Request for comments: New Bugzilla-based contribution process

[Mark Reinhold]

> Date: Thu, 19 Feb 2009 17:05:46 -0600
> From: bradford.wetmore at sun.com

> Comments on the new page first, then a response on the Mark/Neal exchange:

Thanks for the detailed comments.

> Section 1.
> 
> In the second para, there should be a link to the mail.o.j.n page and/or
> the group pages so that people know where to go to find other people
> working in the same areas.

Good idea.  Done.

> Also, add something like the following at the end of this section:
> 
>      Some groups have already generated lists of "starter bugs" that
>      might contain useful ideas to get started.

Hmm; since this hasn't been done uniformly, and likely won't be soon,
I'm a bit reluctant to advertise it.

> Section 3.
> 
> People are still signing up for accounts and registering to watch
> specific product categories and components, so there should be a bit in
> here about "Announce your changes to the appropriate group's mailing
> list and request a sponsor" so that it doesn't get missed.  I'd suggest
> putting it at the end of this section.

I'm not sure this is such a good idea.  The whole point of using Bugzilla
for contributions is to avoid relying upon people reading e-mail to see
that a contribution has come in.  I'd rather not encourage contributors
to send such e-mails, which in the worst case could be perceived by
potential sponsors as little more than spam.  We already intend to
review the incoming sponsorship requests on a regular basis and ping
likely sponsors; that should be sufficient.

> Section 5.
> 
>      but rather to accept only high-quality contributions.
> 
> I would suggest adding an extra sentence here to remind folks that their
> code could be used by millions of users, so it simply has to work.  It's
> an awesome, but fulfilling responsibility.

Good point.  Done.

>> Mark wrote:
>>> Neal wrote:
>>> Is there any mechanism for contributors or committers to "claim" a bug -
>>> that is, notify others that they're working on it - other than
>>> email?  Given
>>> the contribution process you outlined, the main mechanism seems to be
>>> creating a bugzilla bug,
>>
>> As stated this process doesn't really have a "claim" mechanism, since
>> presumably you'd like to claim a bug before doing a lot of work on it.
>> Perhaps the process should suggest that you create a Bugzilla entry first
>> in order to stake a claim, and then attach the patch and request a
>> sponsor once your work is complete.  If two or more developers want to
>> argue about the best way to fix a bug then they can do that in Bugzilla,
>> where a record will be kept for all to see.
> 
> Just a reminder, the majority of bugs will eventually be handled
> directly in Bugzilla,

Right.

>			but for now we're just using Bugzilla for
> accepting contributions for bugs that already have bugids in
> Bugtraq/Bugster (Sun's internal bug tracking system).

Well, not exactly.  Contributors are also welcome to submit patches for
bugs or RFEs that do not already have corresponding Sun bug ids.

> So how does this work with the SUNBUG (TRACKEDINBUGTRAQ) field of Bugzilla?
> 
> If I understood this proposal right, say we have:
> 
> 1)  Internal user has previously filed Bugtraq id 6000000.
> 2)  External user(s) wants to fix it, so they stake their claim in
> Bugzilla (Bugzilla id 100000), listing 6000000 in the summary or
> description field.
> 3)  Several users propose fixes.  Finally one is accepted as "The Fix"
> and is attached to the Bugzilla bug id.
> 4)  Sponsor accepts fix, integrates fix into JDK at 6000000, closes the
> bugzilla bug as "SUNBUG", updates the Whiteboard with "sunbug=6000000".
> 
> Does that correspond to what you're thinking?

Not quite.  The SUNBUG resolution value is intended for the case of a
Bugzilla bug filed against a component whose bugs are only being tracked
in Sun's internal system.  The "sunbug=xxxxxxx" whiteboard entry is more
generally useful.  I think your scenario would work better as:

  1) Internal user has previously filed Bugtraq id 6000000.
  2) External user(s) wants to fix it, so they stake their claim in
     Bugzilla (Bugzilla id 100000), adding "sunbug=6000000" to the
     bug's whiteboard.
  3) Several users propose fixes.  Finally one is accepted as "The Fix"
     and is attached to the Bugzilla bug.
  4) Sponsor evaluates fix, integrates it into the JDK at 6000000, and
     closes the bug as FIXDELIVERED/FIXED.

This does point out the need for a corresponding "How to sponsor" page,
but let's try to get the "how to contribute" page done first.

I've posted a revised draft [1] incorporating the above changes.

On further thought I'm not going to add a "claim your bug" step just yet,
the possibility of which I mentioned last night in response to Neal's
comment.  Let's see how things go and to what extent collisions actually
occur before we add complexity to the process.

- Mark


[1] http://cr.openjdk.java.net/~mr/new-contrib.html