On 12/05/2013 10:15 PM, Christian Boltz wrote:
Hello,
Am Donnerstag, 5. Dezember 2013 schrieb Ancor Gonzalez Sosa:
1. Decision making. As you all know, most final decisions on the technical side rests in Coolo's shoulders. He plays the role of a benevolent dictator, working based on his perception of skills and dedication. Sure he is fair and experienced but still a single human being. And a karma system would change that in which way?
("everybody with >1000 karma points is allowed to submit directly to factory without review" is not the answer, at least I hope so ;-)
With "karma-driven process" I didn't mean automatic. And, of course, I didn't mean removing reviews. Let's try another example with an already working and well balanced karma system (far in the future). Let's say that somebody wants to introduce something in Factory that is kinda "dangerous" because it needs constant care to ensure that does not break stuff in the future if not maintained properly. That is, we need to ensure that the person behind the feature is not going to "fire and forget". Having some real and qualified data about of the previous involvement of that person and its activities in OBS would allow the release manager (or release team, don't assume Coolo is going to be there forever) to know if that person cares about the rules and best practices. Going even further, maybe a mature enough system will allow him to tell that person "according to the information of your profile and to the general guidance, we cannot accept this into Factory unless you find a trustworthy maintainer", which sounds less arbitrary than "I don't trust you".
2. Guidance We lack a clear path from newbie to contributor and then to experienced contributor (like maintainer, reviewer etc). Also, we Maybe because such a path doesn't exist, at least not as _the_ path?
have a wide variety of guidelines which we'd like people to follow better but which aren't hard rules. And you really think rewarding them with a number of points displayed somewhere will help?
Will help to make things more obvious.
People will (hopefully) follow rules that make sense, but I'm quite sure nobody will follow the "jump into the sea _now_" rule just because you offer a karma point for doing that ;-)
Too bad, +100 points for jumping into the sea _now_ was going to be my first proposal :-) Seriously, ask Coolo, the Factory maintainers and the reviewers if people always follow rules that make sense right now. ;-) The common understanding about what are good practices and what are not (I'm always talking about OBS and Factory) are just that: "common understanding". How can the karmafication help? - Making it more obvious and explicit to people who is learning how to do things properly. - Making it more obvious and clear for people who now about the "right" way to do things but simply don't care.
3. Motivation There are areas in openSUSE that do not get the love and attention they deserve in both technical and non-technical terms. People who work on them should be recognized and rewarded. While I agree with the goal - do you really think displaying a numer somewhere is a reward or a motivation?
I expect the final system to be more complex and informative than just "a number somewhere". And yes, it would be a reward or motivation for some people. Maybe not you, maybe not me, but for many people.
As you probably already guessed, I don't care about karma points. It's just a type of paperwork, and if you want that, you can have more fun with an "Antrag auf Erteilung eines Antragsformulars" ("motion for granting a motion form" - and that's just half of the title) If you understand german, I can recommend http://www.youtube.com/watch?v=60cmsLJ--G8 ;-)
How does Karmafication support the factory proposal goals in the above areas? I'd guess "nowhere" ;-)
That was a rhetorical question ;-)
Karma should have an impact in:
1. Decision making. Make contributions visible: credit where it is due. Having a profile page for every contributor in OBS would be useful in decision making at all levels. Is this person a good candidate to become technical reviewer? Should I accept this risky SR to factory? Ah, now things become more interesting.
You just found out that karma points earned by wiki edits say nothing about packaging skills.
In other words: something like "openSUSE karma" (as a total of all areas) is useless.
Of course, it was NEVER the idea.
OTOH, having "OBS karma" at least isn't totally useless (I'm still not sure if I should call it useful ;-) - but even that is too much summarized. Maybe someone earns lots of points with packaging perl modules, but then one day submits a C patch that is totally broken because he rarely writes C. Based on your proposal, you would happily accept it because of the good karma.
Sure we need more than a single "OBS karma" number. Most of the fears in the thread are based on an oversimplification of the proposed idea. We are not proposing to have just a number for measuring all the activity in OBS. We are proposing an effort to find a system that makes sense and which provides a more objective measure than "I feel you are the right person". Look at previous attempts like http://en.opensuse.org/openSUSE:Build_Service_Concept_Trust
Besides that - I'm quite sure that you can answer your questions easily based on a contributor's name and don't need any statistics.
Very experienced people with a lot of knowledge about the project and the community can do it, yes. And it's still totally subjective. In practice, it means depending on the decisions made by a single person that needs full-time dedication to the project and following his very own criteria.
Oh, and there's still...
general rule: if Olaf reports a bug, it is a valid bug. [Olaf Hering while reopening https://bugzilla.novell.com/show_bug.cgi?id=168595]
;-)
In the future, once the system is mature enough, a minimum of karma could even be required to perform some actions. Please forget about this - I understand what your intention is, but it won't work.
Just as an example - maybe someone wants to package "his" software for openSUSE and get it into the distribution, but personally uses $other_distribution. How should this person earn karma points to be "allowed" to submit the package to factory?
The result will be that he won't create the package at all.
Once again, sounds a little "black or white" to me. Even though, let's accept it as a valid example. Let's say that with no karma he can create the package in his own repository, let's say he can also have the package integrated into a devel project. At this point, a lot of people can use the package and the packager is already getting all the benefits from OBS and collaborating with other people. But we already have 6800 packages in Factory so having some criteria for including only things that are trustworthy makes sense to me. But once again, it's just an example (your example ;-) ). Those thresholds are just an idea for a distant future, if we are able to implement a trustworthy system.
2. Guidance By defining tasks and rewards we could 'spell out' a path, or several, from beginner to more experienced contributor. Think for example a number of tasks around your first step to contributing; or tasks related to more advanced hackery like fixing certain type of bugs. It could also be used as a way of define best practices for OBS and encourage people to follow them. I like the idea of having a "how can I help?" list with different levels of difficulty (split by topics like wiki, packaging, ...)
However I wouldn't connect it to karma. ("I'm new here - does that mean I shouldn't do the more difficult stuff?")
Hopefully, already replied above.
3. Motivation Motivate people and visibly reward them for working on things which usually fly below the radar. It would be great to have openSUSE contributors pointing to their OBS profile pages as a reference of their skills and experience, in the same way that most open source developers points to their github page. We could make this recognition more tangible with some special gifts (what about a "I take care of stuff!" t-shirt?). I always like a nice t-shirt ;-) - but again, I'm sure we don't need to count points to know who does the work.
And, as Pascal already pointed out, the better text would maybe be "I tricked the karma system" ;-)
Of course, we never said it would be easy. I already tricked our scores in the bugfix day myself, and it was really easy. We know that implementing a fair system would be a lot of work and will result in a never-ending refinement process. We know that is even risky. The question is: does the community around openSUSE think is worth the effort?
A big benefit of Karmafication over other ways of reaching the same goals is that most alternatives require making rules, commitees and bureaucracy and require much more work. Counter-question: how much time will it take to develop the karma system? (Note: It will _not_ be easy to implement for all areas of openSUSE, and it will at least need rules to define how many points a specific type of contribution earns.)
We can only try to estimate the cost of the implementation in one single point of the infrastructure. Let's say OBS. We will need some months just to have a initial implementation that would be continuously refined in the future.
How often could a commitee meet in that time to choose and announce some "openSUSE heros"?
How much other work with "real" user benefit, for example fixing bugs [1] could we do in this time?
If we agree that Karmafication is worthy in the long term (we haven't yet), it would mean investing manpower now for harvesting bigger "real user benefits" in the future.
Would you like openSUSE to be the first distribution with a karma-driven development process? I'm quite sure you can guess my answer ;-)
:-)
Regards,
Christian Boltz
[1] we have lots of "old" open bugs, and last time I checked, the number of those old bugs was about the amount of bugreports we get for each release. So either we have to ship a totally bug-free release to have time to fix the old bugs, or we need to invest some more time in fixing bugs ;-)
PS: Yes, I know the (random) sig is too long, but it's too good to shorten it. And it's still much shorter than this mail ;-)
Fair enough. I'm trying to learn to summarize... with no success so far. :-) -- Ancor González Sosa openSUSE Team at SUSE Linux GmbH -- To unsubscribe, e-mail: opensuse-project+unsubscribe@opensuse.org To contact the owner, email: opensuse-project+owner@opensuse.org