNL seems happy with what we did).
>>> 
>>> First, Mathew says that this process "can open the door to us helping with 
>>> PETSc more generally."
>>> 
>>> I will just quote the rest of Mathews' message:
>>> 
>>> Thanks,
>>> Mark
>>> 
>>> I can see you don't add license notifications to the headers of the source 
>>> files, but this tends to be the approach we take for copyright. I'm also 
>>> not sure there is a very good alternative?
>>> 
>>> Instead of the monolithic copying of the full BSD-2 license with the NVIDIA 
>>> copyright (which I appreciate might not be desirable), would it be a 
>>> suitable compromise if we used the SPDX identifier, which looks like:
>>> 
>>> // Copyright (c) 1991-2021, NVIDIA. All rights reserved.
>>> // Copyright (c) 1991-2021, UChicago Argonne, LLC and the PETSc Development 
>>> Team. All rights reserved.
>>> // SPDX-License-Identifier: BSD-2-clause
>>> 
>>> This would be for the main source files that we make any significant 
>>> contributions to or author. I would be interested to hear your thoughts / 
>>> ideas on this.
>>> 
>> 
>> 
> 



Re: [petsc-dev] NVIDIA licensing

2021-09-06 Thread Satish Balay via petsc-dev
> > 2) NVIDIA's lawyers have some thoughts that we will need to address 

When I saw this is, my thought was: ANL lawyers would be involved here [as 
generally employees couldn't respond to legal stuff]

>   git grep Copyright locates some files that are copyrighted by other groups 
> (but fit within the BSD license) so I think it is fine to include the NVIDIA 
> copyright information in appropriate files. 

My impression here was we are bundling some external stuff [yaml, 
abi-compliance-checker, khash, valgrid] - so in some sense they are still 
external stuff - not petsc native stuff - where we are to retain their 
copyright - as per terms of use.

So would 'SPDX' be similar module? It wasn't clear to me if - say any current 
petsc sources say src/ksp/pc/interface/pcregis.c gets updated as part of this 
work - would it also need this copyright update? Perhaps thats ok?

And if this a policy (for current petsc sources) - we should have to do this 
for all contributions to be consistent?

Satish

On Mon, 6 Sep 2021, Barry Smith wrote:

> 
>   git grep Copyright locates some files that are copyrighted by other groups 
> (but fit within the BSD license) so I think it is fine to include the NVIDIA 
> copyright information in appropriate files. 
> 
>   We should also add some text to LICENSE indicating certain files have 
> additional copyrights indicated at the top of the individual file.
> 
>   Barry
>  
> 
> > On Sep 3, 2021, at 7:48 AM, Mark Adams  wrote:
> > 
> > All,
> > 
> > Matthew (cc'ed) and I are going to start working on an AMGx interface in 
> > PETSc, PCAMGX, and I suggested that he may want to think about moving the 
> > core AMG PC code into PETSC as a built-in PC.
> > 
> > 1) We would have to decide of we want the maintenance burden of moving the 
> > whole thing in, but we will probably start with a 3rd party library unless 
> > this move starts looking like a really good idea. We have not started to 
> > really dig into this.
> > 
> > 2) NVIDIA's lawyers have some thoughts that we will need to address even 
> > with a 3rd party library (I know Hypre has this license notification and 
> > LLNL seems happy with what we did).
> > 
> > First, Mathew says that this process "can open the door to us helping with 
> > PETSc more generally."
> > 
> > I will just quote the rest of Mathews' message:
> > 
> > Thanks,
> > Mark
> > 
> > I can see you don't add license notifications to the headers of the source 
> > files, but this tends to be the approach we take for copyright. I'm also 
> > not sure there is a very good alternative?
> > 
> > Instead of the monolithic copying of the full BSD-2 license with the NVIDIA 
> > copyright (which I appreciate might not be desirable), would it be a 
> > suitable compromise if we used the SPDX identifier, which looks like:
> > 
> > // Copyright (c) 1991-2021, NVIDIA. All rights reserved.
> > // Copyright (c) 1991-2021, UChicago Argonne, LLC and the PETSc Development 
> > Team. All rights reserved.
> > // SPDX-License-Identifier: BSD-2-clause
> > 
> > This would be for the main source files that we make any significant 
> > contributions to or author. I would be interested to hear your thoughts / 
> > ideas on this.
> > 
> 
> 



Re: [petsc-dev] NVIDIA licensing

2021-09-06 Thread Jed Brown

On Mon, Sep 6, 2021, at 7:16 PM, Barry Smith wrote:
> 
> *  We should also add some text to LICENSE indicating certain files have 
> additional copyrights indicated at the top of the individual file.*
> **

Copyright is held by hundreds of individuals and organizations. I'm not a fan 
of having it in each file as though that were an effective way to track 
provenance. I'm not saying we refuse to put it there for Nvidia, but we should 
make it clear that this is not the expected way to track provenance.

Re: [petsc-dev] NVIDIA licensing

2021-09-06 Thread Barry Smith

  git grep Copyright locates some files that are copyrighted by other groups 
(but fit within the BSD license) so I think it is fine to include the NVIDIA 
copyright information in appropriate files. 

  We should also add some text to LICENSE indicating certain files have 
additional copyrights indicated at the top of the individual file.

  Barry
 

> On Sep 3, 2021, at 7:48 AM, Mark Adams  wrote:
> 
> All,
> 
> Matthew (cc'ed) and I are going to start working on an AMGx interface in 
> PETSc, PCAMGX, and I suggested that he may want to think about moving the 
> core AMG PC code into PETSC as a built-in PC.
> 
> 1) We would have to decide of we want the maintenance burden of moving the 
> whole thing in, but we will probably start with a 3rd party library unless 
> this move starts looking like a really good idea. We have not started to 
> really dig into this.
> 
> 2) NVIDIA's lawyers have some thoughts that we will need to address even with 
> a 3rd party library (I know Hypre has this license notification and LLNL 
> seems happy with what we did).
> 
> First, Mathew says that this process "can open the door to us helping with 
> PETSc more generally."
> 
> I will just quote the rest of Mathews' message:
> 
> Thanks,
> Mark
> 
> I can see you don't add license notifications to the headers of the source 
> files, but this tends to be the approach we take for copyright. I'm also not 
> sure there is a very good alternative?
> 
> Instead of the monolithic copying of the full BSD-2 license with the NVIDIA 
> copyright (which I appreciate might not be desirable), would it be a suitable 
> compromise if we used the SPDX identifier, which looks like:
> 
> // Copyright (c) 1991-2021, NVIDIA. All rights reserved.
> // Copyright (c) 1991-2021, UChicago Argonne, LLC and the PETSc Development 
> Team. All rights reserved.
> // SPDX-License-Identifier: BSD-2-clause
> 
> This would be for the main source files that we make any significant 
> contributions to or author. I would be interested to hear your thoughts / 
> ideas on this.
> 



[petsc-dev] NVIDIA licensing

2021-09-03 Thread Mark Adams
All,

Matthew (cc'ed) and I are going to start working on an AMGx interface in
PETSc, PCAMGX, and I suggested that he may want to think about moving the
core AMG PC code into PETSC as a built-in PC.

1) We would have to decide of we want the maintenance burden of moving the
whole thing in, but we will probably start with a 3rd party library unless
this move starts looking like a really good idea. We have not started to
really dig into this.

2) NVIDIA's lawyers have some thoughts that we will need to address even
with a 3rd party library (I know Hypre has this license notification and
LLNL seems happy with what we did).

First, Mathew says that this process "can open the door to us helping with
PETSc more generally."

I will just quote the rest of Mathews' message:

Thanks,
Mark


*I can see you don't add license notifications to the headers of the source
files, but this tends to be the approach we take for copyright. I'm also
not sure there is a very good alternative?*

*Instead of the monolithic copying of the full BSD-2 license with the
NVIDIA copyright (which I appreciate might not be desirable), would it be a
suitable compromise if we used the SPDX identifier, which looks like:*


*// Copyright (c) 1991-2021, NVIDIA. All rights reserved.*
*// Copyright (c) 1991-2021, UChicago Argonne, LLC and the PETSc
Development Team. All rights reserved.*

*// SPDX-License-Identifier: BSD-2-clause*

*This would be for the main source files that we make any significant
contributions to or author. I would be interested to hear your thoughts /
ideas on this.*