wiki:WeeklyMinutes20100401

Attendees

Jelte
Shane
Likun
Shawn
Hankins
Jeremy
Michael
Fujiwara
Kambe
Jinmei

AP from Previous Weeks

  • Shane to try to get logging for Jabber

Shane: Does anyone have any experience with this?

Michael: I have some experience. Need user ID and password.

Shane: I have this. bind10-bot

Jeremy: Will look into this.

Shane: Will send list and user/password to Jeremy.

  • Michael to write blog entry about SQLite3

Michael: in progress

Y2r01 release (from Jeremy)

  • Deadlines

Shane: focus will be bug fixes, code review, maybe XFR stuff


Shane: boost library dependencies, asio, Python wrappers

Shane: asio from Boost to non-Boost?

Jinmei: Boost requires library just to get Boost versions of error codes. Necessary change should be minimal.

Shane: Useful for next release, or just put on backlog?

Jinmei: Not sure. If we still use the Python binding we'll still need libraries.

Jelte: Only makes sense if we get rid of other Boost libraries.

Michael: Also we might need Boost library in future.

Jinmei: For short term goal I don't think it makes much sense. But I think we should have a more general discussion about how much dependency we should allow. For example, whether we ever want to use the library version of Boost. Or whether we want to use Boost in the first place. Too heavy for conference call.

Jelte: Something for meeting at CNNIC?

Jinmei: Yeah, but maybe a quick opinion on this topic will be useful on this conference call.


Shane: Does using Boost for Python libraries make things harder or easier?

Jinmei: I think this is not as obvious as the asio case. I guess we should first try to take the Francis-like approach, and see how difficult it would be to make the decision. If it turns out to be difficult maybe we should drop that idea. Maybe we should use 1 to 2 weeks to try adapt Francis' previous effort to the current version.

Jeremy: I would suggest that Jinmei does *not* do it, so we could have more eyes looking at the code as they create wrappers.

Jelte: I wouldn't have a problem doing that.

Jinmei: Can just commit Francis' earlier work.

Shane & Jinmei to find this code.

Feng: Going to get rid of Boost Python?

Shane: We don't know.

Feng: Also know OSDI and other projects use Boost libraries to write Python interface. Have several versions, and support older version of C++. Boost Python is not easy to install. But if we write C++ wrapper from beginning I have concern about communication since Python is in C rather than C++. We can look into Francis code at first.

Jelte: I quickly looked into this when it came up. Another library is SWIG but this also has dependencies. If doing it ourselves is too much work then Boost is probably the way to go.

Jelte: I will look at what Francis did and try to do the same for the current DNS library. If that succeeds we can do that otherwise we can keep the Boost or look for something else.


Jelte: We've been talking about the command channel messaging. I hope Michael has followed. If not, then I'd like you to do so now.

Michael: Will catch up to it Monday or Tuesday.

Jelte: I did some quite invasive things into the session stuff, some of which might not be a good thing at all. The problem is that right now the command & control daemon has different threads that read from the c-channel. One thing is to add a lock, the other thing was that if you did not ask for a reply then you only get things that are *not* a reply.


Jeremy: Deadlines. I was hoping to do the release on the 21st or 22nd and not the 23rd. I don't want to be rushed. I was hoping we'd have all the new features & bugfixes planned by the 19th. That loses a whole week. Last time we had 30 commits within the last 24 hours. So I had to retest every single time and that took 24 hours.

Shane: I think that makes complete sense. Do we want a code freeze?

Jeremy: Talk about it on a case-by-case basis. Also note that we don't have beta or alpha releases.

Jelte: Being an optimist, I hope we'll find less show-stopping bugs this time.

Jeremy: Last time there were bug fixes that I purposely did not pull up. We just have to live with them on a case-by-case basis. If everyone decides on a certain day (the 19th) and we do the release on the 21st that will be easier than last time.

Jelte: One thing is that we need to do at least one big testing of the whole system before that. If we only start testing everything on the 19th then we're bound to find one big thing.

Jeremy: I've been doing distcheck and running tests from the tarball, so at least we'll be able to see it from that perspective. It's been failing every time, but at least it's automated. I'll send an e-mail about that.

Michael: Do we actually have tests that run the system?

Jeremy: Not yet, but that's something I'm working on.

Jelte: We may want to clean out the tester directory or clean out the old commits.

Shane: Yeah, archive things older than a week or something like that. Or delete them.

Shane: FYI I'll be Beijing on the 19th then.


Jeremy: If anybody has anything the specifically e-mail the list or speak up now. There may be something you're really concerned about that we might forget.

  • Status of code reviews

Jeremy: Hoping this release will include all reviewed code. I know there's a few more that have had reviews completed. I will copy those into the reviewed branch. Hopefully over time we'll get the list really small. Hopefully by the 19th.

Shane: Do we have testing goals for this release?

Jeremy: Maybe to automate the Python tests that we do have.

Shane: I think that makes sense.

Jeremy: I don't think any tests were done for the Python libdns tests.

Jelte: If we're going to do the hard-coded Python wrappers I'll certainly do tests for that, so that might solve it.

Shane: Do we want the Python API to be similar between the native and Boost versions?

Jelte: I'd like to keep it make it similar to the C++ when possible. CamelCase vs. under_score is an exception.

Jinmei: Question is whether we want to provide exactly the same methods as what C++ provides or fewer or even more

Feng: Currently API for Python should be 1 to 1.

Jelte: The only thing that might cause a difference is type related stuff.

Feng: Difference is about the iterator. C++ can generate iterator, but Python version uses "for".

Feng: Currently we only test the interface. We didn't test the Python API, so maybe this should be done. I don't know whether we need to test this part or not.

Michael: I think we should at least test some of the functions.

Feng: Yes with direct interface that is true.

Jinmei: The test details depend on the bindings.

Michael: If we use an automated tool maybe then we don't need the tests.

IETF report (from Shane)

BIND 10 report to IEPG went well. Presentation linked to. There may be an audio feed archived somewhere - the BIND 10 presentation was 2nd so not too far in.

  • Steering committee

Steering committee meeting was fine. Will be giving updates targeting status for senior management & board-level folks with each mini-release we do.

  • New sponsors

Got committment from a new sponsor, who will also sponsor DHCP with 50% of the grant.

  • DHCP

David, Shawn, and Shane met on Friday morning to discuss DHCP in BIND 10. Working on creating DHCP "story" similar to the 5-year plan for DNS.

Meeting ended due to excessive echo. :(

Last modified 8 years ago Last modified on Apr 7, 2010, 2:54:56 PM