Document: draft-ietf-ivy-network-inventory-yang
Title: A Base YANG Data Model for Network Inventory
Reviewer: Elwyn Davies
Review result: Ready with Issues

I am the assigned Gen-ART reviewer for this draft. The General Area
Review Team (Gen-ART) reviews all IETF documents being processed
by the IESG for the IETF Chair.  Please treat these comments just
like any other last call comments.

For more information, please see the FAQ at

<https://wiki.ietf.org/en/group/gen/GenArtFAQ>.

Document: draft-ietf-ivy-network-inventory-yang-18
Reviewer: Elwyn Davies
Review Date: 2026-06-10
IETF LC End Date: 2026-06-10
IESG Telechat date: Not scheduled for a telechat

Summary:  There is one potential issue that the document itself brings out and
a couple of operational ssues that could make the specification difficult to
use especially on larger systems. Otherwise, the document has a fair number of
nits and editorial issues outside the YANG specification.  I have not looked at
the form of the YANG in great detail.  Please note that some parts of the
appendices are (partly) written in academic paper style and need to be
depersonalized.

Judging from the example in Appendix E, the comments in the Security section
about the risks of very high volumes of YANG data needed to complete the
inventory of a network seem rather too well justified. I don't know if this
could be a show stopper but feel that consideration should be given to this
matter by peole with more hands-on knowledge of inventory handling.  I wonder
if the inventory could be segmented in some way to limit the size of chunks
being handled.  I am unsure whether there is going to be much call for complete
inventories to be interchanged.  It  strikes me that one aspect that has not
been discussed here is the dynamic cnstruction of complete inventories.  I
could see an example where a new component is added to a network:  Would the
expansion of the inventory be conveniently carried out by the new component
carrying its own inventory and passing it over to the inventory server (with
appropriate security checks) for integration into the overall inventory - as
opposed to the central inventory being updated by manual/semi-automatic
interaction at the inventory server.  There are also potential risks of the
consistency of the model being damaged by incomplete or faulty removal of the
inventory associated with a componet tht has been removed.

Major issues:
1. Potential problem:  Size of complete YANG inventory specifications if needed
to be interchanged between devices. 2. Possble problem:  Ease or difficulty of
integrating/de-integrating YANG inventory specifications for added or
subtracted components. 3. Handling inter-component connections and
disconnections. Ensuring that changes to interconnections can be handled in a
way that limits the scope for dangling connections to be left in place after
changes to components.  This is not discussed in the document.

Minor issues:
None

Nits/editorial comments:

S1, paras1 and 2:  The definition of Network Inventory really needs to appear
right at the beginning of the text.  As this document is the primary output of
the IVY WG, I would suggest copying all or at least the beginning of the first
paragraph of the charter of the IVY WG to supply the definition of Network
Inventory:- thus some of all of:

    Network inventory is a foundation for network management in all types
    of: Network operators need to keep a record of what equipment
    is planned and installed in their networks (including a variety of
    information such as product name, vendor, product series, embedded
    software, and hardware/software versions). Network inventories may
    be constructed from management system data to represent the expected
    devices present in the network, and also be used to audit and catalog
   what devices are discovered in the network, and to expose that
   information in a consistent way.

and delete para 2

s1, para 3:  s/overall network management/scheme of network management/

s1, para 3: "which was specified many years ago"  If so, can we have a
reference please.  I could not identify where this specification is located
although maybe RFC 8969 may be relevant.  If there isn't one, just omit this
phrase - I am quite prepared to accept that it is a useful fundamental block.

s1, para 3: s/well-planned/well-managed/ (One hopes that the planning precedes
definition of the inventory;  the inventory essentially documents the plan).

s1, para 4: s/to retrieve network element/to retrieve information relating to
network element/

s1, para 4: s/are key enablers/provides key enablers/

s1, para 4: s/a gap about/a gap relating to/

s1, para 5:  Please delete this paragraph.  This is a standards document and
not a poltical positioning paper.

s1, para 8:  This paragraph does not read well.  Suggest:
OLD:
As outlined in Section 6, the network inventory provides a read-only
perspective of the actual inventory data that a network controller knows of
what it is actually installed within the network. NEW: As outlined in Section
6, the network inventory provides a read-only perspective of the actual
inventory data of which a network controller is aware covering the components
that are actually installed and active within the network. END The term 'spare'
in the second sentence of this paragraph may be misleading. Networks may
contain backup components which are there as spares that can be switched in
automatically in the event of the primary component failing. I would expect the
inventory to account for these. This is clearly different from having offline
spares that might require human intervention to insert them into the active
inventory.

s2, definition of Hardware Component:  s/The generalisation of/A portmanteau
[or alternatively 'An umbrella'] term for any of/  It might be worth
referencing RFC 8348 here which sets out terms for adding items to the lists.

s2, definition of Component:
OLD:
Component: The generalization of the hardware component definition to include
other inventory objects which can be managed, from an inventory perspective,
like hardware components. NEW: Component: A furher extension of the portmanteau
hardware component definition to include other inventory objects which can be
managed, from an inventory perspective, in the same way as hardware components.
END

s3.1 and s3.1.1: The use of 'such as' in the first paragraphs of these sections
implies that there are other common attributes. etc.  I suspect that the lists
givn are complete and different phraseology should be used:

In s3,1;
OLD:
For all the inventory objects, there are some common attributes, such as:
NEW:
All inventory objects have the following common attributes:
END

In s3.1.1:
OLD:
To be consistent with the component definition, some of the attributes defined
in [RFC8348] for components are reused for network elements, such as: NEW: To
be consistent with the component definition, the following attributes defined
in [RFC8348] for components are reused for network elements: END

s3.1, definition of ne_id: 'assigned by the server' -  Which server would that
be?

s3.3, para 1: s/same approach of/same approach as/

s3.3, definitions of hardware_rev and mfg-date:  Should the specification offer
a standard form for use in case these items are not known?

s3.3.1, para 1:s./classifies/classify/, s/and manage all/and manages all/

s3.3.2, para 3: s/an Application software modules/an Application module [or
Application software modules]/ Which is intended?

s3.3.2, para 3: Expand FPGA - it isn't defined currently.

s3.3.2, para5: s/under definition/defined/ (By the time this draft is
published, the model will have been defined)

s3.3.2, para 6:  'have similar replaceable requrements':  Does this mean that
the components have similar requirements or characteristics as to
replaceability?  I don't understand what is meant here.

s3.3.2, para 6: s/one or more software patch information/one or more records of
software patches that have been applied./

s3,3.2, last para: s/like/such as/

s5, module description item:  Suggest adding reference to RFC 8342 for
Datastore Architecture

s6, para 7: s/ and outside the scope/and is outside the scope/

s6, para 8: s/dooes not constraint/does not constrain/

s6, para 10:  The term 'brownfield' is jargon that needs to be explained or
replaced.

 I suggest somethings along the lines of:

      If the inventory system defined by this document is to be deployed
      into a nework which has a  preexistng inventory system....

s9,2:  Arguably RFC7951 and RFC 8340 are normative references.  If RFC 6241
(NETCONF) is nomative then I would suggest RFC  8040 (RESTCONF) is normative as
well... but they could also both  be considered informative.

Appendix A:  The firstt two and last paragraphs and table of this section
section is written in academic paper style  with 'we' do things and 'our'
rather than impersonal standards form  and need to be rewritten.  There are
also two missing deifinte articles in the first sentence of the second
paragraph. before Openconfig and "ietf-hardware".and another one missing in
each of the third and fourth paragraphs.

Appendices B and C; Also have the academic paper style issue.

Appendix C, para 3: s/large scale of networks/large scale networks/

Appendix C, para 4: s/ so to cover/so as to cover/, s/in YANG model/in the YANG
model/, s/it is only needed/it is only necessary/

Appendix D, para 2: s/three type of ports/three types of port/

Appendix E:, para 2: ss/composed by/comprised of/

Appendix E para 3 and para 8: :s/Stacked switch/Stacked switches/

Appendix E, para 6: s/synchronzed/require synchronized/

Appendix E, titles of Figures 5 and 6: s/a stacked switch/stacked switches/

Appendix E, para 9:
OLD:
Using the base network inventory YANG data model each interconnected switch can
be modelled as a chassis within the same network element, which models the
cascaded switch. The ports used to interconnect the different chassis are
normal (traffic) ports and modelled like other ports. NEW: Using the base
network inventory, the YANG data model for each interconnected switch can be
set up as a chassis within the same network element, which models the cascaded
switch. The ports used to interconnect the different chassis are normal
(traffic) ports and modelled in the same way as other ports. END

Appendix E: [Aside: Yes, that's quite a lot of YANG!! ]



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

Reply via email to