P0900 Concerns - insufficiently vetted by Kokkos team

[hcedwar]

P0900 was generated quickly from some initial discussions. This proposal has not received sufficient vetting with the rest of the Kokkos / ISO-C++ team (@hcedwar, @crtrott, @dsunder, @mhoemmen). However, the 2018-03-Jacksonville pre-meeting deadline is too soon to allow sufficient vetting and this topic is too important to not submit a paper.

Decision: Revise abstract to note that this is initial exploration of this design. If the paper is taken up by LEWG at 2018-03-Jacksonville then the objective is to prompt discussion and collect perspectives of the available experts for a subsequent revision to (1) include more stakeholders and (2) make a concrete proposal.

[dhollman]

@hcedwar have you sent it in?

[hcedwar]

In the next hour or so

[dhollman]

What concerns? I was worried about this, which is why I only put my name on the paper so far.

[hcedwar]

It is not just an individual name on the paper, it is the organization you are representing. Sandia / the Kokkos project is sending you, you are not just representing yourself.

[dhollman]

@hcedwar I know that; I didn't mean to imply otherwise.

[crtrott]

OK I am pretty happy with the paper now after more understanding and a bit of discussion with David.

[mhoemmen]

@hcedwar I did actually vet the paper (and proofread it too!). I think it makes a critical point about the risk of proliferating properties, and how we might mitigate that risk. Some of the details, like operator()'s suggested implementation, scare me a bit. Nevertheless, it's worth writing them down and discussing them, so that people realize the implementations. (Otherwise, "properties should be orthogonal" is just a platitude, like "don't kick puppies.")