[netmod] Re: Comment on draft-fedyk-netmod-yang-normal-form

[Balázs Lengyel <[e-mail]>]

Hello,

  1.  I would consider using the new form for retrieval also. While some might consider that incompatible, I dislike introducing a third format lexical, canonical and now normative.
  2.  greater-than lower-than should also use the normalized form.
  3.  Evaluation of the unique statement must also use the new form.
  4.  Evaluation of subtree filtering must also use the new form.
  5.  What format would NETCONF insert before/after use?
  6.  Would a leafref or instance-identifier reference be based on the new form ?
  7.  A general YANG guideline should be created to make patterns strict. E.g. detecting A-Xa-z patterns inside pattern statements.
  8.  This requires a separate identity and description for each data type. I would consider as an alternative a "canonical-pattern" extension. Most of the time that would be enough, but it could also have a description substatement if needed.
Regards Balazs

From: Joe Clarke (jclarke) <jclarke=[e-mail]>
Sent: Thursday, 23 July, 2026 14:44
To: NETMOD WG <[e-mail]>
Subject: [netmod] Comment on draft-fedyk-netmod-yang-normal-form

I said at the mic today this approach reminds me of MIB object DISPLAY-HINTS (but in the reverse).  Those were very helpful to take which would have been generic octets and expressing them in something more logical.  Likewise, I think this is useful work, and I think this should be considered for YANG Next in addition to it being an extension for now.  I think the approach of using identities is good.

That said, the YANG module in this draft is very under-specified.  The mac-48 identifier should include the text in section 5.4.1 in its description (IMHO).  I think any additional identities should be required to express a similar normalization scheme as part of their descriptions (e.g., ipv6).

One other thing.  The examples in 7, 8, and 10 should probably use iynf:normalized-form "iynf:mac-48";  The use of "yang:" was confusing at first.

Joe