BIND 10 #464: Additional processing for authoritative NS

[BIND 10 Development]

#464: Additional processing for authoritative NS
-------------------------------+--------------------------------------------
      Reporter:  zzchen_pku    |        Owner:  vorner   
          Type:  task          |       Status:  reviewing
      Priority:  major         |    Milestone:           
     Component:  Unclassified  |   Resolution:           
      Keywords:                |    Sensitive:  0        
Estimatedhours:  0.0           |        Hours:  0        
      Billable:  1             |   Totalhours:  0        
      Internal:  0             |  
-------------------------------+--------------------------------------------
Changes (by zzchen_pku):

  * owner:  zzchen_pku => vorner


Comment:

 Replying to [comment:3 vorner]:
 > Just few minor points, and one bigger:
 >
 > '''query.cc''':
 >  - the Query::getAdditional header should have the opening bracket on
 the same line.
 >  - Line 75 is too long (81 characters, the guidelines recommend max 79)
 >  - The assert is probably wrong. The fact that one zone does not contain
 a NS RRset at the apex is not a programmer error, an exception that can be
 caught should be used instead (probably derived from BadZone, as the
 missing SOA one is). It is not possible to catch asserts, they kill the
 program right away.
 Done, thank you for checking.
 > '''query.h''':
 >  - „For now, in order to optimize the additional section processing, we“
 ‒ what is optimised by including more data in the answer? I don't see an
 optimisation (possibly simplification of the code, yes, but I guess that
 sending more data will not make it faster).
 IMO, "Optimization" means that we need't to query twice to get all address
 informations of authoritative name servers. For if we don't search
 additional address records in GLUE_OK mode, we may need to query the
 delegation zone again to get glue address RRs of the apex NS.
 Also, this is a suggestion from jinmei, currently bind9 and nsd has
 different behaviors for it, should we send it to bind10-dev and discuss
 about it?
 > The tests:
 >  - Could you add an out of zone nameserver to the auth_ns_rrset? Its
 address should not be included in the answer (nor can be, as it is
 unknown), but just to make sure it does not crash or anything like that.
 >  - Also, it might be worth putting the *_query.process calls inside
 EXPECT_NO_THROW. Not that it would change the result of tests, but now,
 when anything is thrown, the whole test crashes, which means additional
 output (possibly more failure reports) are missing and they can help
 finding the problem.
 Done.

 New change has been committed on r4158, thank you.

-- 
Ticket URL: <http://bind10.isc.org/ticket/464#comment:4>
BIND 10 Development <http://bind10.isc.org>
BIND 10 Development