On Fri, May 18, 2018, laczenJMS wrote:
As all nodes have mandatory configuration server and health serverThis is an interesting idea, and definitely could be made to work.
However, only in some limited scenarios. One challenge is that it'll
work only if you have control of the entire network. The reason is that
it is completely up to the administrator of the network to choose which
application key to bind to the health server models. There's also no
guarantee that all health servers on the network are bound to the same
key. Even if you knew which keys the servers are bound to, each model
can only be bound to one key, so your health client could only talk to
health servers that are bound to the same key as the client.
There's also the issue with the small window between power up and not
having received responses from all nodes yet. Would we ignore any other
messages until we receive the Health Fault Get response? What if one of
the nodes in the RPL is powered off at that moment? This is quite likely
since the local node had just experienced a power cycle as well.
Additionally, this would constitute a layer violation, since the RPL is
handled on the network and transport layers, which have no knowledge of
model layer information.