[CentOS-devel] handling ABRT

Jeff Johnson n3npq at mac.com
Fri Nov 26 17:52:27 EST 2010


On Nov 26, 2010, at 4:01 PM, Ralph Angenendt wrote:

> 
> So it is either abrt without reporting, no abrt, or b.c.o running on a
> bugzilla instance, afaics.
> 

There are fallacies in reaching the conclusion that
those are the only possible outcomes.

The argument chain that leads to those alternative resolutions
goes something like this:

    Q: Are segfault's (tracked by ABRT) bugs?
    A: Almost certainly.

    Q: And where do bug reports get filed go ... ?
    A: In a bugzilla! duh!

which leads to a certain narrow perspective and Q.E.D conclusion:
    1) rip ABRT out of CentOS
    2) cripple ABRT or gimp it up with "no reporting"
    3) set up a b.c.o instance

Bugzilla is almost certainly a reasonable end-point for ABRT
if you have gazillions of paid employees who are
paid (as part of a "service" model) to track
store-bought product defects and paying customer complaints.

But is that the right model for CentOS? Hardly imho ...

The better model is what is being done with kernel.org bugs
(which I claim was studied when ABRT was written. I'm sure
there were other implementations at M$ and with Apple radar blips too).

With kernel.org, bugs are tracked as a software devel process metric, not as
a paid wage slave performance indicator.

SO I suggest that you should look at other alternative end-points
for ABRT automated segfault/bug end-points, and view as a
objective distro "process" metric to prioritize scarce resources, not track,
bug reports. No user id's needed is just one of many benefits.

Unless you really like sipping from a bugzilla fire hose,
with everyone 2nd and 3rd guessing whatever solution you
attempted to solve a reported problem. There are hair shorts and
strait jackets for those who really MUSTHAVE their nagware to
	Get Things Done!

JMHO, YMMV, everyone's does, yadda, yadda.

hth

73 de Jeff



More information about the CentOS-devel mailing list