While I do not have a strong opinion, I do have a preference for a
_simple_ subscription mechanism as part of the protocol since I
believe having special verbs will make it more likely that
management
apps get to actually use the feature. The argument is clearly on the
irrational side of things how I believe many human programmers work
-
they usually tend prefer verbs over data structures and even purely
declarative approaches.
I am not an operator, so I cannot really state what operators would
prefer.
I am not a Netconf implenentor, so I cannot discuss implementation
issues.
I do not know XML or XSD very well, so cannot discuss
language-specific tradeoffs.
But I've heard complaints about SNMP for 13+ years, so I know about
that.
And I've been programming even longer, so I know about that.
One operator complaint is the data-centric rather than a task-based
approach of SNMP. People don't think in terms of "how should I
manipulate data sets to cause a side-effect to occur?", they think in
terms of "what do I need to do" to accomplish a task.
Programming languages migrated from binary to assembly to procedural
to object oriented designs.
Andy often complained about the peek-poke nature of SNMP. I associate
that approach with assembly language and its limited set of operations
that allowed one to peek/poke memory locations and registers.
I think a verb-based approach is associated with the more advanced
procedural style. I am of the impression that operators would prefer
to move from the antiquated peek-poke method of SNMP to a task-based
procedural approach (verb-based), akin to that used by CLI.
In a private email, somebody made this observation:
"In all my years of doing SNMP work, I never saw the
immediate uptake and commitment of real dollars to
an I-D, like I saw for netconf. It's because dev-groups
all over were already migrating their CLI to XML, and all
doing it differently. Netconf is all about CLI, not SNMP.
The SNMP people don't understand that yet. ;-)"
I think I understand that. If Netconf is all about CLI, then it
appears that support for a verb/task-based approach is called for,
rather than the data-centric peek-poke approach of SNMP.
Of course, with that decision comes the possibly-unconstrained growth
of verbs often found in CLIs.
But that's the price of progress, just as the unconstrained growth of
functions written in Fortran/Pascal/C/etc was the price one paid for
moving away from assembly language.
To be continued ...
dbh
--
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/>