Re: Methods, Inheritance, Exceptions, etc. (was: Re: Methods in SMIng ?)

["Joel M. Halpern" <joel@longsys.com>]

Frank, I beg to differ.
What is being sent on the wire is limitted by the existing protocols.  As 
such, the request is sent as a "SET".  It is currently represented as a set 
of elements to be set because that is what the tools can talk about.
What was intended by the user was "remote entity: pig second remote 
entity".  From the users point of view, there was no interest in setting 
any attributes.  The user wanted a behavior to take place.  And there are 
many of these scattered through MIBs, most of them with really bad 
descriptions that do not make it obvious taht setting element X will cause 
the device to perform operation Y, based on other elements A, B, and C.
What I want is the signature and the syntax to make it clear to the reader 
that this is a method to be invoked rather than just a set of objects with 
obscure side-effects.
And, given that there are ways to map this stuff to existing MIBs (and 
probably PIBs), we should not wait for upgrades to the current management 
protocols.  When it was necessary that I read almost every MIB that came 
out for 4 years, these sort of oddities really caused me grief.

Yours,
Joel

At 11:10 AM 5/3/01 +0200, Frank Strauss wrote:
>In both of your examples there are SNMP Set operations that have to
>take place to launch the operations. What happens on the wire is an
>SNMP operation, not a method.
>
>If we want to model signatures that ease usage of these operations
>on the manager side, I would call it procedures. In your examples,
>these procedures might be relatively simple.