Re: Blog: YANG Really Takes Off in the Industry
2014-12-09 09:07:21
Hi Dan,
Hi,
Success comes sometimes together with its own risks. In the case of
YANG, the risk is that six months or one year from now people who
enthusiastically engaged with developing YANG module will have at hand
bad modules that do not meet their expectations of functionality,
quality, interoperability. The pushback can then follow at the same
intensity as the current embracing.
This is exactly my concern, and why we have a limited amount of time to
organize ourselves.
To avoid this situation we need to provide answers to the problem of
non-scalability of the 'Doctors' team model which is IMO real. The
answer cannot be just 'let us focus on what we do in the standards
track in the IETF' because much of the development of YANG modules
happens out of the IETF and this is how we want it to be. We need to
encourage the open source and SDOs that engage on the YANG path to
develop their own Yangathon events and maybe in time their own YANG
doctors teams. We need however to make clear that this is the model we
recommend (rather than developing YANG modules and coming to the YANG
doctors to review) and offer help in organizing these first activities
and teams at their homes.
+1.
Regards, Benoit
Regards,
Dan
*From:*ietf [mailto:ietf-bounces(_at_)ietf(_dot_)org] *On Behalf Of *Ersue,
Mehmet (NSN - DE/Munich)
*Sent:* Monday, December 08, 2014 7:30 PM
*To:* ext Benoit Claise; Thomas D. Nadeau
*Cc:* IETF-Discussion list
*Subject:* RE: Blog: YANG Really Takes Off in the Industry
Hi Benoit, All,
to be able to reduce the necessary amount reviews and publications, I
assume we need to define some categories:
- YANG models might be interesting to have but they don't need to be
standardized in many cases. So, the category for Standard track RFC
has to be defined with a clear criteria.
- Others may be Informational or Experimental. However for these we
need to see the relevance to IETF work and necessity for publication.
The other categorization could be concerning the dispatching of YANG
models within IETF:
- There are particularly important YANG models related to a specific
WG. They should be developed in that WG.
- There might be others which are still important to publish but don't
have any home. I would discuss them in an OPSAWG meeting and decide
whether it should be published in OPSAWG or NETMOD WG
.
The other category I see is related to YANG models which should be
matched to another SDO. If we think a model is e.g. related to IEEE or
MEF, they should be done there and not at IETF.
There might be also other YANG models which are interesting but not
related to IETF. One particular category is concerning models, which
were originally Enterprise models but people would like to have an
IETF publication.
In general, over time we might need to be strict to stick to the
categories and the matched process per category.
That said, I would like to suggest to use the resources of the YANG
doctor team for models, which are essential for the IETF at the first
place, especially those seen under the category Standard track RFC.
Cheers,
Mehmet
*From:*ietf [mailto:ietf-bounces(_at_)ietf(_dot_)org] *On Behalf Of *ext Benoit
Claise
*Sent:* Monday, December 08, 2014 10:32 AM
*To:* Thomas D. Nadeau
*Cc:* IETF-Discussion list
*Subject:* Re: Blog: YANG Really Takes Off in the Industry
Hi Tom,
In light of the numerous YANG models these days, there is the YANG
doctors scaling issue. You're right, even if the number of YANG
doctors recently increased, we need other venues to provide advice to
YANG model designers. This should solve the issue of designing
properly the YANG models.
On the other hand, there is a bigger issue, IMO: the proper
coordination of those YANG models. This is not the YANG doctors
responsibility. This can't be: see the YANG doctors scalability
issue. So who's responsibility is this? Simply asserting "it's the
community responsibility" is the easy answer, but I'm afraid it will
not work.
Regards, Benoit
One of the things that came up in a number of discussions I had in
Hawaii and afterwards was around the coordination and
encouragement topics. A number of people commented both during
these discussions (and I think someone did during one of the
Netmod sessions) that the "MIB Doctor" model we are using is not
going to scale out to the numbers of Yang models that are in need
of advice or review, nor will be scale in terms of progressing
models through the IETF's RFC process. The fact is that we simply
do not have enough Yang Doctors to cover all of the models in
question, despite our best efforts. It is for this reason that I
strongly encourage other venues of review and advice such as a
continued "advice" or "Yangathon" session at each IETF meeting
going forward, as well as encouraging a loosening of the interim
WG meeting rules to encourage more meetings, as well as perhaps
less formalized ones. I also encourage the IETF to start pairing
up with other organizations such as OpenDaylight, Openstack and
OP-NFV and join their Yangathons there.
---Tom
On Nov 28, 2014:8:12 AM, at 8:12 AM, Benoit Claise
<bclaise(_at_)cisco(_dot_)com <mailto:bclaise(_at_)cisco(_dot_)com>>
wrote:
Hi Jari,
Let me open the discussion.
What is important at this point in time is the coordination of
those YANG models.
All of them come at the same time, and this required some
urgent attention.
Focusing on the routing YANG models with
"Rtg-yang-coord(_at_)ietf(_dot_)org"
<Rtg-yang-coord(_at_)ietf(_dot_)org>
<https://urldefense.proofpoint.com/v2/url?u=https-3A__www.ietf.org_mailman_listinfo_rtg-2Dyang-2Dcoord&d=AAMFAg&c=BFpWQw8bsuKpl1SgiZH64Q&r=I4dzGxR31OcNXCJfQzvlsiLQfucBXRucPvdrphpBsFA&m=EldXn4rzRf97Z5Y7A2KouM2hYYBbL5m9IMpucoilFZE&s=14IGX58lMwSyD8bOPQb_MqlsYeQ3u_s0T7YwNmjqqU0&e=>
is a step in the right direction. Indeed the community needs
to agree on how to model IGPs, BGP, the topology, etc...
However, the coordination should also occur with the data
models developed in other IETF WGs. And the IETF might need to
reach out to different SDOs/consortia.
As the operators told me: we can't afford to develop those
data models independently from each others.
Regards, Benoit
Thanks for writing this article, Benoit!
The wave of new data models is obviously interesting and exciting.
But I wanted to open a discussion with you all on what we should do with
regards to serving this need better. Is there something that we could do better
at the IETF to be able deal with this new work?
Jari
|
|