wiki:WeeklyMinutes20110301

Roll Call

Likun
Jerry
Ocean
Jeremy
Fujiwara-san
Aharen-san
Stephen
Larissa
Shane
Kambe-san
Michal
Jinmei
Jelte
Keith

Feedback from Last Release

https://lists.isc.org/pipermail/bind10-announce/2011-February/000009.html

Jeremy: Went well. Had some last-minute releases, but these were already identified by a few of the builders. Did not call for a freeze ahead of time, if we had then there would not be any issues. Everything was good.

Shane: We do spend some time near a release. Do we want to factor into planning?

Jeremy: Already had it planned to do ahead of time, using a branch that we're working on.

Jelte: Earlier occasions we added a bunch of stuff before a release, but this time we were just unlucky.

Shane: This is the last one of this year. Continue these next year?

Jeremy: Not sure... starting in April do we want to do a lot of quality, or continue with snapshots? Or maybe a mix, like doing 6-week snapshots but then every 3 months doing a quality release.

Larissa: Need to discuss further. I would prefer the later.

Jeremy: One concern I have still is... we need to do production releases but we're not production quality yet. We haven't been misleading yet - people know these are prototypes.

Jelte: +1 to Jeremy

Shane: Will put on agenda for f2f meeting.

Difficulties Setting up a Forwarding Resolver

Jeremy reports.

https://lists.isc.org/pipermail/bind10-dev/2011-February/002038.html

Jeremy: I documented almost all the steps in setting up a forwarding resolver. Sent to bind10-dev list. Seems like we're so far away as something we want to ship or give as production quality software.

  1. Using bindctl - if you make a typo it is unclear what the error is
  2. Configuring forwarding addresses had an unreadable error message
  3. Requires you to define a port number (known issues, should not be required)
  4. Running as non-root failed with different error messages - none pointed to where problem was!
  5. Have to start with nohup, make sure stderr/stdout go to /dev/null (with named you do not have to do that)
  6. When doing shutdown, get error messages that are empty
  7. Getting Python tracebacks that are lengthy written to screen hides error message
  8. Kerberos was used for all logins, so when DNS was down I could not login remotely

Please look at my e-mail, if you have tickets let me know, if not we need to create.

This was setting up a forwarder.

Shane: I had a look, but mostly at the boss-related stuff. Jeremy or Stephen had an idea for "bind10 --cleanup" which would stop orphaned processes.

Jinmei: Most of these seem to be generic, rather than forwarding-specific, right?

Jeremy: Yeah, I think they're pretty generic. Other than configuration of the forwarder.

Shane: Turn these into tickets, then run the experience again.

Jeremy: I was working through the document for this, but did not get far because of all the problems. We have the ops team planning on setting up xfrin and xfrout authoritative server so I want to make sure the instructions for that are clear.

Jeremy: Please read the e-mail.

End-of-year Release

Announcements to be sent 2011-03-22, but release ready 2011-03-17 or 2011-03-18.

Larissa: Technical blog articles. We did not want to do a big media push since this not production ready. I would like to have them over a few days before and after the release, and we would have our events person co-ordinate that. I will send an e-mail to the list.

Shane: Put on bind10-dev... possible someone may ask for something to be explained.

Larissa: No press releases. We will do these in September.

A-team status (for R-team)

Shane: Features are done.

Michal: Small features still in review. 2 or 3.

Shane: Brainstorming session about to test things.

Jinmei: Done in that we have completed the in-memory data source. Still ongoing, unrelated bug fixes.

R-team status (for A-team)

Shane: Still some features being worked on.

Stephen: Working on fallback to TCP.

Jelte: Hope people are working on negative cache. I'm working on using the NSAS in the resolver. Going slowly.

Ocean: I think tomorrow the negative cache will be finished and ready for review.

Shane: You can see the ResolverTestingBrainstorming page for testing ideas.

Future Code Stubs

Jinmei notes:

https://lists.isc.org/pipermail/bind10-dev/2011-February/002037.html both about the general sense about how to handle mostly empty implementation, and about the specific point of XXXCacche::dump/load/resize().

Jinmei: I noticed when I tested with cppcheck, there were warnings due to the fact that these were empty. I suppressed warning for these, but I think in the general case if we don't have a short-term plan we should remove from the master tree.

Jinmei: A related point is that for this specific case I would suggest applying this practice and remove them, and put in comments, so we can perform a full check with cppcheck without suppressing warnings. Then we can enable the check throughout the file.

Michal: Would cppcheck complain if there was header but no body?

Shane: I think we should remove this. It goes along with the general policy of not having blocks of commented-out code, get it out of there.

Jinmei: If nobody disagrees I will make a ticket for this task.

"Pacific" Daily Call

Jinmei notes:

"we've found there are several problems in it such as incomplete/missing calls on Monday/Friday? and schedule conflict. Also, there's now only one "developer" in the US side of the region (i.e., me), it's more questionable to keep this style. Considering this call is mainly for developers (in my understanding) and the fact that all but one developers are in Asia and Europe, my personal suggestion is to revert to the previous style, and US people join it best effort basis."

Jinmei: Observations:

  1. We have issues on Mondays and Fridays. On Mondays only Asian people can make the call (still Sunday in the USA). On Fridays we skip the calls (Saturday in Asia, only 2 other participants).
  2. Some days are not good for CNNIC developers, so we miss the calls.
  3. Now Scott has moved to BIND 9 development, the only developer doing coding/reviewing is me (Jinmei). I don't think the Pacific call is really effective.

Larissa: I agree there is an issue. The issue with CNNIC is that if it is a day when they have had a BIND 10 team call the night before, they are not at the office yet when we have the daily call.

Jinmei: We soon have a face-to-face meeting, so maybe we can postpone the decision until then. But my impression is that we should conclude this is not really a good idea and go back to the original organization. If we do that, most of the developers are in Europe and Asia so it makes more sense for the other people to participate in that in a best-effort basis. For myself I don't mind joining it in most of the cases. If I do that then all of the developers are there.

Shane: Jeremy would not be on daily calls...

Jeremy: That's fine. I would prefer it during regular workday.

Larissa: Maybe best possible solution.

Shane: Another possibility is a 10-minute call with Larissa, Jeremy, and Shane.

Shane: Some details to be discussed on list (for example summer time), but in principle I think it is an improvement. Other opinions?

Jeremy: To clarify, these calls are part of the Scrum protocol? [ yes ] I wasn't clear if these were part of the protocol, if they have created success over the past 6 months or so.

Larissa: People think they are helpful or not?

Stephen: I have found them useful. In most cases it is not conflicting, but once or twice it has avoided a conflict.

Jelte: Yes these are definitely useful.

Jeremy: I wasn't sure if it was the same small set finding it useful. I wasn't sure for the rest. But that's fine either way.

A.O.B.

Jeremy: Any more details about the Google Summer of Code?

Shane: Sent mail to ISC staff about this.

Larissa: Need to have this by Friday.

Jeremy: Some hints.

  1. We need to make sure we have multiple mentors listed for all projects regardless of whether they are going to be the ones doing it. We can move mentors once we have students.
  2. Having a good list of projects. We need to open that up to the bind10-dev list now to get others contributing to building that. Also hit a student now who can share some of their project ideas - regardless of whether we get approved by Google.
  3. They do have female t-shirts!!

Larissa: Should also be for BIND 9 and DHCP development. I have no issue with the list going to the bind10-dev list. I'll send that. Appreciate that you have done this before.

Likun: Can we send students?

Larissa: I don't know. I will send a link to the SoC project.

Michal: I tried applying for Jabber foundation 2 years ago, and usually there is not a lack of students applying. Like 10% of them get accepted... problem is convincing Google that the projects are nice. It may also be advertisement for us... students may look and say "hey what's this?"

Jeremy: Need to be very clear that DHCP or other projects don't make BIND 10 lose availability. BIND 10 has a public development list, and the code tree has always been public.

Larissa: I agree there is an issue there. I'll make clear we are evolving our development process at ISC, and the further along is BIND 10 but we're working on moving BIND 9 and DHCP 4 along the path.


Jinmei: About e-mail about reverting code when triggers build failure. I think it's a matter of wider interest for developers.


Jeremy: I'd like to switch both servers to port 53. This might break some systems if you're already doing things on port 53. Concerns about this?

Shane: Better to do now than after release.

Michal: If someone has something running, he cannot configure on a different port. So I suggested 2 weeks ago that we start without any ports configured.

Shane: Send to list.

Last modified 7 years ago Last modified on Mar 1, 2011, 4:02:24 PM