[J3] (SC22WG5.6178) Re: (SC22WG5.6176) [ukfortran] October meeting visa invitation letter

[Ondřej Čertík]

Hi Steve and Anton,

On Tue, Apr 28, 2020, at 1:06 PM, Shterenlikht, Anton via J3 wrote:
> Steve
> 
> > On 28 Apr 2020, at 18:23, Steve Lionel <steve at stevelionel.com> wrote:
> > 
> > In my mind, the biggest step we can take is to get work-list papers written between meetings, with tuning by small teams of interested parties, so that when we do meet in person we can get things done faster. We don't need any equipment to do that, just a shift of mindset and a willingness to siphon off some time from "the day job" to make it happen. (I understand, though, that some members simply may not be in a position where they are allowed to devote resources to J3 between meetings.) I'll note that often J3 votes to change or choose among approaches with straw votes, so we can't simply present papers as a "take it or leave it". Again, needs some thought and discussion.

We need to try it, then refine the process as we go. As long as there is willingness to make it succeed, we can find a way.

> I'm very much against drastic changes.
> 
> Assuming we are talking about
> an evolution, not revolution,

Yes, evolution, not revolution.

> I was trying to imagine
> what remote participation
> might mean for me for the
> October WG5 LAS meeting
> within the current setup.
> 
> UK is +8 hours to LAS time.
> The 0800 plenary discussion
> and voting is 1600 my time.
> I can probably read the papers
> submitted for the plenary,
> during my day and participate
> in the plenary via some remote
> audio/video two-way tool.
> 
> Provided that tool works well,
> and the morning plenary lasts
> no more than about 2 hours,
> i.e. end by ~1800 my time,
> this will probably work,
> i.e. I can convince myself
> that my contribution was useful.
> 
> I can probably agree with my
> management that for that week
> I do WG5 work from lunch onwards.
> I estimate that 3-4 hours is
> enough for me to read and understand
> the papers submitted for voting
> each day, and prepare an opinion.
> 
> In fact, for papers from subgroups
> in which I don't participate
> this is not very
> different from what I do
> when I participate in person.
> Indeed, this might even be better
> because I'll read the papers in
> day time, rather than late at night.
> 
> The subgroup discussions will
> happen while I sleep, so I will
> not participate in those.
> I guess I can work via email
> on papers assigned to my subgroup,
> but the 8h time diff will
> probably mean the progress
> is too slow to be useful.

I suggest that we submit all papers ahead of time to the j3-fortran.org website (1 month would be ideal, but at least 2 weeks ahead) as we do now, but in addition also track them as issues at the J3 GitHub repository (https://github.com/j3-fortran/fortran_proposals) and get real discussion on each paper. For the last meeting we have already done exactly that for 4 papers, here they are for reference:

https://j3-fortran.org/doc/year/20/20-106.txt
https://j3-fortran.org/doc/year/20/20-107.txt
https://j3-fortran.org/doc/year/20/20-108.txt
https://j3-fortran.org/doc/year/20/20-109.txt

As you can see, each submitted paper references the corresponding GitHub issue with the associated discussion.

This is the current state, and the change I suggest is to do extend this process from 4 papers to the majority of papers (eventually all, but we don't need to achieve that at the next meeting). I.e., evolution, not revolution. (Also the above papers were for 202Y, but we should do the same with 202X papers.)


In addition I suggest we set a plan ahead of time, what papers will be considered each day at plenary, and which papers will be skipped (if any) and why. And that we get an agreement with the community that this is a good plan. In particular, here is the list of all papers that need to happen so that we can release 202X:

https://github.com/j3-fortran/fortran_proposals/issues/162#issuecomment-593671596

Many of them are already done, so our goal should be to submit papers for the rest of the items. I talked with Steve some time ago, and I think we both agreed this is a good plan. So let's work as a community to make it happen. 

As Steve said, sometimes as we meet in person, we realize that we need to modify some papers, something that we missed while discussing it online. As an example of this for the 20-107 paper, which we discussed a lot ahead of time online, but then at plenary there were very good suggestions that would require to improve the paper before voting. The summary of the discussion is at the corresponding GitHub issue:

https://github.com/j3-fortran/fortran_proposals/issues/22#issuecomment-591566803

Since this was a paper for 202Y, we didn't spend time at the meeting to modify it. But had this been a paper for 202X, then we can improve it right away at the meeting.  Preferably in a transparent manner at GitHub (thus giving the outside community a chance to see the changes and comment) and resubmit.

These are not drastic changes, it's just an evolution of what we are already doing anyway. Let's do more papers ahead of time as I outlined above and see how it goes.

Ondrej