Pros & Cons of ASSIGNED pointer and NEEDINFO

[Warren Togami]

Jens Petersen wrote:
> Warren Togami wrote:
>> It seems that bouncing of the ASSIGNED pointer is a polarizing issue. 
>> Some people like it, some people hate it.  NEEDINFO is a little less 
>> polarizing, although some people hate to bounce too.
> 
> I haven't followed all the discussion but it seems to me that
> 
>> "ASSIGNED to reviewer, use NEEDINFO as necessary"
> 
> looks more consistent with common bugzilla process and usage.

This is an uncertain point.

> 
> Furthermore for "ASSIGNED to next actor" one has to keep 
> copying'n'pasting the other person's address to reassign to them which 
> can get pretty tedious for quick interchanges IMHO.  With NEEDINFO it is 
> easy just to set "from Reporter" or "from Assignee" etc.
> 

I have not heard any valid argument against "ASSIGNED to next actor" 
beyond this user interface problem.  With the current user interface, it 
can be error prone if reviewers don't read and understand the written 
documentation.

Talked a bit with dkl today about the user interface.  There are 
potential ways to improve the user interface, but it would require 
Bugzilla coding that he currently doesn't have time to do.

My current theory is that most complaints were not specifically about 
this review process sucking, but rather emotional conflating of 
objections to ASSIGNED bouncing with the very real brokenness CVS ACL 
implementation.  (If you review most of the vocal list complaints, they 
were mostly about CVS brokenness and process regressions.)  I ask that 
people here please consider these two processes independently.

CVS is a less contentious issue, and will probably be solved sooner as 
there seems to be general agreement with the "CVS Admin with Flags" 
proposal.  (Please comment there separately, as I intend on making it in 
effect during Thursday.)

This review process I am less certain of.  "ASSIGNED to next actor" has 
been polarizing.  On one hand, some believe the benefits and the process 
itself is more logical.  On the other hand, the user interface may make 
it infeasible to use because it can be error prone if done without 
proper training.

I believe that the user interface problem is only minor.  Proper 
training and documentation make it usable.

I believe that the user interface problem is only temporary.  We can 
craft changes to the review interface on top of Bugzilla that uses the 
existing database fields, but with behavior that we desire.  We just 
have to wait a bit for dkl to get past his current major project, and 
use the less optimal solution in the meanwhile.

Ultimately, we need to figure out if the user interface problem really 
is bad enough.  There is no agreement on this, which is why I am asking 
for more opinions here.

(I am also exploring an entirely different alternative to either 
option... but this idea is less developed at the moment.  I may write 
about this soon if it pans out.)

Warren Togami
wtogami at redhat.com