From mboxrd@z Thu Jan 1 00:00:00 1970 X-Spam-Checker-Version: SpamAssassin 3.4.4 (2020-01-24) on polar.synack.me X-Spam-Level: X-Spam-Status: No, score=-1.3 required=5.0 tests=BAYES_00,INVALID_MSGID autolearn=no autolearn_force=no version=3.4.4 X-Google-Language: ENGLISH,ASCII-7-bit X-Google-Thread: 103376,99ab4bb580fc34cd X-Google-Attributes: gid103376,public From: dewar@cs.nyu.edu (Robert Dewar) Subject: Re: Q: access to subprogram Date: 1996/07/12 Message-ID: #1/1 X-Deja-AN: 168177598 references: organization: Courant Institute of Mathematical Sciences newsgroups: comp.lang.ada Date: 1996-07-12T00:00:00+00:00 List-Id: iJon said "As several have pointed out, this is not particularly convincing. The best evidence indicates that this was a red-herring in this instance." Nope, this was not a red herring, it (the concern about backend reimplementation) was the major focus of the argument. Remember that this feature was being discussed very late on. Jon, did you read all the references that Norman pointed to, they will give you a clear picture of the discussion on this issue. The reason it was discussed late on was Bill Taylor's late plea to reexamine this issue in light of the iterator issue (well worth reading, it is the first of Norman's references). This got a lot of people interested again, and in the absence of implementatoin concerns, it is clear that the design would have been revisited. This late discussion was entirely focused on the issue of importance of this functionality vs impact on existing backends Note that, as I have tried to emphasize before, static links vs displays was an irrelevant issue then and now. I suppose someone could ake the position of the original Steelman and worry about the impact on implementations at a conceptual level, but in fact NO ONE took this position that I can remember. The issue was very much focused on the fact that at least two existing Ada technologies depended on displays (you should also carefully look at Randy's concerns about shared generics -- as always Randy was the one defending the shared generic approach, and there are some interesting concerns here). Tuck is certainly right that the reason that this originally disappeared was in the general mood of simplification, although some would feel that the restriction here is not a simplification but rather an instance of introducing arbitrary restrictions that make things more complex. But the failure of Bill Taylor's arguments with repsect to iterators to succeed in getting the feature in was definitely the above consideration, and as I say, you should read all the articles -- of course the articles only capture a part of the argument -- you should also read Jim's excellent minutes of the Villar's meeting where this issue was discussed and finally decided on.