On Oct 27, 2019, at 1:38 PM, Rolf Rabenseifner <[email protected]> wrote:
The index entries are well chosen according to relevance, i.e., not always just all Locations.
have to admit that I was always confused by the "MPI Declarations Index" -- I thought it was buggy because it was so incomplete. There is no description there of your stated intent. All it says in MPI-3.1 is: "This index refers to declarations needed in C, such as address kind integers, handles, etc. The underlined page numbers is the “main” reference (sometimes there are more than one when key concepts are discussed in multiple areas)." There's a lot of references for each type, but it is nowhere near complete. There's no indication that only *some* of the references to handles are listed. This led me to believe that it was erroneous. It sounds like there are three possibilities: 1. List all the pages where all the types are used. 2. List some of the pages where some of the types are used. But don't say why. 3. List some of the pages where some of the types are used, and give an explanation / rationale as to what exactly is shown, and why. I think we're in state 2. Either state 1 or 3 would be significantly better. I took a stab at state 1, but I don't have strong feelings about it. Specifically: it sounds like you'd prefer state 3. If you have specific rationale for why specific selections were made, can you amend the introduction text to the MPI Declarations Index? I'm guessing this can be an errata, or possibly even a ticket 0...? -- Jeff Squyres [email protected]