On Sun, Feb 25, 2007, Mike Hommey wrote:
> What do you think of the dunc-tank initiative ?
First the good things. I clearly think dunc-tank helped Steve and
Andi be more productive, and all things equal otherwise this improved
the release. There is no denying that.
I see several problems with dunc-tank:
* trying to solve a problem without discussing the solution
* a conflict of interests
* people reacting or overreacting
* no feedback
From the very beginning of the dunc-tank discussions (Anthony's
24 Aug e-mail on -private), it has only ever been about paying Steve
and Andi for one month each. Not anyone else, not for a different
duration, not 4 people part-time. Just that. Of course in corporate
projects the managers are paid more, but if we were to do real release
management it might have been wiser to first identify the blockers and
then act on them. I'm not sure we could have foreseen them so early, but
retrospectively it would have helped a lot to hire experienced hackers
to work full-time on the kernel and on the installer.
After being met with too much opposition on -private, Anthony decided
to set up a separate project. While he tried to ensure it would not
mistaken with a Debian initiative, it was either naive or dishonest to
believe that his DPL hat and the composition of dunc-tank would not make
it look like a Debian initiative. "Debian experiments with funding group
to release 'etch' on time" is not the only misled headline we got.
Really, it's less about what was done than the way it was done. Had
the dunc-tank been a totally separate entity contacting Debian by saying
"We have $12,000 to fund some of your people to release Etch sooner,
please send us your applications and explain what you would do to help
meet the deadline", there is a non-zero probability that Steve and
Andi's applications would have been selected and probably a lot less
drama would have ensued. If you (as in "the reader") see no difference,
then you are simply on the other side of the culture gap.
That culture gap caused people to leave the project in frustration,
and that is probably the worst thing that happened.
> What do you think are the result of the "experiment" ?
On how it was set up we sure learnt a lot of things not to repeat.
On how it helped the release, I think we need to first see the final
report.
An information I am interested in seeing is a graph of how many RC
bugs were fixed everyday, and how many bugs were personally fixed by the
paid release managers. I have a feeling that they spent most of their
time triaging and fixing RC bugs. It's all right to fix bugs, and again
I don't mean to minimise their work, but many talented developers can
fix bugs; I'd like to know what other things were done.
Regards,
--
Sam.
Attachment:
signature.asc
Description: Digital signature