Issue L - Progress references in General Index - for RC-4-June

[RolfRabenseifner]

Problem

This is an addendum to Issue K=#468 and PR564,
see last item of the comment there: https://github.com/mpi-forum/mpi-standard/pull/564#issuecomment-799978311

Proposal

Adding references about "progress" to the global index for:

• In Chapter 3 Point to Point Communication: small updates
• In Chapter 6 Collective Communication, Section 6.14 Correctness:
  Examples 6.31 and 6.36 on pages 306 and 308
• In Chapter 11 Process Initialization, Creation, and Management, Section 11.2 The World Model, 11.2.2 Finalizing MPI:
  Example 11.6 + advice to implementors, page 496 lines 28-48
• In Chapter 11 Process Initialization, Creation, and Management, Section 11.3 The Sessions Model, 11.3.1 Session Creation and Destruction Methods:
  MPI_SESSION_FINALIZE (here, the final wording does not yet exists)
• In Chapter 11 Process Initialization, Creation, and Management, Section 11.6 MPI and Threads, 11.6.1 General:
  Page 517 line 28
• In Chapter 12 One-Sided Communications, Section 12.7.3 Progress: The section title should be referenced.
  (This section includes the interesting rationale on page 619 lines 5-25, which helped a lot to understand the weak-local characteristics of the definition of MPI's local procedure and this rationale includes the summary sentence cited above.)
• Chapter 13 External Interfaces, Section 13.2 Generalized Requests: Page 631 lines 25
• Chapter 14 I/O, Section 14.6.3 Progress: Page 711, line 8
• Chapter 14 I/O, Section 14.6.5 Nonblocking Collective File Operations: Page 711, line 48.

Page refs to https://github.com/mpi-forum/mpi-issues/files/6136676/mpi40-report-rc-4-june-F-PR560.pdf

Changes to the Text

Important: This issue does not change any binding text, advice, rationale or example.
It only adds missing references to the general index. And it formats the referenced keyword "progress" as \mpiterm, i.e., italic.

See PR https://github.com/mpi-forum/mpi-standard/pull/569 and mpi40-report-rc-4-june-L-PR569.pdf

Impact on Implementations

None.

Impact on Users

Progress rules are easier to be found in the MPI standard

References and Pull Requests

https://github.com/mpi-forum/mpi-standard/pull/569

[Wee-Free-Scot]

This might be an issue that should be postponed into the MPI-4.1 timeframe.

There are no "emergencies" here, except perhaps for the new API being added for MPI_SESSION_FINALIZE (because it is new text). However, that is a blocking procedure and is therefore covered by the general statement of progress that all blocking procedures must guarantee progress for all outstanding operations (@RolfRabenseifner found the reference, page 619 lines 5-6).

[RolfRabenseifner]

See also pdf: mpi40-report-rc-4-june-L-PR569.pdf

[RolfRabenseifner]

This might be an issue that should be postponed into the MPI-4.1 timeframe.

There are no "emergencies" here, except perhaps for the new API being added for MPI_SESSION_FINALIZE (because it is new text). However, that is a blocking procedure and is therefore covered by the general statement of progress that all blocking procedures must guarantee progress for all outstanding operations (@RolfRabenseifner found the reference, page 619 lines 5-6).

Especially for MPI_SESSION_FINALIZE, it is more than only to guarantee progress:
MPI_SESSION_FINALIZE (same as FINALIZE) must guarantee the completing of all needed progress before it returns.
See mpi40-report-rc-4-june-L-PR569.pdf page 496, lines 26-48.
For the moment, we provided the possibility that MPI_SESSION_FINALIZE can do this final progress, but we mention it only for MPI_FINALIZE. Of course this issue add the required entry to the genaral index, see
progress --> finalize --> buffered data, 496

[RolfRabenseifner]

I'm not sure whether we treated index entries as chapter committee or editor changes.

[Wee-Free-Scot]

The "final progress" is a good point. IMHO, that one item should be split out into its own issue and PR, so that it can be proposed as an "emergency fix" for MPI-4.0 -- but the rest look like existing carry-overs from MPI-3.1 and before.

We will need to iterate on wording for that today, so it can be presented to the Forum at tomorrow's virtual meeting.

[Wee-Free-Scot]

At this stage of MPI-4.0, I suspect all changes require a vote -- @wesbland can help us with the rules.

[wesbland]

I'm not sure whether we treated index entries as chapter committee or editor changes.

@RolfRabenseifner Is fine here. My general philosophy was that anything that didn't change the PDF (I know this adds entries in the index, but those are extremely minor), I marked as something that we didn't necessarily have to vote on. If it's inside of the 4-week window, the ruling gets tighter.

[RolfRabenseifner]

Okay, I'll try to get my reviews before the 4-week-deadline.

[RolfRabenseifner]

@wesbland This means, the wish to change "Progression rule" into "The progress rule" in normal chapter text is a text change and therefore will require a non-no-vote on next meeting? Or can such chapter committee changes still go witout a vote before the 4 week deadline?

[wesbland]

Yes. It will need an issue to track it and a no-no reading/vote.

[RolfRabenseifner]

@wesbland, @Wee-Free-Scot, and @tonyskjellum:
I removed all text changes from this issue and started a new one only on coll.tex to substitute the 4 locations of progression into progress. See Issue M = #492.

[RolfRabenseifner]

@wesbland It looks like that you removed a wrong issue from "ready to merge".
In the protocol pptx from last meeting, https://github.com/mpi-forum/mpi-forum.github.io/raw/master/slides/2021/03/2021-03-24-rcm-frm-procedures.pptx
you wrote on slide 11:

Issues to continue for MPI 4.0 (12):
......, #491

i.e., this issue here.
Its pull request https://github.com/mpi-forum/mpi-standard/pull/569 is ready and 5x approved (including Bill Gropp).
There is also no comment that these general index entries should not be done. As co-author of the general index, it is also my duty to add such entries (as also Bill did over the time).

Therefore, I expect that you removed this issue by mistake. I was the last days completely busy, therefore I catch this only now. Therefore, please can you put it back to 4.0 Ready to Merge.

[wesbland]

Yes, that's correct. This one should be in the "In Progress" column until it's been reviewed by everyone that needs to review it.