* A new configuration option `CLICON_CONFIGDIR` has been added for loading of extra config files
* If not given, only the main configfile is loaded.
* If given, and if the directory exists, the files in this directory will be loaded alphabetically AFTER the main config file in the following way:
* leaf values are overwritten
* leaf-list values are appended
* You can override file setting with `-E <dir>` command-line option.
* New clixon-config@2020-10-01.yang revision
* Added option for configuration directory: `CLICON_CONFIGDIR`
* Only requests on the form: `<rpc xmlns="urn:ietf:params:xml:ns:netconf:base:1.0"><edit-config>...` will be accepted
* All replies will be on the form: `<rpc-reply xmlns="urn:ietf:params:xml:ns:netconf:base:1.0">...`
* Requests such as: `<rpc><edit-config>...` will not be accepted.
* You can revert this behaviour (to clixon pre-4.6 behaviour) by enabling `CLICON_NAMESPACE_NETCONF_DEFAULT`
* This API change is a consequence of: [copy-config's RPC cxobj parameter does not contain namespace #131](https://github.com/clicon/clixon/issues/131)
* NACM recovery user session is now properly enforced. This means that if `CLICON_NACM_CREDENTIALS` is `except` (default), then a specific `CLICON_NACM_RECOVERY_USER` can make any edits and bypass NACM rules.
* If a default value is replaced by an actual value, RESTCONF return values have changed from `204 No Content` to `201 Created`
* clixon-config.yang: Removed default valude of CLICON_NACM_RECOVERY_USER
* Enforcing RFC 7950 Sec 7.6.1 means unassigned top-level leafs (or leafs under non-presence containers) are assigned default values.
* NACM default behaviour is read-only (empty configs are dead-lockedd)
* This applies if NACM is loaded and `CLICON_NACM_MODE` is `internal`
* Fixed: [default values don't show up in datastores #111](https://github.com/clicon/clixon/issues/111)
* Changed commit lock error tag from "lock denied" to "in-use".
* Changed unlock error message from "lock is already held" to #lock not active" or "lock held by other session".
* Fixed [lock candidate succeeded even though it is modified #110](https://github.com/clicon/clixon/issues/110)
1. Added #ifdef extern "C" to public header files.
2. Changed variables that are key words in c++, to different names (like ns instead of namespace).
3. Changed char* to const char* for constant strings.
* In summary, a user registers an upgrade callback per module. The callback is called at startup if the module is added, has been removed or if the revision on file is different from the one in the system.
* The register function has removed `from` and `rev` parameters: `upgrade_callback_register(h, cb, namespace, arg)`
* The callback function has a new `op` parameter with possible values: `XML_FLAG_ADD`, `XML_FLAG_CHANGE` or `XML_FLAG_CHANGE`: `clicon_upgrade_cb(h, xn, ns, op, from, to, arg, cbret)`
* Added: stats RPC for clixon XML and memory statistics.
* Added: restart-plugin RPC for restarting individual plugins without restarting backend.
* xml-stats moved from clixon-config.yang as state data to an rpc `datastats` in clixon-lib.yang
* Experimental: restart_plugin
* Two new plugin callbacks added
* ca_daemon: Called just after a server has "daemonized", ie put in background.
* ca_trans_commit_done: Called when all plugin commits have been done.
* Note: If you have used "end" callback and usign transaction data, you should probably use this instead.
