Meld Ops and Org 2020-11-05 Meeting notes
Date
Nov 5, 2020
Attendees
@Jim Baker @Heather Kirksey @Scot Steele @Beth Cohen @Al Morton @David McBride @Qiao Fu @Toshi Wakayama @Ulrich Kleber @Georg Kunz @Trevor Cooper @William DIEGO (Orange) @Jonne Soininen @Lincoln Lavoie @Ildiko @Sunku Ranganath (Deactivated) @Phil Robb @Tom Van Pelt (Deactivated)
Discussion items
Time | Item | Who | |
---|---|---|---|
10min | Interim TSC Election plans | @Scot Steele | |
20min | Release artifact discuss and agree | @Mark Beierl | Deferred for @Mark Beierl presence |
10min | Business coordination function
| @Scot Steele | Follow up needed after defintion Nov 12, 2020 |
20min | Release process and naming
| @Jim Baker @David McBride | |
10min | Role of Installers | @Trevor Cooper | |
10min | Conformance projects - what they deliver | @Trevor Cooper | Defer for Release Artifacts disussion |
2min | Proposed Wiki Structure | @Jim Baker |
Notes
Interim TSC plans
OPNFV providing 8 members from the seated TSC as selected by the active community members
CNTT will wait for completion of the OPNFV proposed interim TSC membership
Look for equitable distribution of seats between operators and vendors for Anuket TSC
This is for a INTERIM TSC ONLY - not a permanent change
Business Coordination Function
Expect a short term function - up to two years. Can continue as needed.
Volunteer staffed - likely to be the same folks that are staffing the Anuket MeldMarketing
Focused on the initial project spin-up. Eventually absorbed by MAC function
@Trevor Cooper Please write down the scope and responsibility of this function
CNTT equates Governance to Business Coordination
Notes from @Scot Steele
Strategy, Participant interface; On-boarding, escalation point for complex, inter-related issues; Virtual Conference mgmt, other Admin tasks, Coordination with other projects, Recruitment
Release Process and Naming
Key problem statement: How do we use naming to unify the specification versions with subsequent releases of the RC?
@Georg Kunz Joint release planning between the spec developers and the RC developers could address this.
@Trevor Cooper It's dependency management.
RC project does not equal the OVP program
RC develops test cases and tools to test RA requirements - unlikely to ever deliver 100% of the requirements - MUST be decoupling and independent projects
@Lincoln Lavoie RC is a work stream of individual projects - not a single entity
Perhaps RC should handle the traceability and coverage mapping - by definition it will lag the specification version
What is the stated goal of the lag?
The specifications are a monotonically increasing list (moving target)
RC may not always track the increase in feature sets
@David McBride propose decoupling of specification and RC releases
Likewise use different naming to avoid assumptions of RC fulling implementing a specific RA version
@Lincoln Lavoie RA/RM release Q1 and Q3, RI/RC release on Q2 and Q4
4 releases per year: 2 spec focused and 2 test focused
RC must provide 100% coverage of RI - for whatever features that are implemented in the RI
@Georg Kunz There will always be a gap/offset
@Heather Kirksey Could create a longer release process that just makes the spec a milestone in a longer release cycle.
@Al Morton Community is asking for more synchronized processes
3 month is not sufficient for developing response to the specifications - far shorter than prior OPNFV releases
Possibly overlap the last 3 months of spec dev with first 3 months testing development
@David McBride Concern is development of specifications will outpace the development of tests
@Jim Baker FOCUS is first 3 months of 2021
Role of Installers
@Trevor Cooper Role of installers has been an issue in the history of OPNFV
Proposed statement: Installers are NOT a part of a release. They are a tool to create an RI
OPNFV Jerma has one installer: Airship - it is unclear that this installer meets all the needs of the community
@Georg Kunz Installers may be released as a part of "RI"
How many installers and sets of components (RIs) does Anuket want to support?
Would prefer ONE RI to pool resources
@Trevor Cooper Airship project is a stand-alone tool - it doesn't deliver a RI
RI is delivered as a cookbook or method for deploying the RI
@Georg Kunz Would like to use an RI in the internal testing process for commercial xNFs
@Trevor Cooper Propose deferring the decision to the Release Artifacts discussion
@Al Morton There are many ways to end up with a RI as defined by the RA and validated as correct by RC