Re: [netmod] Text in import to indicate whether a module is needed as import-only or as implemented

[Balázs Lengyel <[e-mail]>]

Hello,

Putting this information on individual schema nodes can be a problem if a module uses imported items a lot. We have modules with 100+ schema nodes that use an import, while  there are only 7 modules imported. Adding the statement (mandatory/optional to implement) 100 times would be a lot of text. I also doubt that people will check all 100 places.

IMHO alternatives a) or b) are feasible. (I like a) the best)

Regards Balazs

 

From: netmod <[e-mail]> On Behalf Of Andy Bierman
Sent: 2020. március 3., kedd 17:14
To: Benoit Claise <[e-mail]>
Cc: NETMOD WG <[e-mail]>
Subject: Re: [netmod] Text in import to indicate whether a module is needed as import-only or as implemented

 

 

 

On Tue, Mar 3, 2020 at 12:45 AM Benoit Claise <[e-mail] <mailto:[e-mail]> > wrote:

Hi,

 

 

On Mon, Mar 2, 2020 at 10:23 AM Ladislav Lhotka <[e-mail] <mailto:[e-mail]> > wrote:

On Mon, 2020-03-02 at 09:52 -0800, Andy Bierman wrote:
> 
> 
> On Mon, Mar 2, 2020 at 8:57 AM Benoit Claise <[e-mail] <mailto:[e-mail]> > wrote:
> > Sorry to resurrect this old email thread.
> > To me, it's an important piece of information to know that ietf-netconf-acm
> > is optional to implement.
> > 
> > It seems that we have 3 potential places where to insert this information
> > 1. The associated document. We could and should insert it into the RFC text.
> >     Drawback: Somehow the YANG module is looked at independently of the
> > associated document
> > 2. import-stmt: people on the list apparently don't like this
> > 3. module description? What harm would it do if the description could
> > contain this info?
> > 
> > 
> 
> IMO it makes more sense to summarize the imported modules that need to be
> implemented
> and not mention the ones that are not required.  The module description-stmt
> is better
> than each import. (YANG 1.1 allows the same module to be imported multiple
> times).

Modules that have to be implemented can be imported only once. Adding this
information to the import statement for such modules is IMO more effective than
having it in the module description. I don't get how it could become a problem.

 

I do not think this info helps very much. It is duplicate info.

It should be in the description-stmts for the objects defined in the module.

Well, I had in mind the module description
The "description-stmts for the objects defined in the module" is yet another place.

 

 

The object descriptions are the correct level of granularity to be useful.

A boolean statement wrt/ module implementation is required vs not required is not very useful.

Rarely is the entire module either fully required or nothing to implement at all.

 

IMO it is wrong to conflate the import-stmt with conformance requirements.

It is merely a mechanism to bind local prefix usage to an external module.

Text like "revision X or derived from X is required" is appropriate for the import-stmt because

it refers to the binding between prefix and imported module revision used.

 

Consider how this extra documentation work scales to a big module like ietf-bgp.

There are multiple submodules, each with overlapping imports and each with their own module header.

 

 

Andy

 

It is also self-evident (and defined in YANG) that if you augment some external node

you have to implement the augmented module.

 

 This is the classic definition of a CLR.

 

Lada

 

 

Trying to decompose the problem space ....
I believe it's common sense to include in the RFC text that ietf-netconf-acm is optional to implement (to take the example in question in https://tools.ietf.org/html/draft-ietf-netconf-notification-capabilities-11)
Do we agree that this information should also be contained somewhere, somehow in the YANG module? I would say: yes

If yes, what are the options?
I agree with Juergen (in a reply in this email thread): ideally a tooling answer would be better. This will take some time. So what do we do now?
3 options:
a. import description (description-stmt in import-stmt) 

module ietf-system-capabilities {
  yang-version 1.1;
  namespace
    "urn:ietf:params:xml:ns:yang:ietf-system-capabilities";
  prefix sysc;
 
  import ietf-netconf-acm 
    { prefix nacm; }
    description "ietf-netconf-acm is optional to implement.";
    }


b. module description (description-stmt in  meta-stmts in module-stmt/submodule-stmt)

module ietf-system-capabilities {
  yang-version 1.1;
  namespace
    "urn:ietf:params:xml:ns:yang:ietf-system-capabilities";
  prefix sysc;
 
  import ietf-netconf-acm { prefix nacm; }
 
  import ietf-yang-library {
    prefix yanglib;
    description "Revision 2019-01-04 or a
      revision derived from it is REQUIRED.";
  }
 
  organization
    "IETF NETCONF (Network Configuration) Working Group";
  contact
    "WG Web:   <https://datatracker.ietf.org/wg/netconf/>
     WG List:   <mailto:[e-mail]> <mailto:[e-mail]>
 
     Editor:   Balazs Lengyel
                <mailto:[e-mail]> <mailto:[e-mail]>";
  description
    "This module specifies a module intended to contain system
      capabilities. System capabilities may include capabilities of a
      NETCONF or RESTCONF server or a notification publisher.
 
      ...
 
     
      ietf-netconf-acm is optional to implement."

c. description-stmts for the objects defined in the module

          leaf node-selector { 
            type nacm:node-instance-identifier;
            description
              "Selects the data nodes for which capabilities are
               specified. The special value '/' denotes all data nodes
               in the datastore. ietf-netconf-acm is optional to implement.";
          }


c. looks weird to me, and might need repetitions in different YANG objects. In the end, as an implementer, I only care to understand whether I need to implement "ietf-netconf-acm" as a prequesite to implement ietf-system-capabilities. 

Between a. and b., I have no strong preferences.
However, a. seems to be the logical place to me.

Regards, Benoit




>  
> > Regards, Benoit
> > 
> 
> Andy
>  
> > > 
> > > On Wed, Jan 8, 2020 at 5:44 AM Ladislav Lhotka <[e-mail] <mailto:[e-mail]> > wrote:
> > > > On Wed, 2020-01-08 at 04:49 -0800, Andy Bierman wrote:
> > > > > 
> > > > > On Wed, Jan 8, 2020 at 1:11 AM Ladislav Lhotka <[e-mail] <mailto:[e-mail]> > wrote:
> > > > > > On Tue, 2020-01-07 at 14:29 +0000, Balázs Lengyel wrote:
> > > > > > > If that is the consensus, I can remove the description statements,
> > > > no big
> > > > > > > deal. (I actually like the statements, but they are not important
> > > > for this
> > > > > > > draft)
> > > > > > 
> > > > > > Of course, it is not that important, but I don't see how this
> > > > information
> > > > > > could
> > > > > > be harmful, if it is included with the import. In my view, it is not
> > > > meant
> > > > > > as a
> > > > > > conformance requirement but just an info from the module author
> > > > about the
> > > > > > meaning of the import statement.
> > > > > > 
> > > > > 
> > > > > It adds a lot of extra work but more importantly the import-stmt is
> > > > the wrong
> > > > > place
> > > > 
> > > > What work do you mean? I thought that it would be just info for
> > > > potential
> > > > developers (or their managers) that implementing the module requires
> > > > implementing other modules and functionality - or not. 
> > > > 
> > > > 
> > > 
> > > 
> > > It is duplication because the individual data-def statements should have
> > > any notes
> > > about implementation requirements. The duplication may even be wrong.
> > > E.g., in the module it says NACM is not required, but what if some objects
> > > have NACM requirements listed in the Security Considerations section?
> > > That is the RFC section to discuss NACM requirements.
> > > 
> > > 
> > > > For example, if a module imports ietf-netconf-nacm only for using "node-
> > > > instance-identifier" type, it is relatively uninteresting. Otherwise,
> > > > implementation of the module may just be out of question.
> > > > 
> > > > 
> > > > > to document such a complex and unrelated issue as server conformance
> > > > > requirements.
> > > > > 
> > > > > 
> > > > > > The root of this problem (and design flaw of YANG, IMO) is that
> > > > import is
> > > > > > "overloaded" with two different purposes, one of which effectively
> > > > requires
> > > > > > that
> > > > > > the imported module be also implemented, while the other doesn't.
> > > > > > 
> > > > > 
> > > > > The import-stmt is only used to map a local prefix to an external
> > > > module.
> > > > 
> > > > But one thing is using a prefix for accessing top-level types and
> > > > groupings
> > > > (i.e. stuff in YANG modules), and another thing is accessing schema
> > > > nodes, which
> > > > have to be present in the schema tree. In complicated data models, it is
> > > > not
> > > > exactly easy to figure out all these dependencies.
> > > > 
> > > > 
> > > 
> > > I do not agree these are different things.
> > > In both cases the prefix is used to determine the imported module that
> > > contains the identifier.
> > > 
> > > 
> > > > So maybe what is really overloaded are the namespace prefixes: they are
> > > > used for
> > > > addressing YANG modules, schema nodes and instances (in XPath). 
> > > > 
> > > > Lada
> > > > 
> > > 
> > > 
> > > Andy
> > >  
> > > > > This proposal to add conformance info the the import-stmt would
> > > > overload it
> > > > > with
> > > > > another purpose.
> > > > > 
> > > > > Not even a typedef is easy to classify  (e.g., leafref requires
> > > > implementation
> > > > > of a data node).
> > > > > I certainly agree that YANG conformance is poorly specified, poorly
> > > > > understood, and
> > > > > in real need of improvement. Likewise, the import-stmt is also in need
> > > > of some
> > > > > improvement
> > > > > since import-by-exact-revision is (and has always been) poorly
> > > > designed.
> > > > > 
> > > > > 
> > > > > > Lada
> > > > > > 
> > > > > > > Balazs
> > > > > 
> > > > > Andy
> > > > > 
> > > > >  
> > > > > > > -----Original Message-----
> > > > > > > From: Martin Bjorklund <[e-mail] <mailto:[e-mail]> > 
> > > > > > > Sent: Tuesday, January 7, 2020 2:38 PM
> > > > > > > To: Balázs Lengyel <[e-mail] <mailto:[e-mail]> >
> > > > > > > Cc: [e-mail] <mailto:[e-mail]> ; [e-mail] <mailto:[e-mail]> ; [e-mail] <mailto:[e-mail]> 
> > > > > > > Subject: Re: [netmod] Text in import to indicate whether a module
> > > > is
> > > > > > needed as
> > > > > > > import-only or as implemented
> > > > > > > 
> > > > > > > Hi
> > > > > > > 
> > > > > > > I agree w/ Andy and others that we should not add this to the
> > > > import's
> > > > > > > description.  I don't think this kind of conformance text belongs
> > > > to the
> > > > > > > import's description.  If you think it is important to state this,
> > > > the
> > > > > > best
> > > > > > > place is probably as plain text in the document itself.
> > > > > > > 
> > > > > > > 
> > > > > > > 
> > > > > > > /martin
> > > > > > > 
> > > > > > > 
> > > > > > > Balázs Lengyel <balazs.lengyel=[e-mail] <mailto:[e-mail]> >
> > > > wrote:
> > > > > > > > As a draft author who was asked to add text about the imports
> > > > IMHO
> > > > > > > > 
> > > > > > > > *   it would be easy for me to remove the description from the
> > > > import.
> > > > > > > > Actually I really just want to know what is acceptable to the
> > > > group, so
> > > > > > I
> > > > > > > > can proceed
> > > > > > > > *   I also think that adding this text is in most cases easy and
> > > > it does
> > > > > > not
> > > > > > > > need updates later.
> > > > > > > > *   The rules in some cases might not be trivial.
> > > > > > > > 
> > > > > > > > *   Imported YAMs need to be implemented if
> > > > > > > > 
> > > > > > > > *   Imported parts are included in Xpath (augment, when, must,
> > > > require-
> > > > > > > > instance)
> > > > > > > > 
> > > > > > > > *   Imported YAMs do not need to be implemented if only the
> > > > following
> > > > > > are
> > > > > > > > used
> > > > > > > > 
> > > > > > > > *   Types
> > > > > > > > *   Features
> > > > > > > > *   extensions
> > > > > > > > 
> > > > > > > > *   Ambiguous if
> > > > > > > > 
> > > > > > > > *   groupings are used
> > > > > > > > *   if the dependency is not formally defined by YANG, but
> > > > functionally
> > > > > > > > needed. (E.g. notification-capabilities does not formally need
> > > > YANG-
> > > > > > > > Push 
> > > > > > to
> > > > > > > > be implemented, however there is no sense in defining
> > > > capabilities if
> > > > > > YANG-
> > > > > > > > Push is itself not implemented.)
> > > > > > > > *   deviation ?
> > > > > > > > *   other cases ?
> > > > > > > > 
> > > > > > > > Regards Balazs
> > > > > > > > 
> > > > > > > >  
> > > > > > > > 
> > > > > > > > From: netmod <[e-mail] <mailto:[e-mail]> > On Behalf Of Andy Bierman
> > > > > > > > Sent: 2019. december 19., csütörtök 17:23
> > > > > > > > To: Ladislav Lhotka <[e-mail] <mailto:[e-mail]> >
> > > > > > > > Cc: NetMod WG <[e-mail] <mailto:[e-mail]> >
> > > > > > > > Subject: Re: [netmod] Text in import to indicate whether a
> > > > module is 
> > > > > > > > needed as import-only or as implemented
> > > > > > > > 
> > > > > > > >  
> > > > > > > > 
> > > > > > > >  
> > > > > > > > 
> > > > > > > >  
> > > > > > > > 
> > > > > > > > On Thu, Dec 19, 2019 at 8:00 AM Ladislav Lhotka <[e-mail] <mailto:[e-mail]> 
> > > > <mailto: <mailto:%0b> 
> > > > > > > > [e-mail] <mailto:[e-mail]> > > wrote:
> > > > > > > > 
> > > > > > > > On Thu, 2019-12-19 at 07:52 +0000, Schönwälder, Jürgen wrote:
> > > > > > > > > On Thu, Dec 19, 2019 at 08:23:27AM +0100, Ladislav Lhotka
> > > > wrote:
> > > > > > > > > > I don't see how YANG syntax defines this. If a module
> > > > imports 
> > > > > > > > > > ietf-netconf- acm, it could be because
> > > > > > > > > > 
> > > > > > > > > > - it just uses a typedef, such as "node-instance-
> > > > identifier", and
> > > > > > then
> > > > > > > > > >   ietf-netconf-acm needn't be implemented (but can be),
> > > > > > > > > > 
> > > > > > > > > > or
> > > > > > > > > > 
> > > > > > > > > > - it augments ietf-netconf-acm, which makes sense only if
> > > > the latter
> > > > > > > > > >   module is implemented.
> > > > > > > > > > 
> > > > > > > > > > It it the YANG library that specifies whether a module is 
> > > > > > > > > > implemented or not, but the "import" statement itself
> > > > doesn't tell
> > > > > > you
> > > > > > > > > > anything.
> > > > > > > > > > 
> > > > > > > > > 
> > > > > > > > > Can we not assume that an implementor will figure out the
> > > > difference?
> > > > > > > > 
> > > > > > > > An implementor should be able to figure it out, but other
> > > > potential 
> > > > > > > > module users may not. For example, if somebody is evaluating
> > > > whether 
> > > > > > > > to use a module for their device or not, it is important to know
> > > > that 
> > > > > > > > NACM has to be implemented (or not).
> > > > > > > > 
> > > > > > > >  
> > > > > > > > 
> > > > > > > > You seem to be talking about a new conformance documentation
> > > > procedure
> > > > > > > > 
> > > > > > > > that attempts to solve the problem "to use modules A, B, and C 
> > > > > > > > together
> > > > > > > > 
> > > > > > > > to achieve some functionality X, all these conditions need to be
> > > > met"..
> > > > > > > > 
> > > > > > > > (Sounds more like a problem for YANG Packages to solve)
> > > > > > > > 
> > > > > > > >  
> > > > > > > > 
> > > > > > > > IMO this is a much harder problem than something that can be
> > > > solved
> > > > > > > > 
> > > > > > > > with extra description-stmt text. The writer is likely to get it
> > > > wrong 
> > > > > > > > or not
> > > > > > > > 
> > > > > > > > keep it up to date, so a tool to analyze the file seems like a
> > > > better
> > > > > > > > solution.
> > > > > > > > 
> > > > > > > >  
> > > > > > > > 
> > > > > > > > Lada
> > > > > > > > 
> > > > > > > >  
> > > > > > > > 
> > > > > > > >  
> > > > > > > > 
> > > > > > > > Andy
> > > > > > > > 
> > > > > > > >  
> > > > > > > > 
> > > > > > > > 
> > > > > > > > > Or someone writes a pyang plugin to determine from the schema
> > > > tree 
> > > > > > > > > the kind of imports there are (for a given set of features).
> > > > > > > > > 
> > > > > > > > > /js
> > > > > > > > > 
> > > > > > > > --
> > > > > > > > Ladislav Lhotka
> > > > > > > > Head, CZ.NIC Labs
> > > > > > > > PGP Key ID: 0xB8F92B08A9F76C67
> > > > > > > > 
> > > > > > > > _______________________________________________
> > > > > > > > netmod mailing list
> > > > > > > > [e-mail] <mailto:[e-mail]>  <mailto:[e-mail] <mailto:[e-mail]> > 
> > > > > > > > https://www.ietf.org/mailman/listinfo/netmod
> > > > > > > > 
> > > > -- 
> > > > Ladislav Lhotka
> > > > Head, CZ.NIC Labs
> > > > PGP Key ID: 0xB8F92B08A9F76C67
> > > > 
> > > > _______________________________________________
> > > > netmod mailing list
> > > > [e-mail] <mailto:[e-mail]> 
> > > > https://www.ietf.org/mailman/listinfo/netmod
> > > > 
> > > 
> > > 
> > > _______________________________________________
> > > netmod mailing list
> > > [e-mail] <mailto:[e-mail]> 
> > > https://www.ietf.org/mailman/listinfo/netmod
> >  
-- 
Ladislav Lhotka
Head, CZ.NIC Labs
PGP Key ID: 0xB8F92B08A9F76C67