Andy Bierman writes:
Andy> This doesn't mean lots of agent code couldn't be better layered
Andy> -- it just means there are plenty of protocol specific details
Andy> that make the 'one piece of code' goal unrealistic.
I continue to believe that proper layering of access mechanisms on a
central resource manager is the correct approach - some access
mechanisms might be more troublesome, others less.
The problem lies more with the organization of information. Whereas all
interface information might well be suited for the IF-MIB that very same
organization does not work for configurations. One exemplar reason for
this is templating. Another has more to do with seemingly unrelated
physical device characteristics at the abstract layer that are actually
intimately related in implementation.