BIND 10 #169: notify recv

[BIND 10 Development]

#169: notify recv
----------------------+-----------------------------------------------------
 Reporter:  larissas  |        Owner:  hanfeng                                    
     Type:  task      |       Status:  reviewing                                  
 Priority:  critical  |    Milestone:  04. 2nd Incremental Release: Early Adopters
Component:  b10-auth  |   Resolution:                                             
 Keywords:            |    Sensitive:  0                                          
----------------------+-----------------------------------------------------

Comment(by jinmei):

 Replying to [comment:5 hanfeng]:
 > > Please at least add detailed tests first.
 > The code is direct hacked into main.cc just like axfr logic, there is no
 head file for the two added function, so it's hard to add it into unit
 test. Actually, the axfr and ixfr logic should be handle all in
 handleMessage, but current interface can't support them.
 > I have write separate script to send package and check the reply. I will
 add the test code into unit test, once the refractor of auth server is
 done.

 IMO, this cannot be an excuse for skipping tests.

 If the reason you cannot  write tests is dependency on another
 refactoring, we should refactor the code first, then write tests for the
 new feature, and then design and code the new feature.

 We can easily find an excuse for skipping tests (because they are often
 boring and implementing new features tends to look more interesting), and
 IMO there should be very high bar to allow that to fight against that
 temptation.  We lowered the bar for the year 1 release due to a hard
 deadline based on contract, but we should consider it a very exceptional
 case, rather than a precedent.

 Also, if we found some new feature may be difficult to test just a day
 before a code freeze, whether it's due to dependency on another task or
 other reasons, that would mean we underestimate the planning, either or
 both about dependency consideration and about necessary time for that
 work.  We should learn from that so that we can make a better estimation
 next time.   If we allow to skip tests (or documentation or review) for
 some easy reason, it would make it difficult to revisit and improve the
 estimation because that means "we were accurate".

 Anyway, this is a more general topic, and many of the points are my
 personal opinion, for which others may disagree.  Perhaps we can discuss
 it tomorrow?

 Meanwhile I'll take a closer look at the branch to see how we can move
 forward practically.  Unfortunately, however, I suspect we must give up
 including this feature in the next release.

-- 
Ticket URL: <http://bind10.isc.org/ticket/169#comment:6>
BIND 10 Development <http://bind10.isc.org>
BIND 10 Development