[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Methods in the NIM requirements

At 01:33 PM 5/1/00 -0400, you wrote:
>I have not seen a convincing explanation of how NIM will model the
>existing sets of attributes and protocol commands. I think that is what
>Walter is trying to get from you and Andrea and others, who seem to be
>ignoring the existing bottom-up body of work. The abstractions used in
>the NIM should be representative of the existing data models, such as
>SNMP and LDAP, not just soemthing pulled from blue sky theories.

There are lots of configuration parameters in SNMP which can be easily 
modelled in NIM.  That is not the question.  There are lots of statistics 
which are gathered in SNMP which may or may not belong in some form in 
NIM  -- lets defer that till after we get past this point, as it has little 
impact on the current discussion.

There are a number of SNMP objects which exist as mechanisms to represent 
operations in a pure attribute syntax.  If one is trying to represent those 
operations in a Network Information Model (when they belong in the model) 
then I would hope that one would use operations / methods rather than 
attributes.  The mapping of said model to SNMP will involve mapping back to 
the attributes.  That is equally true for any new methods we create in the 
model.  But we should not be restricting the model capabilities just 
because the common limited practice was a pure attribute modeling tool.

To put it another way, the description of an information model I read on 
another (policy?) list implied that the definer of an information model 
understood and expected that there was a necessary mapping step to use the 
information model with a specific implementation tool.  The fact that to 
implement some or all of the model in SNMP will require mapping to 
attributes does not mean that we should refuse to use methods.

While I can not construct a strong example of what methods the information 
model requires, I am very much of the opinion that the modeling tool must 
include support for methods.  For most attributes, read / write attribute 
semantics will suffice, but for some cases methods with signatures and 
semantics will be necessary.  I rather hope and expect that that is one of 
the place where the power of inheritance will come into play as 
well.  Otherwise it is purely a textual convention.

Joel M. Halpern