Hi -
On 2020-12-29 12:31 PM, Andy Bierman wrote:
...
I am not that interested in lessons from the past because it just sounds
like an old-timer's negativity. But I am not interested in boiling the
ocean either.
There is an invalid assumption with ECA that the current client
development logic
(which is completely unbounded in its complexity, purpose, side effects,
design,
or server interaction model) can somehow be replaced with a highly
constrained
YANG model abstraction.
IMO a script-language-based solution is needed.
The ECA framework is good progress, but Programming in YANG is not.
However the IETF is not well-suited for standardizing programming languages
Open-Source is a much better solution path for this approach.
The solution also has to do a great job of gluing the model components to
the ECA logic and actions so there is maximum reusability and scalability.
Most importantly, it has to be easy for an operator to program and debug
ECA.
There aren't going to be any magic tools to hide the complexity.
There's a tension between the convenience of scripting languages and the
desire to be able to reason about how policies interact, particularly if
one is approaching the problem from a perspective of goal-oriented
management. To me that suggests that more structured (than mere
scripting) mechanisms are needed to to identify what aspects of what
resources are inputs or outputs in the operation of a given policy.
(That also has the nice side-effect of making policy formulations
vastly more re-usable.)
For the ISO/ITU work, my recollection is that the direction in which
it was headed was to use facilities of the management information
model and protocol (e.g. scoping, filtering, allomorphism, attribute
re-use, etc.) for the parameterization, but to use scripting for the
policy's "body". I don't know if that's how things ultimately turned
out; I've been away from the work for decades.
If all the semantics, including specification of applicability,
are embodied in the script itself, one ends up with something
like the script MIB. A potentially useful thing, but not particularly
helpful if one's goal is to be able to reason about policy interaction
or to formulate things in terms of goals.
Randy
_______________________________________________
netmod mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/netmod
