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

Re: no NETCONF WG meeting planned for Paris



David B Harrington wrote:

Hi Andy



There are some pretty powerful features within a
decent transaction model to act on that bucket of XML content.
I know some of us on the design team we're really just
trying to eliminate screen-scraping the CLI with 1.0.
Even if NETCONF is just used as a programmatic interface
to the CLI at first, this is better than screen-scraping.



And I strongly support that vision. Screen-scraping needs to go.


But we also need to move past the constraint that any operator must be
able to lock the whole configuration to modify it, and a large number
of other issues that you have collected during the development of
netconf 1.0. Just changing from screen-scraping to XML isn't enough to
meet operators' needs, even if it is a big step in the right
direction.



I disagree. The 2 main deferred features -- partial locking and notifications -- are both simply speed optimizations, which impact a relatively small number of use cases. We don't have any serialization in SNMP or CLI at all! Concurrent writes can clobber config now, but this doesn't seem to be a big problem. I'm not convinced the lack of locked concurrent writes in NETCONF is a problem yet.

Notifications essentially provide a speed improvement over polling
techniques.  Currently, many CLI applications use SNMP notifications
and SYSLOG for this purpose.  The convenience of receiving async
messages within a NETCONF session is probably balanced out by
the increased localized cost of the NETCONF implementation, and
most likely does not represent a significant percentage of either
the implementation or deployment complexity.

IMO, the lack of these deferred features has no significant impact
on the usefulness of NETCONF 1.0.

As for NETMOD -- I don't see much hope of a WG agreeing on the
one true NETCONF data modeling language.  I agree that would
be the ideal from a standards-POV, but not realistic or even
close to operationally optimal.  E.g., A scheme tuned for SNMP
integration isn't going to work well for CLI-oriented configuration.

Andy




-- to unsubscribe send a message to netconf-request@ops.ietf.org with the word 'unsubscribe' in a single line as the message text body. archive: <http://ops.ietf.org/lists/netconf/>