To: J3 J3/19-237 From: Malcolm Cohen Subject: Non-sibling access masquerading as sibling access Date: 2019-October-14 1. Introduction Paper 19-180 contains three purported interpretation requests glommed into a single paper. This paper discusses the second of them. 2. The alleged problem From 19-180: "Section 9.6, paragraph 3 explains the semantics of image selectors that specify TEAM=, TEAM_NUMBER=, or that specify neither. There are several problems with this paragraph. ... "The TEAM_NUMBER= specifier in an image selector may specify a value of -1 if and only if the current team is the initial team. This is different than the behavior of the TEAM_NUMBER arguments of the IMAGE_INDEX and NUM_IMAGES intrinsics which allow -1 to indicate the initial team, no matter what the current team is." 3. Response from /INTERP Firstly, it is unclear why the intrinsic functions in question should have any bearing on this facility. TEAM_NUMBER is, as noted, not a specifier in an intrinsic function reference but an argument, and intrinsic functions frequently have special cases. Secondly, and more importantly, TEAM_NUMBER= in an image selector is for accessing sibling teams. It would be quite surprising for an image selector written with TEAM_NUMBER= to access an ancestor team instead of a sibling team. Accessing ancestor teams is the purpose of the TEAM= specifier. Thirdly, the requirements on TEAM_NUMBER= for coarray establishment make it crystal clear that the access cannot be to an ancestor team but to a sibling team. Permitting such accesswould introduce an inconsistency. Therefore /INTERP concludes that this looks more like a new feature request, and one that is not obviously desirable. Therefore /INTERP politely declines to accept this request at this time. Instead, /HPC is requested to review its history to see whether this alleged missing functionality was both omitted by accident and is so highly desirable that we need to add it to the standard right now. If the conclusion of /HPC is that this was an accidental omission that warrants immediate correction rather than being processed as a new feature request, /INTERP will accept it and process it accordingly. 4. Comment It would appear that the arbitrary decision to number the initial team with the special value "-1" has prompted some unusual and interesting design decisions and questions. In hindsight, perhaps "+1" might have been better (this would have removed the possibility of such strange and unusual designs and interpretations). ===END===