-
Notifications
You must be signed in to change notification settings - Fork 22
Meeting 2019 05 03
Josh Hursey edited this page May 9, 2019
·
1 revision
Kathryn Mohror (LLNL)
Stephen Herbein (LLNL)
Barry Rountree (LLNL)
Aurelien Bouteiller (UTK)
Swaroop Pophale (ORNL)
Thomas Naughton (ORNL)
Dave Solt (IBM)
-
Continue discussion on PMIx standardization process
- Discuss PR associated with: https://github.com/pmix/pmix-standard/issues/181
- (Josh) I have a PR version of Issue 181 almost finished. I will have the PR posted by Monday (if not sooner).
- Discuss: https://github.com/pmix/pmix-standard/issues/179
- Try to limit this to first 1/2 of the meeting
- Discuss PR associated with: https://github.com/pmix/pmix-standard/issues/181
-
(Dave) Working group: Implementation agnostic document
-
(Stephen) Working group: Slicing/Grouping of functionality
-
(Josh) Use case gathering
-
Need a concept of “reasonable” time period for review so that someone can’t postpone voting/adoption indefinitely
-
4/26 notes on issue 181: lack of dissent model
- Last time we said that Github voting by “thumbs up” on PR within some period
- This may limit what can move to ‘experimental’ status. People who vote thumbs down must include a comment stating their technical reasons for voting against
- This model informs authors of what they need to address for more formal adoption
-
Issue #179
- To discuss:
- Names of statuses
- How many users to move between Experimental to widely used?
- What about use cases?
- Do both multiple users and use cases prove generality?
- Do multiple PMIx implementations supporting the feature show generality?
- Moving to ‘widely used’ gives some stability guarantees and makes it harder to change the interface so we don’t want the bar to be too low
- (users, use cases, multiple implementations) could create a metric based on counts of items in the tuple, say 1 user but 3 use cases or 2 implementations and 1 user
- Experimental, widely used, stable
- Experimental: Does it have a negative connotation?
- Unstable?
- New?
- Proposed?
- Level 1, 2, 3
- Evaluation, accepted, stable
- Plasma, gas, liquid, solid
- Experimental: Does it have a negative connotation?
- What are the requirements on implementations wrt to the levels? Do they have to implement all the experimental interfaces?
- We still want to leave most of the interface optional although there may be pieces of functionality in the interface that will be required for a baseline PMIx implementation
- “Rating” of PMIx implementations will be determined later, e.g. PMIx implementation supports all stable bootstrapping interfaces, etc
- To discuss:
-
Working group: Implementation agnostic document
- Fill out the doodle poll!!
- What is the best approach to managing the work in this WG? Will we assign chapters to people?
- Then have everyone come back with their notes, compare commonalities across chapters, get a feel for what people think
- If we can have multiple people look at the same chapter to tease out differences of opinion
- Start off with everyone looking at the same section of the document to try to find the patterns/sections/ideas that we think are implementation dependent and then we can break up the document into sections, once people know what patterns they are looking for
- Dave will close the poll if he can get enough responses late Monday and plan to start next week hopefully. Perhaps we should perform more outreach first
- People are encouraged to come prepared with opinions!
-
Working group: Slicing
- https://github.com/pmix/pmix-standard/issues/182
- The plan is to hold off until after there have been several meetings of the implementation agnostic group meets
- Whole groups of functionality could be eliminated if they are decided to be implementation specific, rearranging of the document could cause headaches
- Might think about big picture of functionality before moving forward with text changes
- Outreach needed for groups that are interested in use cases who may have input on this topic
- Stephen will send out a doodle poll and try not to have the time conflict with the other WG
-
Use cases
- Josh made a wiki: https://github.com/pmix/pmix-standard/wiki#use-cases
- What are the basic pieces of information we need for use cases?
- In the PR for #181, that will have the templates for PRs and issues. Would we use those as a basis for the use cases, modified appropriately
- Brief description
- Long description
- Why PMIx is the right interface to support it
- How else could you do it, examples of other support
- References to papers or other examples where useful
- (Josh) Move notes to PMIx Standard wiki
- (Josh) Create a PR form of Issue 181 for reading.
- (Josh) Sketch a template issue for use cases with a single example
- Open discussion on suggest template for these use case issues
- Create a few Use Case Issues on GH to experiment with the template and set some examples for new participants.