RE: SNMP improvements

["Harrington, David" <dbh@enterasys.com>]

Hi,

SNMP has never been good at configuring an entire device because the
mibs needed to do so are not available (except in rare instances), and
the efficiency characteristics of the SNMP protocol don't map well to
full-box configuration needs.

I think SNMP will continue to be used for monitoring and notifications,
for some incremental provisioning and operational adjustments, and for
troubleshooting some aspects of device operations. 

CLI scripting is much better suited to full-box configuration, but
suffers from a lack of standardization and a consistent programmatic
interface. In my view, the proposed design of netconf focuses on
standardizing the protocol aspects of a full-box scripting-like approach
to configuration, providing a programmatic interface, and, over time,
standardized data modeling.

So netconf is focused more on replacing (or standardizing) a CLI
scripting style of configuration, not on replacing SNMP. At least that's
my view of where things are going.

My $.02

dbh

> -----Original Message-----
> From: chintan sheth [mailto:shethch@yahoo.com] 
> Sent: Thursday, September 18, 2003 10:48 AM
> To: Wes Hardaker; Andy Bierman
> Cc: Wes Hardaker; Wijnen, Bert (Bert); Harrington, David; 
> j.schoenwaelder@iu-bremen.de; Eduardo Cardona; mibs@ops.ietf.org
> Subject: Re: SNMP improvements
> 
> 
> So are we going on the path where netconf protocol is
> expected to replace SNMP completely? For SNMP lovers
> this won't be good news.
> 
> thx,
> chintan
> 
> --- Wes Hardaker <wjhns1@hardakers.net> wrote:
> > >>>>> On Wed, 17 Sep 2003 22:10:34 -0700, Andy
> > Bierman <abierman@cisco.com> said:
> > 
> > Andy> Also, I don't understand why "fixing" SNMP is
> > so important
> > Andy> anyway.  The industry has decided SNMP is good
> > for monitoring
> > Andy> and notifications, and it does a great job in
> > both cases.
> > 
> > I agree that its good for monitoring and
> > notifications, but I disagree
> > it does a great job in both cases.  It still suffers
> > problems, like
> > bulking as Dinakaran pointed out.  It also suffers
> > from lack of
> > current data types, a lack of security that matches
> > current security
> > deployment sceneries, ...  It *has* been good and
> > continues to be good
> > for many situations.  It does, however, suffer from
> > all the problems
> > that started EOS and SMIng in the first place.  Some
> > of those
> > problems, like dealing with hierarchal
> > configuration, can be thrown
> > out since netconf might take care of them (note I
> > only say might
> > because it still needs to be proven in an
> > interoperable fashion
> > [juniper has proven it can work in a
> > non-interoperable way which is a
> > huge start]).  I still believe we need to fix SNMP
> > in a few cases, and
> > I agree that the number of cases has hopefully been
> > reduced.  If,
> > however, we expect it to be used in the future along
> > with netconf, as
> > you suggest, then it could use some minor revamps.
> > 
> > -- 
> > "In the bathtub of history the truth is harder to
> > hold than the soap,
> >  and much more difficult to find."  -- Terry
> > Pratchett
> > 
> 
> 
> __________________________________
> Do you Yahoo!?
> Yahoo! SiteBuilder - Free, easy-to-use web site design software
> http://sitebuilder.yahoo.com
>