BIND 10 and Microsoft Windows

Jelte: What to do with open design issues Francis raised?

Jinmei: Can we safely rely on the fork() model if we want higher performance? Depending on whether Windows supports this we have flexibility about this kind of discussion. If we have a specific policy we don't have to repeat that argument again.

Shane: I think most of Francis' questions are being addressed.

Shane: I don't know!

Jinmei: My question is whether we need to support Windows or not, or whether we have the option of not supporting Windows.

Shane: I don't know. I think Windows is important from the resolver side. ISC will make sure that we implement Windows support.

Shane: Like I said in Prague: "If we're going to do it, we'll do it right."

Jinmei: My concern is that if this is an open question we may have no constructive argument if, for example, someone proposes an optimization and someone says "Windows can't do that".

Shane: Yes, that is a valid concern.

Jelte: Have any of the sponsors?

Shane: Not to me.

Michal: Maybe the answer to that question depends on the optimization. If a performance optimization and it works reasonably without it, then I don't see a reason not to leave it out. OTOH, if we need something like fork() and we would have only minimum support for Windows - like 1 query / second - then it is probably not best to do it.

Jeremy: An ISC customer had 300 servers running on the Windows platform. I don't know if we had a second customer running Windows.

Shane: So at least someone uses it heavily.

Jelte: As a short continuation is to replace it with actual TCP sockets... but we can't do that because we chose domain sockets for the file system protection.

How to prevent Trac spam (Jinmei)

Shane: There is a Trac plug-in for spam.

Jeremy: Yes, it has like 15 different features. It's pretty comprehensive. As an easier solution there is a math Captcha. I wanted to see if it could be modified for new user creation.

Shane: Yeah, maybe that's easier. Currently lower priority than getting to the new server.

Shane: I have an RSS feed, and I see spam. And I delete it. We don't get too much spam.

Michal: Yes, I do that too, but I hope if we delete it fast enough that they will consider our site not interesting.

Jelte: They don't look if it works, they just run it.

Michal: In the past 24 hours there were like 4 spams.

Jelte: Every spam is bad!

Jinmei: I did not raise this due to a specific concern but since we talked about it on Jabber I thought it might be good to have a short discussion on the call.

TSIG Status (Michal)

Michal: Which tickets are implemented already and which tickets are waiting on which? Which tickets can I take? #782 is new...

Jelte: Already almost done in another ticket.

Stephen: How many of these tickets are waiting on something else to be completed? Is #782 complete?

Jelte: If we don't want optional algorithms, when #781 is done.

Stephen: We have #813, #816, #814, #815, #816...

Jinmei: #813 will depend on signing, #812.

Jelte: All the other ones depend on both #812 and #813. These could be implemented without crypt stuff...

Stephen: #812 -> #813 -> #814 -> #815/#816

Michal: #856 can be started once #812 is done

Jelte: #782 can wait...

Shane: #814 can start before #813 is done then?

Jinmei: We may be able to separate signing and verifying for #814.

Stephen: To summarize:

Stephen: We're waiting on #781, and so on.

Stephen: If you take non-TSIG tasks and wait for #781 and #812, then other TSIG tasks can be started.

Shane: There are a number of these. Either "new" or "assigned" that belong to UnAssigned.

Jinmei: Maybe it would have been a bit better if we had done #812 not depending on #781. And concentrated on the abstract-level interface, with empty hash algorithm or something.

Jelte: In retrospect it should have been designed a bit more and not assumed we had the same ideas about what the interface would look like.

Next BIND 10 face to face meeting


DHCP        2011-08-26 to 2011-08-30
Community   2011-08-29
DNS         2011-08-29 to 2011-09-02

The idea is:

  • a DHCP meeting on Friday and Saturday
  • a meeting with the community on Monday
  • a joint DHCP/DNS meeting on Tuesday
  • a DNS meeting on Wednesday, Thursday, and Friday

Amsterdam, Netherlands @RIPE NCC office Redwood City, California @ISC office

Stephen: If we can get people to come in and discuss their BIND 10 experiences and what they want from BIND 10 that would be great.

Larissa: More of the steering committee & sponsors are close to Amsterdam, but if we meet in RWC we can have Silicon Valley types and ISC customers.

Shane: Larissa had an idea of doing one in RWC and one in Europe this year.

Whither Hooks?

Michal: We agreed with Jelte that both our proposals are kind of the same, looked at from a different point of view. Everybody else did not notice, because it was during the IETF and everybody was busy. So can we ask everybody to look at it or something?

Jelte: I tried to remember it, but that discussion was buried in my mailbox.

Shane: Okay, please everybody have a look! (Subject: "plugin proposal")

Stephen: We have a lot of discussions that get buried in mailbox. Would it be worthwhile taking a pro-active approach and ask someone to write a summary for the wiki?

Shane: I think it makes sense. To be discussed (not on the mailing list!).

Michal: During face to face meeting we had a discussion. Proposal was to split processing of server into parts, and anybody could put any other part in-between. Other proposal was to identify specific points to place a function. Opposition was it might be more complicated. So a kind of study showing that it was less complicated - or the same complication. I think people need to read it and point out errors in thinking or say they agree or don't agree or something.

Last modified 7 years ago Last modified on Apr 26, 2011, 4:08:13 PM