Current Mode OPNFV/CNTT Gaps/Issues/Scope Definition
The CNTT and OPNFV communities agree to follow the premises below and subscribe to the associated RACI chart for scope boundaries:
CNTT defines specifications and provides only the specifications to OPNFV
CNTT: Reference Model, Reference Architecture, Reference Implementation
OPNFV: Reference Implementation, Reference Compliance
OVP: Testing, Badging
OPNFV performs builds from the Reference Implementation specifications.
"Cookbook" artifacts are the domain of OPNFV, CNTT should participate in the management of the cookbooks, and selects the Test cases to be executed.
A release management method including feedback loops is needed between OPNFV and CNTT. LFN/OPNFV has an existing meeting in which some members of the CNTT community will participate.
CNTT Participants that will participate include @Scot Steele, etc
Cross Community Collaboration must increase without overloading the participants.
Recommendation: WS Leads should identify WS participants that can attend the other community TSC meetings
Regularly held cross community meetings should be held. Community Chairs should schedule/delegate to community participants to implement/manage
General group agreement
The RACI doc and other docs related to the organization are all living docs.
Long term there will need to make a decision related to the final home for CNTT project work. – @vincent.danno@orange.com
We need to see how successful "1" arrangement is. → no need to rush in thinking about "2" and we still didn't figure out "1". I STRONGLY SUGGEST NOT TO RUSH INTO THIS WITHOUT FIGURE OUT "1" AND PROVE MODEL WORKS. @Rabi Abdel
Questions need to be answered: Current Mode OPNFV/CNTT RACI Chart
@Cédric Ollivier How can it help the short term issues? an implementation is needed for the conformance on the VNF side and we need additional relevant resources in both sides: specifications (CNTT) and OPNFV (OPNFV)
@Cédric Ollivier How would be leverage on it in a long term view? We shouldn't reinvent the wheel every release - the current model works and the separation between CNTT and OPNFV is clear for a few main CNTT contributors
What happens when CNTT develops a requirement in, say, RI-1?
does CNTT goes directly to prospective OPNFV project and get it implemented (via that project PTL).
how will the implementation link back to CNTT releases?
@Cédric Ollivier : CNTT RC lists the deliverables (containers/tags/testcases) from OPNFV. What's the additional point here? Could we detail the gap targeted?
Will each project have their own release cycle (with appropriate baggings) or will OPNFV have an overall release cadence ? how does the model work?
@Cédric Ollivier I think the model works already.
CNTT is in charge of test case description/selection and playbooks (see RC) and then CNTT selects the test cases from OPNFV according to them.
OVP should implement the test result verification from the archive produced by CNTT playbook - Compared to first OVP model the test case selection is now in RC
or will CNTT has to follow some kind of process to get their requirements implemented?
@Cédric Ollivier : we may consider the classical opensource model here and then the answer could be straightforward: by putting developpers
@Cédric Ollivier : opensource cannot follow a subcontractor model by design
or will OPNFV itself regularly get CNTT releases and find the right OPNFV projects to implement any new requirements.
via CIRV? @Cédric Ollivier: CIRV is mostly an umbrella for incubation projects. It should be done at the project level possibly between CNTT RI/RC leaders and the PTLs/core developpers
what is the role of OPNFV TSC on all this? who is accountable for what?
we need to talk about accountability and make it clear who is accountable for what!
How does OVP fit into this matrix?
Given the GSMA process is it mostly a consumer and ratifier of the output from CNTT, or does it have a role in providing feedback? I.e. how can this be set up to be a two way street?
@Scott Steinbrueck proposal 6/9/2020
The challenge is how to organize the content in an efficient way that reduces complexity for the authors/implementors to create, while also producing a quality product that is simple for the end user base to navigate, read and comprehend.
Seeking agreement to the following premise:
RA-1 is the requirements and specs
RI-1 is an example implementation (“example” word is used because no company is required to follow this exact lab setup / installer approach – it’s an example)
@Cédric Ollivier
Does it mean that RI-1 would be strictly removed from CNTT repository?
Yes why not considering Airship as an implementation amongst others which tries to implement the CNTT requirements (RC is currently rather verified thanks to Functest SUT and Field trials).
I would agree to fully remove RI1 from CNTT but CNTT needs a reference implementation somewhere which passes RC successfully for the next steps (conformance of VNF/CNF).
By the way, it's becoming urgent to take CNTT results into account to find the best path (RI1 doesn't pass RC - we can't leverage on the existing resources allocated to Functest for the VNF conformance)
If FMO targets to improve the process, having an implementation fulfilling CNTT requirements is a very short term key issue
RC-1 is a test conformance suite
@Scott Steinbrueck proposal 6/16/2020
Big Picture
@Cédric Ollivier
Sorry I'm lost here. It's introduced as an executive summary but I would rather consider that it increases so much the complexity both regarding the collaborative work and the end users.
If we target an executive summary it could rather be a minor change in the top CNTT document listing (it could be a new Executive summary bullet in between Governance and Technical specifications)
RA1 requirements
additional relevant requirement details (e.g. RA1 chapter 5)
the related mandatory test cases (please see 3.4 Test Cases Traceability to Requirements)
RC playbook (please see 4.3 NFVI Testing Cookbook)
RA1 would decide for everything which is not right regarding the current work and false from a collaborative mindset.
I'm lost by considering RI1 as an example. Why should we increase the complexity of RA1 in that case?
If the overall work is about the separation between CNTT and OPNVF, everyone should know that RC1 Baldy and older is only driven by the CNTT part:
testcase descriptions, selection and integration are part of CNTT RC
OPNFV implements the test cases and offers the deliverables (e.g. containers) as asked by CNTT (see first bullet)
In my view, Ch 10 is just complexifying the current model which should firstly be demonstrated as false to avoid any sophism as it seems here.