* Hello message semantics has been made stricter according to RFC 6241 Sec 8.1, for example:
* A client MUST send a <hello> element.
* Each peer MUST send at least the base NETCONF capability, "urn:ietf:params:netconf:base:1.1" (or 1.0 for RFC 4741)
* The netconf client will terminate (close the socket) if the client does not comply
* You can set `CLICON_NETCONF_HELLO_OPTIONAL` to true to use the old behavior of essentially ignoring hellos.
* New clixon-config@2020-03-08.yang revision
* Added: `CLICON_NETCONF_HELLO_OPTIONAL`
* The base capability has been changed to "urn:ietf:params:netconf:base:1.1" following RFC6241.
* Not backward compatible: All uses of the ca-auth callback in restconf plugins must be changed
* New version is: `int ca_auth(h, req, auth_type, authp, userp)`
* where `auth_type` is the requested authentication-type (none, client-cert or user-defined)
* `authp` is the returned authentication flag
* `userp` is the returned associated authenticated user
* and the return value is three-valued: -1: Error, 0: not handled, 1: OK
* For more info see [clixon-docs/restconf](https://clixon-docs.readthedocs.io/en/latest/restconf.html)
* Replaces the hardcoded "config" keyword.
* Implemented by a compile-time option called `DATASTORE_TOP_SYMBOL` option in clixon_custom.h
* Tests: added endtest to all tests. Removed all premature exits if BE=0
* Introduced a delay before making process start/stop/restart processes for race conditions when configuring eg restconf
* For restconf `CLICON_BACKEND_RESTCONF_PROCESS`, restart restconf if restconf is edited.
* evhtp restconf closes session on -1 fatal error (no hang)
* restconf auth-type=none call ca_auth callback
* main example for auth_type=none uses basic auth to get user but without passwd check
* Fixed: [Augment that reference a submodule as target node fails #178](https://github.com/clicon/clixon/issues/178)
* Fixed a memory error that was reported in slack by Pawel Maslanka
* The crash printout was: `realloc(): invalid next size Aborted`
* Except `CLICON_RESTCONF_DIR` which remains in clixon-config.yang due to bootstrapping
* -d <dir> option removed
* This applies to both evhtp and fcgi RESTCONF
* Both can also read config from backend, and be started from backend
* You may need to move config as follows (from clixon-config.yang to clixon-restconf.yang)
* CLICON_RESTCONF_PRETTY -> restconf/pretty
* CLICON_RESTCONF_PATH -> restconf/fcgi-path
* Not backward compatible: All uses of the ca-auth callback in restconf plugins must be changed
* New version is: `int ca_auth(h, req, auth_type, authp, userp)`
* where `auth_type` is the requested authentication-type (none, client-cert or user-defined)
* `authp` is the returned authentication flag
* `userp` is the returned associated authenticated user
* and the return value is three-valued: -1: Error, 0: ignored, 1: OK
* For more info see [clixon-docs](https://clixon-docs.readthedocs.io/en/latest/restconf.html)
* New clixon-restconf@2020-12-30.yang revision
* Solution description using openssh and utility functions, no changes to core clixon
* Example: test/test_netconf_ssh_callhome.sh
* RESTCONF Call home not done
* Applies to cli/netconf/restconf/client-api code
* Previous behaviour:
* Close socket after each rpc, but now keeps the socket open until the client terminates
* Kept locks over socket life-cycle, but according to RFC 6241 7.5 a lock should be relaeased when session ends
* Netconf split lines input (input fragments) fixed
* Netconf input split on several lines, eg using stdin: "<a>\nfoo</a>]]>]]>" could under some circumstances be split so that only "</a>]]>]]>" be properly processed. This could also happen to a socket receiving a sub-string and then after a delay receive the rest.
* Fixed by storing residue and add that to the input string if later input is received on the same socket.
* Use [https://github.com/clicon/libevhtp](https://github.com/clicon/libevhtp) instead of [https://github.com/criticalstack/libevhtp](https://github.com/criticalstack/libevhtp) as a source of the evhtp source
* New clixon-lib@2020-12-30.yang revision
* Added callback to process-control RPC feature in clixon-lib.yang to manage processes
* Changed behavior of starting restconf internally using `CLICON_BACKEND_RESTCONF_PROCESS` monitoring changes in enable flag, not only the RPC.
* Changed: RPC process-control output parameter status to pid
* This is an alternative to manage a clixon daemon via sudtemd, containerd or other
* One important special case is starting the clixon-restconf daemon internally
* This is how it works:
* Register a process via `clixon_process_register(h, name, namespace, argv, argc)`
* Use process-control RPC defined in clixon-lib.yang to start/stop/restart or query status on that process
* Example code in the main example
* Added: autocli-op extension (see new features)
* Added: rpc process-control for process/daemon management
* Added enable flag and removed presence in clixon-restconf
