BIND 10 #1976: use meta-or-container-of data source in b10-auth

[BIND 10 Development]

#1976: use meta-or-container-of data source in b10-auth
-------------------------------------+-------------------------------------
                   Reporter:         |                 Owner:  UnAssigned
  jinmei                             |                Status:  reviewing
                       Type:  task   |             Milestone:
                   Priority:         |  Sprint-20120703
  medium                             |            Resolution:
                  Component:         |             Sensitive:  0
  b10-auth                           |           Sub-Project:  DNS
                   Keywords:         |  Estimated Difficulty:  8
            Defect Severity:  N/A    |           Total Hours:  0
Feature Depending on Ticket:         |
        Add Hours to Ticket:  0      |
                  Internal?:  0      |
-------------------------------------+-------------------------------------
Changes (by vorner):

 * owner:  vorner => UnAssigned
 * status:  accepted => reviewing


Comment:

 Hello

 It is somewhat ready for review. However, it is problematic, as much of
 the functionality is missing and some of it has dependencies on other
 tickets. So I'm putting whatever I have to review and to discussion what
 to do with it.

 First, the diff is large, so there are some hints on how to split it in
 review:
 * Starting with `c558fd6d7a67fa70d3ba78f43c6536f5c006fd3d`, it changes the
 auth server to hold the client lists. But it does not do anything with
 them. Also, the virtual module for the data sources is created here.
 * Starting with `07236d6cc1068f21d8362c5bb47c4928741d5421`, I introduce a
 new class, DataSourceConfigurator, that'll handle configuration of the
 data sources virtual module. This part also contains a small bugfix in the
 config session (`343dc8e1fefc65298ddf3cc7d82adf8b6b9a60bf`).
 * In `6107757547e99687e84c8c19f412801ac926a342`, the server is configured
 using the DataSourceConfigurator.
 * From the `6a56ac9b0e5bd227af1730fe4cd8adbd36d03d00`, the real switch
 happens. The interface of Query is changed and the server pushes the list
 into it. The real changes here are small, but the amount of changes in
 tests is large.
 * From `be542e8e2c04b89bc6788112e4006e91bcff2418` to the end, there are
 cleanups.

 To list the problems:
 * The built in data source is not used. It should be a simple change, but
 I'm proposing to do it in a follow-up branch, this one is large enough by
 itself.
 * Many tests are disabled. This is mainly because the built in and the
 #2046 not being implemented yet.
 * The `loadzone` command is gone for now. It makes no sense until #2044
 and #2046 are implemented. This means more disabled tests.
 * The system and lettuce tests are not updated. Some of them need change
 to configuration, some won't work at all, due to #2044 and #2046.
 * The old configuration options are ignored, but left in place, since
 other modules refer to them. Once the other modules are switched over too,
 we can remove them.
 * The virtual module does not check the config, since it needs the python
 wrapper to do so.

 Also, the way the DataSourceConfigurator is done might be little bit
 controversial, but I had no idea how else to do it in a manner which
 wouldn't involve a lot of work.

 So, what I'd like to propose is to review this branch, but not merge it
 yet. We would create the follow-up tickets (the built-in data source and
 the system/lettuce test update) and solve the #2044 and #2046 and then
 finish enabling the tests again. Then do the merge.

 Once merged, it will need a changelog entry:
 {{{
 The server now uses never API for data sources. This would be an internal
 change, however, the data sources are now configured differently. Please,
 migrate your configuration to the top-level "data_sources" module.
 }}}

-- 
Ticket URL: <http://bind10.isc.org/ticket/1976#comment:4>
BIND 10 Development <http://bind10.isc.org>
BIND 10 Development