Hi All,

I'm wondering about the usefulness of our issue life cycle.

Aside from reopened/unconfirmed issues, the usual workflow as defined in
the cookbook
(http://argouml-stats.tigris.org/documentation/defaulthtml/cookbook/ch09.html)
is:

New
Started
Resolved
Verified
Closed

It appears that this is largely based on the stock tigris tooling. 

Currently, at argouml, we have:

New 934
Started 45
Resolved >1000
Verified 55
Closed >1000

The subversion.tigris.org project also use the tigris issue tracker, but
they seem far less bothered about verifying/closing issues:

New 526
Started 27
Resolved 2129 
Verified 2
Closed 218

subclipse.tigris.org even less so:

New 70
Started 8
Resolved 710
Verified 1
Closed 4

I understand that the reason for verifying/closing issues has to do with
quality assurance, and I don't doubt that having a process where a
second person reviews a supposed fix is a good way of ensuring that
issues are indeed fixed.  However, I do also notice that: 
 - users are probably not concerned in a particular issue once the
 problem appears to have gone.
 - users probably judge the quality of the overall product on the number
 of outstanding bugs rather than the formal closure of historic bugs.
 - developers are evidentially more interested in fixing problems than
 verifying them.
 - some issues are very hard to verify and as a result remain in the
 database forever.
 - we are not building a space shuttle.

So I am questioning whether verifying issues at all is a valuable
process.  For a life-critical system, such a workflow would be ideal, as
you would need to ensure that problems were really being solved, but is
this applicable to ArgoUML?  The evidence above suggests that other
projects do not want to use the Verify-Close parts of the tigris
workflow, yet they are still successful projects (arguably more so than
ArgoUML).  I also notice that tortoisesvn.tigris.org now use an external
issue tracker.  In my workplace, we use trac, which by default omits the
'verified' status, and instead reporters tend to re-open issues if they
are unsatisfied with the fixes (they do after all get automatically cc'd
when the issue is marked as resolved).

Perhaps our verification time would be better spent on other things than
verifying issues.

Maybe we could only make Verify/Close a requirement if the project
manager specifically requests it, e.g. for a particularly stubborn bug
or an important feature.

Maybe we could just automatically close issues that have been resolved
for one whole stable release cycle.

I strongly feel that the simpler processes are, the more likely they are
to be used, and the more useful work gets done.  If any flexibility is
offered with the new tigris set-up I think we should review the
workflow, and if not we should definitively decide what we want the
workflow to be.  The cookbook and reality differ at the moment.

Regards,

Dave

---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]

Reply via email to