On Monday 22 March 2010 13:26:12 Josef Reidinger wrote:
Both solution use notifications. How this works? Simple you use in rest-service your status module and use static method notify problem. You can notify that problem appear or disappear.
I just discussed the issue with Stefan and we found another problem with this
approach:
The notification approach might sound nice, but it reverses the way that
information flows for the status information. Stefan designed it the way, that
all status information is queried for. The notification approach relies on the
fact that the module puts its status information somewhere. That means, that
the module must have run bevore. But we can not guarantee that.
Easy example: a module that needs to be was not or was skipped (like
registration maybe). There is no chance for the module to put some information
if it did not run.
So the status information needs to be queried for - only this assures that the
module itself at least ran once and the module itself gathered all information
relevant for the status information. And it assures that the information is
uptodate.
A status information might even change without any change on the system itself
- the registration is a good example: When we in the future fix the
registration protocol and can get more information from NCC then the
registration state can change with a change on the registration server -
without any change on the appliance.
Ciao,
Daniel
--
J. Daniel Schmidt