In Tokyo, Guy shared a couple concerns:

  1. The definitions of concepts in the proposal would be looked at skeptically because they are impossible to change once created. My response was that a graph (adjacency list) is a range-of-ranges, which is different than the one-dimensional ranges in the standard today. Because of that, they warrant their own concepts.
  2. After showing early work using coroutines as a possibility for implementing more versatile BFS, Guy was concerned that it might be viewed with skepticism as a "novel" approach.

Since then I have added comments to all concept definitions with "For exposition only" in both the paper and the code.

In our last SG19 meeting with Andrew, Michael and Jens we discussed Guy's concerns:

  1. There was agreement that marking concepts as "For exposition only" was a reasonable approach and that LEWG would weigh in later.
  2. It was pointed out that iterating on a tree using a coroutine-based generator was a common example used, so the use of coroutines for something like DFS and BFS wasn't novel.

I am working on comparing both coroutines and visitors WRT performance and usage.