RE: [YANG] new pyang errors

["Bert Wijnen" <[e-mail]>]

Inline

Bert Wijnen 

> -----Oorspronkelijk bericht-----
> Van: Balazs Lengyel [mailto:[e-mail]]
> Verzonden: vrijdag 25 januari 2008 12:19
> Aan: Ladislav Lhotka
> CC: [e-mail]
> Onderwerp: Re: [YANG] new pyang errors
> 
> 
> Mandatory in YANG means mandatory to configure.
> Today YANG considers all mandatory and optional elements 
> mandatory to implement.
> 

OK, I like that. Can you make that statement explicit in the
draft, so everyone understands the intention.

> Later we will probably define a conformance mechanism, then YANG 
> will be able to say:
> 
> module foo defines 100 leafs and these 85 are mandatory to 
> implement and the rest is optional.
> 
> We consider this conformance mechanism to belong to step to in 
> the DML. AFAIK it is not really used in SNMP either.

In some MIB modules we actually have pretty good MODULE-COMPLIANCE
statements that try to make it clear which objects must be implemented
and what their minimum implementation needs to support. Such MIB modules
often have multiple such module-compliance statements if that makes
sense for different environments. 

Other MIB modules just have one big MODULE-COMPLIANCE statement that
- requires to implement all objects
- but leaves it open to the implementation to implement them
  read-only or read-write. 
Supposedly that can be reported by the agent via a AGENT-CAPABILITIES,
(in which one can also report deviations) which results in an OID 
that one can store in the sysOrTable and so a manager can pick up the 
compliance that an agent claims and then supposedly can read the
AGENT-CAPABILITIES statement and so it can determine (in machine
readable form) what exactly is supported by the agent.

Well, at least that was/is the theory.

But other than during early SNMP interoperability tests in the mid-1990s
I have never seen it used much (if at all). 
So we should learn from this and think hard before we fall into the
same trap of designing mechanisms that no-one uses.

Bert
> (We could start using mustImplement or 
> something similar as a term for 
> mandatory to implement.)
> 
> Balazs
> 
> Ladislav Lhotka wrote:
> > Phil Shafer píše v Pá 25. 01. 2008 v 01:14 -0500:
> >> "Randy Presuhn" writes:
> >>> In the SNMP/SMI worlds, the "optionality" Ladislav refers to is
> >>> handled by conformance statements, rather than the model per se.
> >> Ah so maybe the confusion is "mandatory to implement" versus
> >> "mandatory to configure".  YANG's mandatory statement is the latter.
> > 
> > Oh yes, this is the Pudels Kern :-) I am sorry I wasn't able to express
> > myself so concisely. However, I still think Sec. 7.6.4 in the YANG draft
> > can be understood as "mandatory to implement": >>If "mandatory" is
> > "true", the node must exist in a valid configuration if its parent node
> > exists.<< I guess it's again due to the different meanings of validity.
> > 
> > Lada
> > 
> >> YANG has no conformance mechanism.  Whether it appears in YANG-2.0,
> >> I can't predict, but I'm hoping we can make base models that devices
> >> can implement, with augmentations for additional config.
> >>
> >> Thanks,
> >>  Phil
> >>
> >>
> >> _______________________________________________
> >> YANG mailing list
> >> [e-mail]
> >> https://www1.ietf.org/mailman/listinfo/yang
> 
> -- 
> Balazs Lengyel                       Ericsson Hungary Ltd.
> TSP System Manager
> ECN: 831 7320                        Fax: [phone]
> Tel: [phone]     email: [e-mail]
> 
> 
> _______________________________________________
> YANG mailing list
> [e-mail]
> https://www1.ietf.org/mailman/listinfo/yang
> 



_______________________________________________
YANG mailing list
[e-mail]
https://www1.ietf.org/mailman/listinfo/yang