Skip to content

Meeting 2019 06 21

Jump to bottom
Josh Hursey edited this page Jun 21, 2019 · 1 revision

Agenda

Agenda:

Notes

        Joshua Hursey (IBM)
	Kathryn Mohror (LLNL)
	Stephen Herbein (LLNL)
	Nick Radcliffe (Cray)
	Ralph Castain (Intel)
	Anatoliy Rozanov (Intel)
	Jithin Jose (Microsoft)
	Howard Pritchard (LANL)
	Geoffroy Vallee (Sylabs, Inc)
	David Solt (IBM)
	Artem Polyakov (Mellanox)
	Swaroop Pophale (ORNL)
  • Naming discussion: Reference Implementation vs Standard document
    • Reference implementation development community open to changing name to “Open PMIx”, but not immediately. Desire to see how far the standardization can go with PMIx as the base.
    • Desire to know what folks have in mind as far as major/minor changes to what’s in the PMIx standard today.
      • Are these radical changes or evolutionary changes?
      • If, down the line, someone joins that proposes something too radical then the PMIx community can accept/reject the proposal.
      • Adding things generally is not a problem, deprecation/removal is ok. Rewriting APIs is problematic.
    • Suggest to postpone naming until we gain some clarity on the changes to expect.
  • PMIx Standardization Process:
    • Things left to discuss:
      • Stability classes (https://github.com/pmix/pmix-standard/issues/179)
        • Suggestion - let’s draw up a specific proposal
          • L1 - everything starts here
          • L2 - requires two release cycles to be eligible
        • Link to the implementation
        • Links to users or use cases for the API
        • Chicken & Egg issue
          • User needs the interface standardized to use it in their application, but standard needs a use case to accept it.
          • Allowing interface to be brought into the standard as L1 lowers the barrier to entry.
        • Who gets to decide as it moves between the levels? What is the process to include the necessary folks?
          • Mailing list and weekly meetings can be a lot of overhead for some folks not down in the weeds. Having a less frequent fixed point in time gives the general audience a lower overhead way to participate.
          • Suggestion of a quarterly (virtual?) meeting to decide on PRs to the standard.
          • Everything goes in as L1
        • Suggestion on voting using flexibility of stability classes
          • L1 - lighter process
            • PR based process with straw voting and no unresolved objections.
            • Accepted quarterly
          • Move to higher level requires more formal voting structure and/or more time.
            • The organization should be present at most of quarterly meeting to vote.
            • One of the quarterly meeting is face-to-face?
          • Who gets to vote? Per organization or per person?
          • Ralph to draw up a formal proposal along the lines above to Issue 181
        • Stephen to work with Ralph on the stability portion of the process PR.
      • Mission statement and scoping (https://github.com/pmix/pmix-standard/issues/190)
      • Top two intertwined:
        • When considering a PR what should you be thinking about?
          • Ok to bring in provisionally/L1
          • Depreciation requirements
          • Implementation requirements
  • (Dave) Working group update: Implementation agnostic document
    • Still working through Chapter 1.
    • Emailed a proposal that moves some text around, and rephreases some of the architecture portion a bit more generally.
    • Preserves rationale about interfaces/attributes being optional.
    • Next step is to divide up the sections and work through them.
    • Target: Draft of chapter 1 for review in a month.
      • Renaming chapters should be a bit easier, and brought forward for review on a chapter-by-chapter basis.
  • (Stephen) Working group update: Slicing/Grouping of functionality
    • About ½ way through RFC
      • Documenting interfaces/attributes in each RFC and linking them to higher level use cases.
      • Once complete, then define use cases with required and optional dimensions. Optional will have discussion about implications of not providing it.
    • Canceling the next two meetings. Next meeting July 10.
    • Targeting Mid-Aug to have a proposal for review.

Home

Working Groups

Events

Helpful How To...

ASC Quarterly Meetings

ASC Monthly Meetings

Clone this wiki locally