Changes between Version 1 and Version 2 of WeeklyMinutes20091217

Dec 17, 2009, 6:35:14 PM (9 years ago)



  • WeeklyMinutes20091217

    v1 v2  
    1 WARNING: badly formatted version, just for checkpointing
    41Notes from the BIND 10 call.
    6 *Agenda*
     3= Attendees =
     17= Agenda =
    819      * Roll call
    2738      * AOB
    30 *Attendees*
    31 {{{
    32 Jeremy
    33 Feng
    34 Shane
    35 Jelte
    36 Fujiwara
    37 Larissa
    38 Kambe
    39 Jinmei
    40 Michael
    41 Hankins
    42 }}}
    44 *Additional topics*
     40= Additional topics =
    4642    * Database experiments
    4743    * Testable code
    49 *AP*
    50     1) Not done
    51     2) Done
    52     3) Not done
    53     4) Not done
    55 *Status checkpointing*
    58 DNS Message API: not much since last week. Name class implementation almost complete. Feedback from Francis based on work on Python bindings. Probably making some additional changes to base design, going to finish within a couple days and ask for official review and merge to head branch.
     45= Action Points =
     46See above.
     47  1. Not done
     48  1. Done
     49  1. Not done
     50  1. Not done
     52= Status checkpointing =
     54== DNS Message API ==
     55Jinmei: not much since last week. Name class implementation almost complete. Feedback from Francis based on work on Python bindings. Probably making some additional changes to base design, going to finish within a couple days and ask for official review and merge to head branch.
    6059Michael: what is the test coverage like?
    6161Jinmei: number of test cases?
    6263Michael: edge cases covered, things like that?
    6365Jinmei: tried to cover that, for example using long names, invalid wire format names, and so on. This process cannot be perfect, but I did try to cover some edge cases.
    6467Michael: code coverage test tool would really help in writing tests.
    6569Shane: anyone know any tools?
    6671Michael: all static these days...
    6773Larissa: what about Coverity? maybe some other tools?
    6875Jelte: gcc also has test coverage...
    69 AP: Jeremy to look into this
    71 Python DNS Message API:
     77'''AP''': Jeremy to look into this
     79== Python DNS Message API ==
    7280Shane: Francis did some work.
    7382Jinmei: had a quick look at this. Don't know whether he used any framework to connect C++ and Python. Seems trivial wrapper for C++.
    7484Shane: would prefer to see DNS Name message first.
    7586Jinmei: yes, affects implementation, but doing Python binding in parallel may help us think about it, which we need to do anyway
    76 AP: Shane to give commit access to Francis
    78 msgq:
    79 Michael: still done
     88'''AP''': Shane to give commit access to Francis
     90== msgq ==
     91Michael: still done :)
    8093Shane: waiting on review?
    8195Jelte: yes, I don't know the review process yet
    8297Jinmei: documentation and testing?
    8399Michael: not much documentation... but no torture tests or anything like that
    84101Jelte: once I get to porting to new msgq might run into things
    86 Configuration:
     103== Configuration ==
    87104Jelte: seen a lot of airports in the last few days... only home for 2 hours
    89 bind-control:
     106== bind-control ==
    90107Likun: Got review from Michael, seems that current implementation has some problems. Not easy to map command to URL properly, more work needed on this.
    91109Michael: Anything working is good, we can change it! Running code is better than a perfect solution.
    92111Feng: We'll try to put command in GET or POST so we have some code that can run.
    93113Michael: When you issue commands over the API, you are not using a RESTful interface. You're changing state! So go with what you have and we can change it later.
    94115Shane: You mean stick with XML-RPC?
    95117Michael: No, it was re-written using REST.
    96119Feng: REST should be resource oriented, using standard method like POST/GET to modify resource. If we want to evoke some command, then...
    98122Shane: how do you, for example, say "send notifies"
    99124Michael: probably add a state to say "you need to send notifies"
    101127Jelte: on configuration, I was also heading this way... if you go this way then I can fit it in with the changes I have in mind.
    102129Michael: also change later... use working stuff now
    104 Stats:
     131== Stats ==
    105132Kambe: don't have much progress, but researching Net::SNMP, because could not find a Python SNMP implementation. Now think I need to continue to research the problem now.
    107134Jinmei: Google shows lots of libraries...
    108136Michael: these tend to be client side... we may need to work with UNC SNMP thing
    110 BoB:
     138== BoB ==
    111139Shane: no changes
    112 AP: Shane to check in TODO
     141'''AP''': Shane to check in TODO
     144= BIND 10 DHCP meeting =
    115145[ Shane talks about BIND 10 DHCP ]
    117 1st deliverable:
     147= 1st deliverable =
    118148Jinmei: first deliverable authoritative server implementation?
    119150Shane: AS112, but not totally separate!
    122 Face-to-face meeting
     152= Face-to-face meeting =
    123153Last week of January... will do it on-list
    124155Question about time for visas.
    125157May also do one after IETF - week before due date!
    128 Database experiments:
     160= Database experiments =
    129161Michael: We assumed SQL is slow, but SQLite3 seems pretty fast. Want to decide what to use for database layer format thing.
    130163Shane: We should test an in-memory version...
    131165Jinmei: we need a customized in-memory database for performance
    132167Michael: eventually...
    133169Shane: So we may be able to save a lot of work for the Y1 deliverable by not shipping an in-memory database?
    134171Michael: Yes.
    136 Michael: Nex t step to figure out what format would look like.
    139 Testable code:
     174Michael: Next step to figure out what format would look like.
     178= Testable code =
    140179Michael: smaller functions are more testable. Some parts may not be reachable, but we need unit and system tests.
    141181Michael: I have been converted to test-driven development! Give it a try.
    143 AOB:
     184= AOB =
    144185Jeremy: working on wiki page for logging concepts
    146189Jeremy: for AS112 deliverable, confused on knowing which components we are actually using since I see development in different branches. Maybe we need a file in Subversion or a Wiki page talking about what the branches are.
    147191Michael: make a "review" directory like experimental...?
    148193Jinmei: when we have a certain amount of progress, will concentrate on main trunk
    149194Shane: for now, just send mail to list about this, won't be a long-term problem
    152198Jinmei: possible extension to automake... Francis asked whether we can identify the Python version at configuration time. I defer the question... put to Trac ticket system.
    153200Shane: I think this is a good use for Trac.
    155204Jinmei: about schedule. Some of us will have vacation in 2nd half of this month, possibly beginning of next month. Should we worry about that?
    156206Shane: Can everyone e-mail me their vacation plans?
    158208Jinmei: Is the 24th a problem?
    159 [ many people ]
     210[ many people say "yes" ]
    160212Jelte: Move to Wednesday?
    161 AP: Shane to mail move to list to confirm.
     214'''AP''': Shane to mail move to list to confirm.
    162216Michael: also the call after?
    163218Shane: probably move to Wednesday also...