Minutes IETF97: lime
minutes-97-lime-00

Minutes: Rob Wilton

Frank Brockners:: There is a mistake in the slides - Connection oriented doc is
further along. Ron Bonica: agreed, it is a mistake. Greg Mirsky: Work on Packet
Switch Network Connection oriented and connectionless abstract model. Could
factor out common functionality between the two models. Benoit: Confused - goal
of LIME was to get one YANG model, then need to split into two, no want to go
back to the abstraction.  What am I missing. Ron: If you can do it quickly,
then show me quickly, otherwise ... Benoit: Conclusion was that it was not
possible to have a common abstraction. Qin Wu: Discussion to split from one
draft into several drafts was based on WG consensus. Greg: Initial draft came
from Trill model, which represents connection oriented.  Didn't split it, but
re-designated it.  At the time of the discussion, although we are are working
on technlogy specific technologies, agreement was to have a common abstraction.
Ron: How to describe the abstraction layer in YANG in 4 weeks, then the WG can
adopt it, otherwise missed opportunity. Greg: I sent a proposal before Berlin,
but didn't get any response. Qin: I don't follow Greg's proposal.  Abstraction
is already built into the models, and have already tried to achieve this. Greg:
In correction oriented and connectionless networks, the topology could be
described by the same model, but OAM is differnet and can't be described in
this way.  Some distinct differences in the models which is why we agreed to
have two separate models. Ron: Please make the propasal in end on the WG list. 
Cut off is Dec 14th. Benoit: If abstractions are just end points then that
could just be typedefs.  Shutdown by IETF 99.  Having been hearing that the
models are almost ready for so long then we need to get them finished. Greg: We
already have YANG models that shouldn't be impacted by LIME models.  It would
be better that LIME TWAMP, BGP, MPLS LSP TE ??? Ron: If these OAM models
already exist then what value does LIME add Greg: LIME adds a common
abstraction. Ron: Considering existing models.  How does the LIME abstraction
Greg: LIME model can augment BFD. Ron: What you are really doing is turn LIME
on its head.  If LIME augments other protocols with something that already has.
Greg: LIME is the model that allows the operation to execute or instiante
certain network OAM functionality. Qin: Based on clarification from Greg, it
seems like a contradiction, I feel confused. Ron: Any other comments?

Moving on to the connection less update (Michael Wang presenting)::
Greg: Current state of connection less model does not address my comments
during adoption call.  I believe that the augmentation of  technology specific
models such as BFD. Michael: What to see a more abstract generic model.  Don't
want to describe nots of parameters in OAM specific YANG model. Greg: If you
don't want to be tied to specific technology, why doesn't the contain the list
of RFC xxx (tbc).   2. if there are other mechanism to do continuity check then
I would like to understand what you mean. Greg: But you listing BFD single hop,
BGP multi-hop, etc. Frank: Is the proposal to remove the refences on the
specific models (i.e. the RFCs), the person needs to know what the RFC is. Rob:
It doesn't make sense if LIME augments the specific technology modules because
that requires devices have to implement all of those technologies. Greg: Scope
of LIME is only for IETF OAM protocols. Lada: Concerned that schema mount is
using in places where it shouldn't be.  Schema mount is useful when the same
YANG model is used in different places.  If you want to augment then you can't
augment in both places.  Hence, adding technology specific then augment is a
better choice.  Confused about the groupings Michael: Groupings are taken from
the technology specific modules. Lada: His conclusion is to an augment. Greg:
The augmention has to be part of the base model. Michael: LIME model is only
providing common subset. Greg: The module lists numerous RFCs,  If you can't
use existing methods then what is the value of the LIME model Benoit:
Connectionless and BFD, which model augments which model? Qin: Import topoogy
and network instance model, don't import from BFD model. Benoit: Want LIME
model augmented with technology specific models. Ron: Ultimate value of LIME is
to provide a common base that other OAM models that augment. Lada: If you have
modules that are used already, you could possibly use schema mount. Benoit:
Charter is about consistent configuration, reporting, and state.  BFD are not
going to use the same hierarchy.  For BFD, what is the value?  Perhaps the
consistent reporting? Frank: Too late for BFD or TRILL, other OAMs might be
willing to use it. Benoit: OK, but if none of those OAMs are considering using
LIME. Jeff: BFD does allow for configuratoin by other means (e.g. OSPF can
enable BFD sessions, LIME could do the same thing). Ron: Summarize: Window of
opporunity has gone for BFD and TRILL, lets work fast and get it done. Qin:
Window for TRILL might still be open. Lada: If LIME proves to be useful, then
the BFD module could be modified.  Some models such as BFD don't fit in. 
Trying to convince people that the framework is useful.

Michael is presenting on connection orientated model:
Greg: Several comments between this model and MPLS-TP model.  See this as a
TRILL model.  Has stale reference to MPLS-TP.

Ron: Proposed next steps: Two big obstacles:
Greg has the action to produce some text by Beethoven's birthday, and decide
what to do what this. Second question is regarding augmention.  Proposed
solution is that LIME model won't reference any RFCs, but will ask other
working groups to augment LIME. Greg: Ask other WGs to review LIME model Jeff
Haas: BFD YANG module has to be self contained, so it can't augment LIME.  IETF
technologies are losely coupled. Greg: OAM has often being regarded as a lonely
child. Jeff: LIME has changes in building  a common base model.  Missing a
common OAM layer.