Re: [core] draft-bierman-core-yid-00 open questions

[Andy Bierman <[e-mail]>]

On Wed, Aug 24, 2016 at 2:09 PM, Michel Veillette <Michel.Veillette@
trilliantinc.com> wrote:

> Hi Andy
>
>
>
> My comments inline [MV]
>
>
>
> *From:* Andy Bierman [mailto:[e-mail]]
> *Sent:* Wednesday, August 24, 2016 4:48 PM
> *To:* Michel Veillette <[e-mail]>
> *Cc:* [e-mail] WG <[e-mail]>
> *Subject:* Re: draft-bierman-core-yid-00 open questions
>
>
>
>
>
>
>
> On Wed, Aug 24, 2016 at 9:53 AM, Michel Veillette <
> [e-mail]> wrote:
>
> Hi Andy
>
> Following if my answers to the questions listed in
> draft-bierman-core-yid-00
>
> 1) When to assign a module-id
>
> Developers should be in control of when this activity is performed based
> on the estimated amount of work required to reassign IDs (from temporary
> assigned IDs within the experimental range to the final IDs) and the
> impacts on the encoding efficiency of assigning a non stable list of YANG
> items which generate non sequential IDs. If the developer already own a
> range of YID/SID, the activity can be perform at any time.
>
> It is important to note that the registration of the resulting .sid
> file(s) is done independently of the registration of the SID range. When
> registering a SID range, the module or list of modules targeted is not
> provided. This give the full freedom to developers to assign IDs when
> needed.
>
>
>
> I think the registration process issues are the same whether a module-id
> or a range is allocated.
>
>
>
> [MV] Not exactly, registering a module-id imply that you need to know the
> module information in advance.
>
> [MV] If you register a range, you can assign this range as needed to one
> or multiple modules.
>
> [MV] You don’t need to go back to the registrar each time you create a new
> module.
>
>
>

But it's not like anybody can use the numbers without knowing how they are
assigned.


Andy

2) How to support private numbering
>
> A range called "Experimental" spanning both the 16 and 32 bits unsigned
> integer have been assigned to this purpose.
> https://tools.ietf.org/html/draft-somaraju-core-sid-01#section-7.1
>
>
>
>
>
> Yes, it seems trivial to assign a range for temporary numbers.
>
>
>
>
>
>
>
> 3) How to combine registries
>
> Not needed, all registrars, SDO shall receive a specific range from IANA.
>
>
>
>
>
> OK, so there is a tiered registration process?
>
> And there will not be a land-grab for all the numbers?
>
> Not clear to me how IANA approves allocation requests.
>
>
>
>
>
>
> 4) Should more YANG statements be numbered
>
> Yes
> rpc, action, notification, module, submodule, feature
>
>
>
> Numbering submodule and feature?
>
> These are conceptual statements within the schema language and
>
> do not affect anything on the wire.
>
>
>
> [MV] SID assigned to submodules and features are required to support
> discovery
>
> [MV] See https://tools.ietf.org/html/draft-veillette-core-cool-librar
> y-00#section-3.1
>
>
>
> 5a) How to support anyxml
>
> anyxml represents a single data node with an arbitrary content.
> This data node have a single ID assigned to it the same way as any other
> data node, nothing special is required.
>
>
>
>
>
>     anyxml foo;
>
>
>
>
>
>     { "foo" : {
>
>          "bar": 42;
>
>          "baz" "fred",
>
>          "a-list" : [
>
>             ....
>
>          ]
>
>        }
>
>      }
>
>
>
>
>
> So what numbers do you give to bar,, baz, a-list, etc.?
>
> They are all numbered the same as foo?
>
> That will be decoded with all nodes named foo, all anyxml.
>
> The name, type, value info contained in anyxml and anydata
>
> instances have no schema.  These nodes cannot be numbered.
>
> This is a YANG-to-CBOR issue, not a SID issue.
>
>
>
> But in real YANG these child nodes sometimes represent real objects
>
> from some other module.  In this case the server may know the "hidden SID".
>
>
>
> http://www.netconfcentral.org/modules/yuma-ncx/2013-09-23#root.554
>
>
>
> [MV] If the inner members of this CBOR map need to associated to SID, the
> developer will need to assign these SIDs as described in 5b.
>
> [MV] If this is the case, the author should have used an anydata.
>
> [MV] Normally, an anyxml contain an arbitrary CBOR content which have
> nothing to do with YANG and SIDs.
>
> [MV] This is the approach proposed by "draft-ietf-netmod-yang-json" and
> reused by "draft-ietf-core-yang-cbor"
>
>
>
>
>
> 5b) How to support anydata
>
> The anydata define an instance of an unspecified schema node (leaf,
> leaf-list, container, list).
> All possible schema node supported by an anydata need to be register to
> one of the .sid files.
> It is the responsibility of the developer(s) to add these SIDs to one or
> multiple .sid files manually if not already present.
> A section need to be added to the draft to clearly describe this approach.
>
> Regards,
> Michel Veillette
>
>
>
>
>
> Andy
>
>
>