river-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Craig L Russell <Craig.Russ...@Sun.COM>
Subject Re: new site added to SVN and published
Date Wed, 03 Jan 2007 19:19:27 GMT

On Jan 3, 2007, at 9:25 AM, Geir Magnusson Jr. wrote:

> On Jan 3, 2007, at 10:28 AM, Mark Brouwer wrote:
>> Geir Magnusson Jr. wrote:
>>>> I disagree. For cheiron.org I do most of the work myself, but
>>>> nevertheless anything substantial is going through JIRA if it  
>>>> was only
>>>> to keep track of things done for the release notes. Especially  
>>>> when you
>>>> already have a stable codebase gradually improving and getting
>>>> bug-fixed. But I admit with complete new projects I tend to be less
>>>> bureaucratic.
>>> Do you mean
>>> a) you make all the changes, make a patch, submit it to JIRA,  
>>> download the patch, apply the patch to a fresh tree, and then commit
>> Not exactly. I work either on the main codeline or in some version
>> branch, but not before there is an JIRA issue created for the  
>> task, bug
>> or improvement at hand. The JIRA issue is sent to the developer  
>> mailing
>> list (no seperate list for generated artifacts) so other interested
>> people can see whether they think the issue is relevant or not.
> I see
>> When I start working on it, JIRA sends a message to the dev list so
>> people know I'm about to start coding.

JIRA is nice in that any change of status of an issue gets sent to  
the distribution. It's a matter of discipline that committers  
actually use JIRA to "start an issue" so the notices get sent.

>> After I check in my code
>> directly into the repository I resolve the issue which leads to a  
>> post
>> from JIRA to the dev list. In case of commit before review I  
>> suggest one
>> waits a while with resolving the issue.

Both JIRA and svn will generate notices to the rest of the team.
>> No patches are submitted to JIRA, for those that have no commit  
>> rights I
>> also think that sending the patch to the dev list is the best  
>> thing to
>> do. No need to attach that one to the isse. When the patch is  
>> accepted
>> and committed by one of the committers the issue can be resolved.
> That's one thing you may want to modify - ask that patches are  
> posted to the JIRA, as it requires the submitter to explicitly  
> acknowledge that the stuff is contributed under the terms of the  
> Apache License.  Keeps any questions about what's a contribution to  
> a minimum...

This is a very important part of the Apache Way. Contributors really  
should sign CLAs which legally binds contributions. The tick box on  
JIRA attachments also reminds folks that the patch is in fact a  
>> I realize this won't work exactly the same way as for the projects  
>> I'm
>> working on, but still hope to see maybe a somewhat more formal way of
>> working or at least for some parts of the project.
> It's up to you guys.
>>> b) note changes in JIRA afterwards
>> Or during the process of receiving patches, discussing and  
>> committing them.
>>> c) something else?
>>> I think b) is fine, if you are simply looking for a change  
>>> tracking tool, although if you build a decent commit message  
>>> discipline into the community, you can also just ask SVN what  
>>> changed between releases...  and that should be far more accurate.
>> Often it requires multiple changes to the code for the same issue  
>> before
>> everybody is happy (I think/hope this will be a very picky  
>> community :-)

JIRA works well for this case as well. I like to have code uploaded  
to JIRA instead of scouring emails for patches. JIRA allows you to  
replace a patch with "the latest" code in progress, and it's easy for  
reviewers to apply the patch and comment, iterating on the proposed  
fix before svn commit.

>> and I admit I'm an infant with SVN (Perforce addict) but wouldn't a
>> report generated by SVN contain a lot of clutter in that case.  
>> Also for
>> release notes I really like the issue description instead of the
>> comments people in general use for their commit.

I agree. Once an issue has been beaten to death on JIRA, the commit  
message can be relatively brief. commit -m "RIVER-123 added  
snarfgargle to mediation requests" src/java

>> If I recall correctly the Jini team didn't write a single character
>> before there was an issue in their tracking system. I adopted this  
>> style
>> and found it worth while on a rather stable codebase. I'm curious to
>> know what the others think of this.

I like this approach as well. I've seen both discuss on the dev alias  
and discuss on the JIRA issue work well, so this is a matter for the  
team to work on. It might not even be necessary to formalize where  
the discussion takes place. Some people like having it all on the  
JIRA issue; some prefer to follow threaded discussions on aliases.

But basically having a JIRA to track all code changes is a statement  
of formality that should be agreed by the community.
>> And maybe I'm completely out of bounds here but I think that even  
>> some
>> form of issue planning (in what version to expect what) can be
>> beneficial as well, JIRA is great for that too.
> Yep.  It is.  But be careful to keep things as open and welcoming  
> to volunteers.  Rigid process is off-putting to some people.
>>>>> JIRA is great for tracking, but there are things to be cautious  
>>>>> of -  I've experienced technical discussion moving into the  
>>>>> JIRA itself, which I think is not good, because it loses the  
>>>>> bigger audience (IMO).  I hate having to read JIRA discussions  
>>>>> on the JIRA-generated emails, and it makes it hard to  
>>>>> participate offline (unlike email) or search (unlike email...)
>>>> I'm aware of that but in case of tasks to be implemented  
>>>> somewhere in
>>>> the future I like to see any conclusions of discussion in JIRA  
>>>> versus
>>>> digging them from my email archive.
>>> That's fine - that's a good use of JIRA.  But you can simply just  
>>> put a URL to the mail archive in the JIRA...  easy to find later.
>> That is possible, but I'm inclined to describe the summary of such an
>> issue neatly in JIRA. But as I've not that much experience with  
>> working
>> in an ASF style project opposed to playing a dictator I really don't
>> know what will work the best.

Well, this should be interesting. There are a wide range of working  
styles in Apache projects that depends a lot on the community who are  
actively working on the project. It might be good to propose some of  
the guidelines as updates to the site and see what people think. That  
way, it's not just talk; it's proposed action.


>> -- 
>> Mark

Craig Russell
Architect, Sun Java Enterprise System http://java.sun.com/products/jdo
408 276-5638 mailto:Craig.Russell@sun.com
P.S. A good JDO? O, Gasp!

View raw message