Update on bug system for OpenJDK (web-discuss)

[David Holmes]

Volker Simonis said the following on 05/31/11 03:03:
> Ok, I see the value of an access controlled system. But we should try
> to make it as easy as possible for potential contributors to submit
> valuable bug reports and patches without the need to become "Group
> Members" with full commit rights.
> 
> As mentioned earlier, this may be done by defining a new class of
> "Contributors", which is explained in the proposed bylaws as follows
> (copied from Andrews mail above):
> 
> "A Contributor is a Participant who has signed the Oracle Contributor
> Agreement (OCA), or who works for an organization that has signed that
> agreement or its equivalent and makes contributions within the scope
> of that work and subject to that agreement. Only a Contributor may
> submit anything larger than a simple patch."
> 
> And these contributors should not only get direct and easy access to
> the bug tracking system, but also to the newly proposed code review
> system and the site which hosts the Webrevs. This would make the
> communication between "Group members" and  "Contributors" a lot easier
> and place the "Contributors" somewhere in between "Group participants"
> (access to he mailing lists only) and "Group Members" (full commit
> rights) as described at http://openjdk.java.net/groups/.

I think there is some confusion between Group roles and Project Roles 
here (perhaps they have changed between the interim rules and the new 
proposed bylaws). Groups don't have repositories and Group members don't 
have commit rights. The basic proposed hierarchy in the draft-bylaws in 
terms of code submission and review is:

- OpenJDK Participant
- OpenJDK Contributor
- Project Author (must be a Contributor)
- Project Committer
- Project Reviewer

then in addition we have plain old users.

So my initial categorization would be:

- Anyone should be able to submit an "incident report".
- Contributors should be able to submit "bug reports".
- Contributors should be able to have read access to the code review 
system (they can always report an issue via email).
- Project Authors and above should have write access to the code review 
system.

Whether "incident reports" and "bug reports" go into two different 
systems depends on how well the systems can handle the distinction both 
in terms of easy access/use for incidents, and isolation of incidents 
from bugs.

David

> What do you think?
> Volker
> 
> On Mon, May 30, 2011 at 11:48 AM, David Holmes <David.Holmes at oracle.com> wrote:
>> Steve,
>>
>> Steve Poole said the following on 05/30/11 19:10:
>>> On 27/05/11 18:47, Georges Saab wrote:
>>>> The 'noise' problem is real and currently Oracle funds a number of people
>>>> who contribute time
>>>> and effort to the thankless task of weeding through the noise.   But
>>>> there has to be a better way
>>>> of handling it than the current one.
>>> I assume that you end up pointing a lot of these requests to some external
>>> documentation?   I'm thinking that if OpenJDK's landing page included
>>> pointers to FAQs , user discussion mailing lists etc that would help reduce
>>> the amount of noise.
>> Such pointers already exist and are ignored by many bug reporters.
>>
>>>> I wonder if part of the issue is the current lack of distinction between
>>>> bug reports from end users,
>>>> developers in general, and developers with 'skin in the game on OpenJDK'?
>>>>
>>>> Perhaps the first step is to have separate places to report issues if you
>>>> are an end user
>>>> versus a Java developer (and making it clear which is which), and then if
>>>> there are still
>>>> an overwhelming number of 'I am a CS student learning Java and could you
>>>> please do
>>>> my homework' coming into the developer bug database, look at how to raise
>>>> the bar?
>>> I'd rather have a system where anyone can raise issues without regard to
>>> any role.  If there is too much "noise"  then steps can be taken  to help
>>> redirect people to the right place.   End users  asking questions - however
>>> naive  or incoherent - are still part of the community.   Having said that I
>>> think we need to be pro-active - see my comment above.
>> We already know that the noise exists and is extreme. Naive questions don't
>> belong in bug reports they belong on mailing lists (not current openJDK
>> ones) or in forums. As a developer who does a lot of bug triage I do not
>> want a system that will suddenly inundate me with bogus reports (the bar is
>> often too low as it is!).
>>
>> So I'm all for role-based access control on the bug system.
>>
>> Regards,
>> David Holmes
>>
>>>>    /GES
>>>>
>>>> On 27 maj 2011, at 08.07, David Herron wrote:
>>>>
>>>>> I have to say that while what Volker said is compelling, I know that
>>>>> David Holmes is speaking the truth. Way-back-when I saw some of that stuff
>>>>> coming in and much of it was more user error or confusion and needing
>>>>> education on proper use of the API rather than it being a bug.
>>>>>
>>>>> The team when we were still at Sun had a system letting them construct
>>>>> semi canned semi customized responses in many cases, and give proper
>>>>> disposition to the entries. But they did have a serious problem in keeping
>>>>> up.
>>>>>
>>>>> But, one can plausibly imagine a crowd sourced system that handled both
>>>>> actual bug reports and the other stuff like "I'm confused about method Y
>>>>> ...".  I always thought those questions and answers were a potential source
>>>>> of a crowd sourced knowledge base of typical confusion and answers.
>>>>>
>>>>> David Herron
>>>>>
>>>>>
>>>>>
>>>>> On May 27, 2011, at 5:21 AM, David Holmes<David.Holmes at oracle.com>
>>>>>  wrote:
>>>>>
>>>>>> Volker Simonis said the following on 05/27/11 20:11:
>>>>>>> I think it shouldn't be only writable by anyone - it should be
>>>>>>> possible to enter bugs which will become visible instantly.
>>>>>>> It wouldn't make sense to make the new system writable only by
>>>>>>> contributors taking into account that the old system at bugs.sun.com
>>>>>>> is already writable by everybody. But the way how bugs.sun.com
>>>>>>> currently handles new entries which are first kept private and then
>>>>>>> only published after a certain amount of time without any notification
>>>>>>> to the submitter is at least insane. That way you can not enter a bug
>>>>>>> and then start a discussion about it because you simply don't have a
>>>>>>> reference to it. Instead you have to manually poll the system to see
>>>>>>> when it finally appears publicly.
>>>>>> The lack of notification may be "insane" but the process isn't insane
>>>>>> once you see the volume of crud that gets submitted as bug reports. The
>>>>>> current process certainly has some issues but you need a first-level
>>>>>> filtering mechanism of some kind.  Most bug reports don't come from
>>>>>> contributors, they come from users.
>>>>>>
>>>>>> So I'd say that contributors should be able to make entries to the bug
>>>>>> system, but users should still submit via some other front-end.
>>>>>>
>>>>>> David Holmes
>>>>>>
>>>>>>> On Thu, May 26, 2011 at 9:06 PM, Michael Klishin
>>>>>>> <michael.s.klishin at gmail.com>  wrote:
>>>>>>>>> mark.reinhold at oracle.com escribió:
>>>>>>>>> I completely agree. The primary users of this bug system are OpenJDK
>>>>>>>>> Contributors. It will be readable by anyone, of course, but I'm not
>>>>>>>>> sure it should be writable by anyone other than Contributors.
>>>>>>>> Mark,
>>>>>>>>
>>>>>>>> I think that if the bugs system will be read-only for "regular
>>>>>>>> developers", they will post their issues
>>>>>>>> on various mailing lists or won't post them anywhere at all, and in
>>>>>>>> the end it will be significantly harder for everyone, including contributors
>>>>>>>> and Oracle employees, to keep track of issues.
>>>>>>>>
>>>>>>>> Given that OpenJDK is a pretty large project with many moving parts,
>>>>>>>> that would be a loss-loss situation for everyone, IMO.
>>>>>>>>
>>>>>>>> My 2¢.--
>>>>>>>> MK
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>>