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

Re: snmpconf taxonomy discussion(s) - suggested resolution

Hi David...

At 12:31 PM 8/17/2001 -0700, David T. Perkins wrote:
>If the taxonomy is removed, the resulting document, is not a BCP and
>needs much work.

Hmmmm....do you mean that the inclusion of the taxonomy in and of itself is 
the crux of making it suitable in your opinion?  Your next statement makes 
me wonder if you're trying to identify broader concerns you have (but that 
doesn't jibe with how I parse your prior statement :).

>  I work closely
>with one of the co-authors (Mike MacFaden) and have pointed out to him
>many problem areas. He agrees that there are problems. He is vacationing
>this week and should be back on Monday. We can discuss an action plan
>when he returns.

Unless he's changed his mind since last week, he's in favor of excising the 
taxonomy section given where we're at now.

>On the SNMP Policy issue, I've been reading all that I can about policy.
>(It's not too entertaining (sorry Bob Moore, Andrea, etc), so it is taking
>a long time to get through.) I believe that it is very important to
>take the existing documents and provide an SNMP-based interpretation
>of them.

Well, a very strong objection has been raised to the limited degree to 
which the taxonomy attempted to do just that, which is one of the major 
reasons we're chucking it.  There is not a great degree of experience we 
can speak of in policy-based-configuration-using SNMP (the qos device and 
network information model documents do not, to the best of my knowlege, 
describe a shipping implementation of anything in commercial use, but a 
proposed pcim-consistent information architecture with qos as its primary 
domain of current applicability).  So, said objection-holders would say, 
how can a discussion of policy here be a "best current practice"?  Based on 
extensive discussions I've had with Mike McF. in the preparation of this 
document, I believe he was struggling with that very view himself.

On the other hand, I'm not sure the PCIM was an RFC at that point 
either...but based on discussions with him in London, I know Mike was an 
adherent for removing the section in the interests of focus and consensus.

Inasmuch as IETF-based SNMP policy configuration technologies are 
concerned, PM is it at this point in history.  Since the snmpconf pm draft 
is being advanced at the exact same time as the BCP, one can definitely 
argue (and they have, to me) that calling on it for experience of a "best 
current practice" is inflating the term more than a little.


P.S.: By the way, the reason I found the policy fwk documents 
excruciatingly difficult to get through was the lack of a roadmap--they 
require a quite recent understanding of diffserv, and the DMTF network 
information model deepthink documents are pretty required reading.  Even 
then it's difficult to find the right balance of essential theory and 
direct applicability in the policy wg docs as you read through them (I 
needed a cold shower after the PCIM).  The only way I found to overcome it 
was to do a complete loop of reading everything *twice*, and I'm still just 
conversant enough to be dangerous in nondiscriminating crowds. :)