On Mon, 19 Mar 2018 at 12:33, Nevin Liber <nevin@eviloverlord.com> wrote:
[adding SG12]

On Mon, Mar 19, 2018 at 1:43 PM, Geoffrey Romer via Lib-Ext <lib-ext@lists.isocpp.org> wrote:
1. The straw poll results recorded in the Jacksonville wiki (10-4-3-1-3 for constructing string_view from null) are inaccurate.
2. The straw poll results are accurate, but do not constitute consensus.
3. There was consensus in the LEWG meeting room in Jacksonville, but that decision has somehow been superseded.
4. There is consensus for this change in LEWG, but not in LWG, WG21, and/or national bodies.

I believe the answer is 4 (1-3 are ridiculous, given that both Nathan and I were in the room).

There was at least one person who told me ahead of time they would have voted "no" if they could have been in the room, but because it was extremely unlikely they could escape their room (and in fact they couldn't), they didn't have a vote.

This proposal was not seen by SG12, even though it clearly has to do with defining previous ub.

I think we should have a discussion within SG12 as to whether we consider changes like this to be within our remit. I think some library changes clearly are (eg, I expect we'd want to be involved in discussions related to the behavior of less<T*> or vector<const T>), but lots of library functions have narrow contracts and I don't think we should consider all changes that widen any of those contracts to necessarily require our input. This case seems somewhat borderline to me.

There was a 93 message thread on this very thing in 2016 <http://lists.isocpp.org/lib-ext/2016/10/3138.php>, so clearly more than just "the people in the room reviewing every LEWG paper" have an interest in this.

Talking with folks afterwards, I know that there are a few people who would have voted "no" had they been in the room, but I readily admit there is a non-zero chance that some or all of those folks might have just been humoring me while I was ranting. :-)

I think you should also consider the possibility that they may have changed their mind had they been present for the presentation and discussion of the paper. While the makeup of LEWG during the discussion is unlikely to be exactly representative of that of the committee as a whole, the people involved in that discussion are likely the best-informed people on the topic within the committee.

Honestly, I'd rather discuss it instead of going through the dance of objecting during whatever plenary this gets put forward and voting on it there.

Procedurally, I consider it to be a valid and useful response of LWG or CWG to observe that they are very far from having consensus for a particular proposal or design point thereof within their group, and to accompany the proposal back to the corresponding evolution group to reconsider in a combined LWG+LEWG or CWG+EWG session. The combined view of LWG+LEWG or CWG+EWG is probably closer to representing that of the committee as a whole than that of LEWG or EWG alone.

Another valid response, particularly if you can bring new information that was not available in the original discussion, would be to write an opposition paper for presentation in LEWG, and make sure that all people with strong opinions on both sides of the debate are invited to the discussion of that paper.