BIND 10 #983: Python wrappers for ACLs

[BIND 10 Development]

#983: Python wrappers for ACLs
-------------------------------------+-------------------------------------
                   Reporter:         |                 Owner:  jinmei
  vorner                             |                Status:  reviewing
                       Type:  task   |             Milestone:
                   Priority:  major  |  Sprint-20110802
                  Component:         |            Resolution:
  Unclassified                       |             Sensitive:  0
                   Keywords:         |           Sub-Project:  DNS
            Defect Severity:  N/A    |  Estimated Difficulty:  5.0
Feature Depending on Ticket:  ACL    |           Total Hours:  0
        Add Hours to Ticket:  0      |
                  Internal?:  0      |
-------------------------------------+-------------------------------------

Comment (by jinmei):

 Replying to [comment:17 vorner]:

 > I think, when it is already written and is quite short, we might want to
 include it. It will be more convenient and more the python way.

 Okay.

 > But, the `PyImport_AddModule` ‒ it does not ensure the module is
 actually loaded into the interpretter (if somethind didn't load it before,
 it returns an empty module). Can we expect it is already loaded, as in the
 case of isc.acl.acl?

 Ah, good point.  Since the tests passed even if I didn't import json
 explicitly, I blindly thought it was somehow imported as part of the
 core system or something (but simply invisible until the script does
 'import json').  The fact is that the json module is imported via
 the isc.__init__.py, which imports isc.cc, whose __init__.py imports
 cc.message, which imports json.

 So, it's actually true that we could (normally) rely on the json
 module has been imported by the time the acl.dns module is used, but
 after figuring this out I now think it's probably too implicit and
 fragile.  In fact, a buggy or evil python script could explicitly
 delete sys.module['json'] after importing isc.acl and before importing
 isc.acl.dns, in which case the assumption isn't met.

 So I revised the code further: First, isc.acl.__init__.py now
 explicitly imports json (just to be explicit; as noted above, this is
 actually a redundant operation in normal cases).  Second, I moved the
 access code to json.dumps() from load() to the class initialization
 function.  This way, we could defend us against some time of evil
 behavior such as deleting sys.module['json'] after importing the dns
 module.

 > The documentation for the loader is taken from the C++ one and it speaks
 about loading the checks. There's no function and no binding for the
 checks currently, should that part be removed?

 It at least notes create/registration (if you mean it by 'loading the
 checks') is not available in the python version:
 {{{
 To allow any kind of checks to exist in the application, creators are\n\
 registered for the names of the checks (this feature is not yet\n\
 available for the python API).\n\
 }}}

 Basically, I wanted to keep the pydoc as close to the original C++ doc
 as possible because the latter would also be modified in future, and
 we want to incorporate the changes to pydoc.  To make the merge easier
 it would be helpful if the two versions are not so different.

 And, in this case, according to you there's some plan (maybe at a
 lower priority though) to add this feature to the python version, and
 if so, I'd keep these parts with a simple note that it's not currently
 available.

 At the moment I didn't touch the documentation string.  If you still
 want to remove unimplemented part, could you be more specific about
 exactly which part should better be removed?

-- 
Ticket URL: <http://bind10.isc.org/ticket/983#comment:18>
BIND 10 Development <http://bind10.isc.org>
BIND 10 Development