I'm not sure if this should stay on this list or not, but I thought it
was useful to have in the archive...
I've been planning to send a "guidelines" email to this list for several
days, and I'm still not quite there. Others will, no doubt, respond in
more detail to your questions, but I thought I should send out an
> Mike et al.,
> Well, perhaps nanofaq, cast into prose. From reading the web and lists I've
> seen, I'm still a little unclear on the actual modus operandi for us
> ex-officio engineers.
> I gather that the community buglist is world-readable--all bugs, save perhaps
> in the still-encumbered components. We can see a sanitized subset of the ARC
> proceedings, with Sun proprietary stuff redacted.
> The sponsorship process is individual bug driven, so that, for each selected
> bug, we apply, get selected by (perhaps) a different sponsor than for the
> last bug, do our thing, submit the proposed code change to the sponsor. The
> sponsor will surely change if we move across different communities, but
> otherwise might not.
> Sponsor reviews our code, then if s/he approves, s/he does the actual
> putback; we never touch the real gate. Prior to that, sponsor stress tests on
> you guys' test farm. We are assumed to have already tested as far as
> possible on our own farm.
> Putbacks are done in the name of (sponsor, external engineer), so that
> Bonwick knows whose address space to "tear a new address space gap" in, when
> something breaks :-) [cf., "The Quality Death Spiral" by Jeff]
This is correct.
> Assuming I have this right, does this procedure remain the same for everyone?
> It seems fine for an occasional contributor, but I can image an "Outside
> Jack" feeling a bit frustrated by the protocol.
> If the entity being sponsored is just an engineer, and not the ordered pair
> (engineer, specific bug), then within that community, Jack could work with
> the same sponsor for all his code, and the sponsor would get a personal feel
> for Jack's work. This I think would expedite the mechanics and improve
> product quality as well.
There are pros and cons to each approach, but we thought "rotating"
sponsors was the best way to start. If for no other reason than sponsors
go on vacation (allegedly).
If it looks like this method isn't working in the short-term, we can
always change it. We do plan to revisit the mechanism for getting
sponsors assigned in the future, though I don't know if we are thinking
about changing the model. I suppose a lot will depend on the results we
get trying it this way.