Re: Closing on NIM requirements

[Jon Saperia <saperia@mediaone.net>]

on 04/17/2000 6:16 AM, Harald Tveit Alvestrand at Harald@Alvestrand.no
wrote:

>> Two points:
>> 
>> 1. Unless there must be some other discussion I have missed with regard to
>> SNMP and configuration - I must have missed something. Perhaps you can
>> enlighten me, short of that I am not sure I buy your last sentence.
> 
> I was thinking of informal talks that led up to the SNMPCONF bof in
> Adelaide; I wasn't at the BOF, so don't know how they concluded.
> At least before the BOF, it was thought to be (close to?) impossible to use
> SNMP with current MIBs to change from one consistent configuration to
> another without going through a lot of undesired intermediate states - no
> transaction semantics across multiple updates.
> 
The only BOF was the configuration BOF in Washington at the prior IETF. The
SNMPCONG Working Group met in Adelaide. One of the items in the charter is
the definition of a policy MIB module to make global changes in
configuration (e.g, policy) - there are other Objects that will be defined
as well. I think you should familiarize yourself with the current work. This
is thread is orthogonal to the point of the modeling language however except
to the extent that you feel it is not necessary to convert to the SMI.
> 
>> 2. Different tools can be 'a good thing' as long as they are not attempting
>> to do effectively the same thing. This is a discussion many of us on this
>> list have had quite a lot and my point is not to restart it. My point is
>> that the modeling language should/must take into account mechanisms that are
>> going to be used to realize functions. If they are deficient you fix them,
>> replace them, or change the language.
> 
> The most important thing is that the modelling language reflect reality.
> If the current tools are part of that reality - I agree with you.

Then we agree.
> If the model can serve as a basis for choosing the appropriate tool -
> perhaps I disagree.
> 
> 
/jon