[bind10-dev] Minutes, was Proposed agenda for BIND 10 call on 2009-09-23

[Shane Kerr]

All,

Hm... I neglected to send out the minutes from last week's conference
call! Apologies, here they are.

Attendees:

Likun
Feng
Jeremy
Kambe-san
Shane
Jinmei
Larissa
Michael
Evan
Fujiwara-san (but with a broken microphone)

On Wed, 2009-09-23 at 18:05 +0200, Shane Kerr wrote: 
> All,
> 
> Agenda for the BIND 10 team call:
> 
>       * Agenda bashing

Shane added an additional topic: diagrams.

> * AP from previous meetings
>               * Shane to summarize code standards discussion and put on
>                 Trac site

T.B.D.

>       * Component diagram

Shane: I put an overall system diagram up. I like it.

Shane: Would be nice to have a diagram per component. C-channel and
config certainly. Also the BigTool.

Shane: Is this useful for the DNS message API?
Jinmei: Yes, it is useful for DNS message API, but not a trivial task.
Jinmei: As I have been thinking about the API, will be good to have
something like a diagram explaining the relationships.
Shane: Doing on paper and scanning is quite okay. Whatever is easy.

> * Component status/discussion
>               * c-channel (Michael)

Michael:  Have a ruby version of the API. Going to do the Python version
next. Ruby does not use callbacks, Perl does. Will write without (easier
for C++).

Jinmei: Is this for other component experimentation?
Michael: You mean the Ruby version?
Jinmei: Yes.
Michael: I don't know. I don't think so.
Shane: No, proof of concept.

Feng: Ruby version is about general idea about message API?
Michael: Yes
Feng: Can look at code and understand the design?
Michael: Yes
Feng: Committed code? 
Michael: Not yet, but shortly after this call.

>               * BigTool (Likun)

Likun: Want to explain plan communicated on Jabber. Plan to write
mock-up and use Python for this, to make sure that the UI style is
acceptable.
Likun: Can be committed tomorrow or day after tomorrow.
Michael: What language?
Feng: Use Python to write demo first.
Michael: Why not use Python for final implementation?
Feng: I think that will be the case. We can also provide a ruby
version. ;)

>               * DNS Message API (Jinmei)

Jinmei: I have been thinking about the actual DNS message object and the
related API. Thought it could be pretty straightforward, but after
thinking about additional features like RRSET ordering it became
complicated and might affect the basic design of the API. So this is
making it take longer than expected. Mixture of goals, including
performance and intuitive interface design and rich functionality. Still
thinking, but think current ideas are relatively reasonable, so can
propose something within a day or two.
Shane: Just the API, or also a prototype?
Jinmei: Prototype header files, and how could be used. Will include
actual rendering and parsing logic, which will be much more than
proof-of-concept level.

Jinmei: Am I explaining the difficulty?
Shane & Michael: not sure
Jinmei: To make rendering fast using optimized data structure, would
prefer to just copy wire format image. But if we also want to support
functionality like RRSET ordering, this will make the process
complicated. Would prefer a unified API to do both.
Michael: Pre-supplied / pre-compiled stuff is probably an optimization,
prefer to do without now... willing to change the API slightly later?
Jinmei: That is my idea too. Not too worried about balance, but will
affect core part of design, so better to give thought to extensibility
at this moment.

Jinmei: Will need feedback. Initial version will not be final.

>               * mob (Shane)

Now "Boss of Bind".

Michael: some of the things in the requirements will be hard, like
in-place upgrades.
Shane: Yes, if they are too hard they can be eliminated, but I thought
it best to include them anyway.

> Configuration and stats left off since Jelte and Kambe-san were
>         away last week.
>       * AOB

Jinmei: also thinking about how to move forward from the prototype to
the actual implementation.
Michael: how much do you think design will change?
Jinmei: difficult question! Need to expect many changes as we implement
further details.
Michael: thinking that this API and probably the c-channel will need
changes in the API... I don't know how much time on the design before
people start using it
Shane: Jeremy testing with libbind? Maybe use Jinmei's library to do
similar things?

Jinmei: notion of design document... expected format?
Shane: two consumers: users of API, 
Shane: no format
document before finished version
systematic check from Jeremy in middle of process?

Jinmei: no decision now, but think about

[ Shane ended the call abruptly here as he was feeling quite poorly, 
  sadly ending a possibly useful discussion ]