@manuel @david I think # would be better if unspecified assumptions were discovered and specified. One of the best ways to do that is for someone to reimplement it (not necessarily as part of the GS project, but sharing info with the project). Obviously, I think # will be the area where most such assumptions will be found. I do agree that # is the best language for the project, at least for the next few years, until # and # hosting almost entirely replaces shared-hosting. (Some shared hosts claim to offer # hosting, but have out-dated versions 2.2 or 2.4.)