Closed vlopezalvarez closed 2 months ago
Hi, Victor, we thought that since model-name was designed to use as part-number, it was more straightforward to use part-number. Part-number is more common in Operators' network management.
Part-number is widely used in the community. Besides it matches with openconfig part-no, so industry-wide there is no confussion.
2024-01-17 Network Inventory weekly call
The rationale is described in section 2.4.3 of the draft
It has been noted that the YANG description needs to be updated
[ ] @YuChaode update the YANG description for the part-number
[ ] @sergiobelotti check offline with Victor if he is ok with the proposed resolution
Model changes proposed by Chaode has been reviewed during the call
The definition of part-number of the component has been updated based on the discussion
The part-number attribute is not applicable to NE. It has been agreed to rename the attribute as assembly-id as a working name and open a new issue to refine the name and description of the assembly-id attribute: see issue #33
- [ ] @YuChaode update the YANG description for the part-number
Similar with Victor, I suggest that "model-name" in RFC 8348 be reused. This node applies only to components, not NEs.
I tend to agree with Bo's comment: one thing is discussing the opportunity or not to use part-number on behalf of model-name for component only another thing is mixing things and introduce another term , not defined in RFC8348, considering also NE while model-name does not .
"The vendor-specific model name identifier string associated with this physical component. The preferred value is the customer-visible part number, which may be printed on the component itself."
@sergiobelotti @lana-wu I am a bit confused about whether your comments are referring to:
I would suggest using this issue to discuss point 1 and issue #33 to discuss point 2
Regarding point 1, I have understood from the previous calls that we have agreed to use "part-number" or "model-name" attribute for the component, pending confirmation with Victor and marked this issue as agreed accordingly
Please let me know if my understanding is not correct: if needed, we can discuss it again in the next calls (and mark this issue as under-discussion to avoid confusion)
@italobusi @sergiobelotti I agree with the first point, instead of a new name "assembly-id“. For the second point, I am not sure about "asset-id", which is NOT a vendor-specific but a customer defined tag.
"The vendor-specific model name identifier string associated with this physical component. The preferred value is the customer-visible part number, which may be printed on the component itself."
@sergiobelotti @lana-wu I am a bit confused about whether your comments are referring to:
- use of "part-number" or "model-name" attribute for the component
- use of "asset-id" or "model-name" (or "product-name" for the NE
I would suggest using this issue to discuss point 1 and issue #33 to discuss point 2
@italobusi: my comment is related to the above comment from Bo https://github.com/ietf-ivy-wg/network-inventory-yang/issues/15#issuecomment-1918268011, and is referring to the first comment from Victor regarding modul-name vs. part-number. I agree on your suggestion to separate the two issues .
Regarding point 1, I have understood from the previous calls that we have agreed to use "part-number" or "model-name" attribute for the component, pending confirmation with Victor and marked this issue as agreed accordingly
Please let me know if my understanding is not correct: if needed, we can discuss it again in the next calls (and mark this issue as under-discussion to avoid confusion) @italobusi : yes you understanding is correct
Part-number is widely used in the community. Besides it matches with openconfig part-no, so industry-wide there is no confussion.
Actually I noted that OpenConfig defines both part number and model name attributes:
https://github.com/openconfig/public/blob/master/release/models/platform/openconfig-platform.yang
Since the model name is already defined in RFC8348, an alternative solution could be to define:
part-number as “The system-assigned and vendor-specific part number of the component”
product-name as “The system-assigned, vendor-specific and human-readable string describing the component”
Following the similar approach, we can also define:
serial-number as “The system-assigned and vendor-specific serial number of the component”
asset-id as “The user-assigned asset tracking identifier of the component”
2024-02-21 Base Network Inventory weekly call
The following attributes have been agreed:
part-number as "The vendor-specific part number of the component type. It is expected that vendors assign unique part numbers to different component types within the scope of the vendor."
product-name as "The vendor-specific and human-interpretable string describing the component type. It is expected that vendors assign unique product names to different component types within the scope of the vendor."
serial-number as "The vendor-specific serial number of the component instance. It is expected that vendors assign unique serial numbers to different component instances within the scope of the part number."
[ ] @YuChaode : update the draft and the model accordingly
2024-03-06 Base Network Inventory weekly call
The YANG descriptions in the I-D version uploaded for IETF 119 are not fully aligned with the descriptions agreed during the call on February 21, 2024
Closed with 683ea0729a10a841b3fa7de3ec57eb8f33920e0c
Hi, I've seen that we have in the draft the concept of "part number". RFC 8348 defines the parameter model-name to obtain this information.
"The vendor-specific model name identifier string associated with this physical component. The preferred value is the customer-visible part number, which may be printed on the component itself."
As we're changing the model of 8348. Is there any reason to include the "part number" instead of reusing model-name?