Re: Signed-off-by and aliases

[Alex Elsayed <eternaleye@gmail.com>]

Sage Weil wrote:

> On Mon, 17 Aug 2015, Alex Elsayed wrote:
>> Loic Dachary wrote:
>> 
>> > Hi Alex,
>> > 
>> > On 17/08/2015 22:19, Alex Elsayed wrote:
>> <snip>
>> >> This is where I see a subtle, but meaningful distinction: Accepting
>> >> from aliases *which have submitted a DCO* means that the person behind
>> >> the alias, even if we don't know their name, has bound themselves to a
>> >> standard ov behavior.
>> >> 
>> >> Accepting from arbitrary aliases does _not_ carry that meaning.
>> > 
>> > Yes. Although we don't do formal background checks, we make sure that
>> > each commit is Signed-off-by: the author as required by
>> > https://github.com/ceph/ceph/blob/master/SubmittingPatches#L22 which is
>> > linked from the
>> > https://github.com/ceph/ceph/blob/master/CONTRIBUTING.rst document that
>> > shows whenever someone submits a pull request.
>> > 
>> > I also believe this is an important distinction and I would feel
>> > uncomfortable if Ceph accepted contributions (aliases or not) that are
>> > not Signed-off one way or the other.
>> 
>> The kernel does something slightly different, in a very careful manner:
>> Signed-off-by says that you have _submitted_ the DCO - as in, you must
>> have, from the same address as you signed off by, emailed the DCO itself
>> to the list.
>> 
>> The S-o-B tag, then, simply says "If you look, you'll find my affirmation
>> of intent to follow the DCO" - it is not, in itself, anything other than
>> a pointer. This prevents people from copypasta'ing the S-o-B line as a
>> magic incantation, without understanding the meaning. (Which the kernel
>> has found _does_ happen _anyway_, but with the "actually submitted a DCO"
>> requirement they can _detect_ that.)
> 
> I can't find any reference to emailing a copy of the DCO to any address in
> SubmittingPatches or elsewhere.  Are you sure this is the case?  See
> 
> https://www.kernel.org/doc/Documentation/SubmittingPatches
> 
> FWIW, LWN's coverage of James's talk on the DCO last year matches my
> understanding:
> 
> https://lwn.net/Articles/592503/
> 
> ...
> 
> As far as aliases go in sign-offs, I don't see that it's an issue.  In my
> opinion the sign-off is more about due diligence than anything else--that
> we have made a good faith effort to ensure that the code is made in
> compliance with the license.
> 
> James contends that it's also more cleaning out offending code when a
> problematic contributor is identified and less about liability for that
> individual, so as long as an alias is used consistently I'm not sure it
> makes a difference.  Lots of people go by names that are not technically
> their legal names, but shortened or anglicized versions of them, but as
> long as they are sufficient to associate the contribution with the
> contributor it serves its purpose.

Hm, you're right. Not sure where I got that, then - weird.