RE: WG Review: NETCONF Data Modeling Language (netmod)

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

 



Hi Dave,

Good questions. Let me see if I can answer some of them.

For perspective, I have not been involved in the developoment of any
of the proposed technical directions, but I have been a general
technical commentator with 16 years of IETF NM experience ;-)

> -----Original Message-----
> From: ietf-bounces@xxxxxxxx [mailto:ietf-bounces@xxxxxxxx] On 
> Behalf Of Dave Crocker
> Sent: Wednesday, April 23, 2008 12:04 AM
> To: Eric Rescorla
> Cc: iesg@xxxxxxxx; ietf@xxxxxxxx
> Subject: Re: WG Review: NETCONF Data Modeling Language (netmod)
> 
> 
> 
> Eric Rescorla wrote:
> > Which is why it is now returned to the broader community for
> > additional perspectives from those not already committed to a
> > particular path

Dave, my impression of your questions is that "they" means the
"broader community" - those not already committed to a particular path
- that EKR references. I will answer your questions from that
perspective.

> 
> 
> Are they committed to doing the work?

I believe the answer to this is yes.

The Netconf community raised the potential need for a new data
modeling language because XSD was too human-unfriendly, and both XSD
and RNG lacked features needed for network management purposes. We
have performed multiple comparison exercises between XSD and RNG
(e.g., modeling Diffserv configuration), and all have fallen somewhat
short in terms of expressing the things the OPS area feels are
important to express, based on 20 years of experience with SNMP and
SNMPCONF and COPS-PR, and based on experience with CLI-based
configuration, and operator feedback about configuration requirements
exprsessed during the IAB Network Management Workshop in 2002 an dthe
subsequent "world tour" of NANOG, RIPE, and other operators' groups.

People from the broader community (especially the APPS area) with
experience in XSD and RNG came forth and prepared multiple concrete
proposals to compare data modeling language approaches. All of these
previous efforts have tried to be inclusive of the broader community,
but many have been unofficial meetings, so the broader community may
have been under-represented in some of these comparisons, but XSD and
RNG have been prominent proposals.

After multiple comparisons, the rough consensus of those involved was
that the results remained human-unfriendly, especially the XSD format,
and efforts at producing XSD schemas in WG documents had real
difficulties producing valid XSD. While RNG was more human-friendly,
it still was less human-friendly than desired.

Unfortunately, despite going to this effort, the CANMOD BOF was
prevented from actually comparing the various concrete proposals (the
"beauty contest"), which would have shown XSD versus RNG versus YANG,
relative to the stated requirements for network management purposes. 


> 
> Do they have their own constituency?
> 
The supporters of XSD have their own constituency. The supporters of
RNG have their own constituency. The supporters of the YANG proposal
have their constituency. And there are constituencies for other
proposals that have not been widely accepted.

Folowing a proposal for a BOF, the APPS area and some IAB members
wanted some extra input on the need for a data modeling language. A
design team composed of members of the OPS community and the APPS
community was created to document a set of requirements. The OPS
community had already been through this exercise multiple times
already, as documented in multiple existing RFCs on requirementsa for
configuration, and new requirements were allowed to be added to the
existing requirmeents by represntatives of the various consistuencies.

It was decided by the OPS ADs that concrete proposals should be
prepared for presentation and comparison at a BOF to compare
alternatice approaches. Multiple proposals were prepared, including
proposals from OPS area and APPS area people. These proposals were
prepared for a "beauty contest" becauser there was strong aoncensus
amongst the various constituencies that we needed a data modeling
language, and some felt that the existing XML-based schema languages
might be sufficient. The proposals, however, reflected the fact that
the existing languages fell short when trying to represent information
necessary for network management **based on operator input**. Existing
XML-based tools would be unable to validate the data models without
having specfic extensions provided through annotations, and requiring
modifications to existing tools to process those annotations.

At the CANMOD BOF, the "beauty contest" between proposals was not
allowed to be held, because certain members of the "broader community"
insisted that the question of whether the existing languages could
suffice be discussed even further, even though there was strong
consensus from the OPS community (and recently from the APPS
community) that the existing schema languages fall short of the
requirements for network management data modeling.

Following the CANMOD BOF, the constituencies from the OPS and APPS
areas came together at the OPS Open area meeting and at the OPS/IESG
breakout meetings to discuss a way forward, and the various
constituencies reached agreement that YANG provoded a reasonable
human-friendly language, and a translation to either XSD or RNG would
meet the needs for a machine-friendly language suitable for XML-based
tools. Consensus of these constituencies was that RNG + Schematron
(i.e., DSDL) provided better capabilities for validation of the
network management specific requirements than XSD alone or RNG alone.
The consensus of the these constituencies was also that YANG was more
human-friendly then either XSD or RNG or DSDL. 

So constituencies of the "broader community" have been very active in
establishing the proposed direction, in crafting the proposed charter,
and have committed to working on the documents identified in the
proposed charter. 

> Since the topic is not new, where have they been and why have 
> they not 
> developed their own group consensus?

This topic is not new, and constituencies of the "broader community"
have been active in establishing the proposed direction, and the
proposed charter, and have committed to working on the documents
identified in the proposed charter. 

> 
> Rather than "perspectives" where are the technical concerns 
> that Bert asked about?
> 
> d/
> -- 
> 
>    Dave Crocker
>    Brandenburg InternetWorking
>    bbiw.net
> _______________________________________________
> IETF mailing list
> IETF@xxxxxxxx
> https://www.ietf.org/mailman/listinfo/ietf
> 

David Harrington
dbharrington@xxxxxxxxxxx
ietfdbh@xxxxxxxxxxx
dharrington@xxxxxxxxxx


_______________________________________________
IETF mailing list
IETF@xxxxxxxx
https://www.ietf.org/mailman/listinfo/ietf

[Index of Archives]     [IETF Annoucements]     [IETF]     [IP Storage]     [Yosemite News]     [Linux SCTP]     [Linux Newbies]     [Fedora Users]