Re: [gentoo-project] rfc: copyright attribution clarifications

[Sarah White <kuzetsa@poindexter.ovh>]

[-- Attachment #1.1.1: Type: text/plain, Size: 4803 bytes --]

On 11/13/2018 09:46 PM, William Hubbs wrote:
> On Tue, Nov 13, 2018 at 06:17:17PM -0800, Rich Freeman wrote:
>> On Tue, Nov 13, 2018 at 10:32 AM William Hubbs <williamh@gentoo.org> wrote:
>>>
>>> Since we do not do copyright assignment any more and the glep allows for
>>> traditional attribution, if some entity
>>> (company, person etc) has a desire for a copyright notice in
>>> their work, the case for not allowing this is very weak at best, so we will
>>> end up with more and more ebuilds that want to use traditional copyright
>>> attribution, and once an ebuild is switched over, it is problematic to
>>> switch back.
>>
>> So, the purpose of allowing specific copyright holders to be named was
>> to cover cases where we're forking foreign code, not to basically
>> introduce a variant on the BSD advertising clause.  IMO people who are
>> only willing to contribute FOSS if their name gets put in a prominent
>> location might do better to contribute elsewhere.
> 
> Do you feel this way about corporations as well? Do you think the Linux
> kernel maintainers should go and rip out all copyright notices other
> than Linus Torvalds and maybe the Linux Foundation?
> 

perceived similarity to any version of the BSD advertising clause is a
misunderstanding. copyright is not a license or license restriction:
licensing is not copyright. holding a copyright, and/or having a
copyright notice doesn't bring back the controversial 4-clause BSD
license, which itself is an artifact from the 90s

multiline (standard form) copyright attribution doesn't have anything to
do with licensing, and only serves to strengthen copyleft due to the
presence of additional copyright notices which clearly lay out a list of
entities / people with a stake in protecting the interests of an
opensource project remaining FOSS/Libre. gentoo's license policy already
requires FOSS/Libre licenses and correctly using copyright law for
copyleft purposes makes everything work correctly when it's used correctly.

SPDX-style license blocks have a well-defined layout
(I'm a fan / several linux kernel developers are too)

... and SPDX displays the copyright notice in a way which is
fully compatible with, and improves transparency for copyleft

>>
>>>
>>> As you can see from my example, line length will quickly become
>>> problematic in this format because all lines in in-tree ebuilds are
>>> supposed to be under 80 characters.
>>
>> Indeed, this is tone of he problems with allowing people to spam the
>> copyright notice.  It is basically the advertising clause in a
>> different place.
>>
>>>
>>> It is also problematic because the relationship between the years and
>>> contributors becomes unclear unless we allow ranges and single years in
>>> the copyright notice, which would lead to something like this:
>>>
>>> # Copyright <years1>, <years2>, <years3>, ... <yearsn+1>  [contributor1,] [contributor2,] [contributor3,] ...  [contributorn] and others
>>
>> The purpose of a copyright notice is to declare that the file is
>> copyrighted, and that is it.
>>
>> It isn't a comprehensive list of everybody who holds a copyright on the file.
>>
>> It isn't a revision history.
>>
>> There is no need to list various mixes of years and authors.  Just
>> list the first and last year, and whatever copyright holders are
>> necessary.
>>
>>> Multiple-lines would be much easier to maintain, and
>>> there is no cost performance wise for them.
>>
>> Except for spam in our files.
> 
> And how does that affect performance?
> 

it shouldn't. most interpreted languages have sensible handling for
comments / JIT compilation, and for compiled languages there's normally
zero runtime penalty for comment blocks of any kind.

even if a QA tool has a bottleneck when scanning comments, there's no
reason to believe this is a mission-critical failure and the performance
bottleneck will slow down development.

>> Heck, repoman complains if you stick two newlines in a row in the
>> file, and now we basically want to add a revision history to the file?
> 
> No, a revision history comes from vcs.
> 

yep

>>
>> Just say no.  Fit it on one line.
>>
>> But, if you had to have multiple lines, then just wrap the existing
>> notice.  Don't turn it into some kind of revision history.  Just list
>> one year range and whatever list of entities you feel compelled to
>> list.  That is the proper way to do a notice.
> 
> No sir, it isn't.
> 
> Look anywhere outside the Gentoo tree. For that matter, take the Linux
> kernel, or even in the systemd source, there are several places with
> multiple copyright notices in them.

indeed. it's done in a sensible way too (see comments above)

> William
> 

-- kuza

[-- Attachment #1.1.2: 0xBE4D4EFD05297C67.asc --]
[-- Type: application/pgp-keys, Size: 30819 bytes --]

[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 833 bytes --]