Open Stack

On 11/11/2013 12:08 PM, Mark McLoughlin wrote:
> On Mon, 2013-11-11 at 11:41 -0500, Russell Bryant wrote:
>> On 11/11/2013 10:57 AM, Mark McLoughlin wrote:
>>> Hi Nick,
>>>
>>> On Mon, 2013-11-11 at 15:20 +0100, Nicolas Barcet wrote:
>>>> Dear TC members,
>>>>
>>>> Our companies are actively encouraging our respective customers to have the
>>>> patches they mission us to make be contributed back upstream.  In order to
>>>> encourage this behavior from them and others, it would be nice that if
>>>> could gain some visibility as "sponsors" of the patches in the same way we
>>>> get visibility as "authors" of the patches today.
>>>>
>>>> The goal here is not to provide yet another way to count affiliations of
>>>> direct contributors, nor is it a way to introduce sales pitches in contrib.
>>>>  The only acceptable and appropriate use of the proposal we are making is
>>>> to signal when a patch made by a contributor for another comany than the
>>>> one he is currently employed by.
>>>>
>>>> For example if I work for a company A and write a patch as part of an
>>>> engagement with company B, I would signal that Company B is the sponsor of
>>>> my patch this way, not Company A.  Company B would under current
>>>> circumstances not get any credit for their indirect contribution to our
>>>> code base, while I think it is our intent to encourage them to contribute,
>>>> even indirectly.
>>>>
>>>> To enable this, we are proposing that the commit text of a patch may
>>>> include a
>>>>    sponsored-by: <sponsorname>
>>>> line which could be used by various tools to report on these commits.
>>>>  Sponsored-by should not be used to report on the name of the company the
>>>> contributor is already affiliated to.
>>>
>>> Honestly, I've an immediately negative reaction to the prospect of e.g.
>>>
>>>   Sponsored-By: Red Hat
>>>   Sponsored-By: IBM
>>>
>>> appearing in our commit messages.
>>>
>>> I feel strongly that the project is first and foremost a community of
>>> individuals and we instinctively push as much of corporate backing side
>>> of things outside of the project. We try to spend as little time as
>>> possible talking about our affiliations as possible.
>>>
>>> And, IMHO, the git commit log is particularly sacred ground - almost
>>> above anything else, it is a place for purely technical details.
>>
>> This was exactly my reaction, as well.  I just hadn't been able to come
>> up with a good alternate proposal, yet.
>>
>>> However, I do think we'll be able to figure out some way of making it
>>> easier for tools to track more complex affiliations.
>>>
>>> Our affiliation databases are all keyed off email addresses right now,
>>> so how about if we allowed for encoding affiliation/sponsorship in
>>> addresses? e.g.
>>>
>>>   Author: Mark McLoughlin <markmc+ibm at redhat.com>
>>>
>>> and we could register that address as "work done by Mark on behalf of
>>> IBM" ?
>>
>> That doesn't seem any better to me.  It actually seems more likely to
>> break, since someone could be using an email address with '+' in it for
>> some other reason, right?
> 
> Oh, I'm not saying we parse the "ibm" bit and key off that. Just that we
> can associate affiliation with email address while still allowing people
> to use the same email address for everything.
> 
> A better example - say Robert De Niro works for Nebula but uses his
> gmail address for everything:
> 
>   Author: Robert De Niro <taxidriver at gmail.com>
> 
> but if he did some work sponsored by the NSA, he could do:
> 
>   Author: Robert De Niro <taxidriver+soldmysoul at gmail.com>
> 
> and we'd have the tracking tools associate the first email address with
> Nebula and the second with the NSA.

Oh OK, I get it now.  That seems fine.  It keeps the git log more
pristine, and uses the same thing (email address) that we use already
for building.  There's really no technical work required to implement
this, which is nice.  It would just be a recommended convention.

>> I think it may be worth looking at this from a different angle.  Perhaps
>> we should tone down the focus on company metrics, and perhaps remove
>> them completely from anything we control or have influence over.
> 
> I'm down with that. We've definitely jumped the shark on this.

But Monty makes a good point that if we don't produce this information,
someone else will with potentially questionable accuracy.

It's a tough situation, because we really do need to work to encourage
viewing the project as a community of individuals, at least within the
technical community.

-- 
Russell Bryant