From: Mahesh Jethanandani <[email protected]>
Sent: 30 September 2025 02:00

[Changing the subject line as the discussion is tangential to the work at hand]

Hi Andy/Tom,

Here is how I am trying to rationalize this:

  *   Should (complete, and not code snippets) example modules (or instance 
data examples) be validated? Per rfc8407bis, Section 3.2.1, it is pretty clear 
that they should be validated.
  *   Are those examples normative? The answer is no, and that is why the 
examples should exist in the non-normative section of the draft.
  *   Separate from that, the question is how the examples can be validated? 
The tools can validate them only if they are standalone files. To create those 
standalone files, they have to be extracted from the draft. Would you agree 
that how we extract them is less important than being able to do so?

We can discuss whether we need a different tag to extract (complete) examples, 
but given the current toolset, I do not believe we have a conflict in how the 
examples are extracted and named.

<tp>

An excellent suggestion for 8407bis to expand on.  Ah, it is too far  down the 
track for that, except that looking at 8407bis, I think that it is 
inconsistent.  It says

"  Example modules are not code components.  The <CODE BEGINS>
   convention MUST NOT be used for example modules.  However, example
   modules MUST be validated (Section 3.10)."

but how do you validate them when you do not have CODE BEGINS to automate the 
extraction?  We need SAMPLE BEGINS and an update to the tool chain for 
validation!  Without that, I think we are in a confusing state.  The addition 
of that third sentence to the paragraph I quote has consequences that seem not 
to have been thought through.  I would strike that third sentence, even at this 
late stage, and produce a separate I-D on the issue.

The other unresolved issue is migration.  For years to come, we are going to 
have modules that do not conform with 8407bis but there is no way of knowing 
which they are.  We need something in a module, going forward, to say that it 
is intended to conform to 8407bis and is an error if it does not, just as when 
we updated YANG, we changed the version to 1.1
.
Tom Petch


Cheers.

On Sep 24, 2025, at 9:36 AM, Andy Bierman <[email protected]> wrote:

On Wed, Sep 24, 2025 at 4:26 AM tom petch 
<[email protected]<mailto:[email protected]>> wrote:
From: Mahesh Jethanandani 
<[email protected]<mailto:[email protected]>>
Sent: 23 September 2025 21:47

Hi Tom,

My take is that code markers only say where the code begins and where it ends. 
It does not distinguish between it being YANG,  MIB, an example or any other 
piece of code. The code markers can specify which file the said code should 
extract into. We do not need separate code markers for different pieces of code.

<tp>

Colour me confused.  If we do not need separate code markers for different 
'types' of code (which I agree with), then that tells me that the
CODE START and CODE END should not appear in this I-D but should be replaced by 
CODE  BEGINS and CODE ENDS.



Correct.
https://www.ietf.org/archive/id/draft-ietf-netmod-rfc8407bis-28.html#name-code-components

Note that the MUST NOT rule in 3.2.1 is not being followed, and example modules 
are commonly
documented as code components.

There was discussion about EXAMPLE BEGINS/ENDS, but a new macro like that was 
not added.
There was discussion about the importance of IETF Trust IPR issues associated 
with code components.
That is why the guidelines draft distinguishes between normative and 
non-normative YANG modules.

Since examples are quite common in all I-Ds, this seems like an important 
precedent being set here.
As a consumer of RFCs, I do not really want dozens of extracted files for 
ad-hoc example snippets.
These snippets are not code components, and it would be better to handle 
example validation another way.



Mahesh Jethanandani
[email protected]






_______________________________________________
netmod mailing list -- [email protected]
To unsubscribe send an email to [email protected]

Reply via email to