Re: Clear Screen

[dewar@merv.cs.nyu.edu (Robert Dewar)]

Marin said

<<    Would one practical solution be to throw the features out, but
    insist that compilers maintain an "Ada95 Switch" which makes them
    available for the people who still might need upward
    compatibility? Then after "Ada0x" is superseded by "Ada1x", the
    "Ada95 Switch" becomes invalid - thus providing users plenty of
    time to migrate their code. The advantage is that the features
    aren't really there for the programmers building "Ada0x" code, but
    you haven't hurt those still maintaining Ada95 code.>>

This is not worth the effort, because you would still have to formally
define the old features. Note that in Ada 95, we eliminated the old
Ada 83 attributes, not because we expected compilers not to support them --
any reasonable Ada 95 compiler *will* support them -- certainly GNAT does, 
but because it would have been too much effort to explain in Ada 95
terms what they meant.

In practice Annex J is such a small collection of such small features,
that it is simply not worth introducing incomaptibilities by trying
to get rid of them. You get rid of things to simplify the language or
to clean it up, but in this case, the simplication and clean up is
minuscule in the total scale of things anyway.

Annex J ends up being little more than a comment by the designers on
what they felt they would have liked to leave out. A lot of people
disagree with the choice, including me, but I really don't care to
argue the point strongly, since Annex J features are as much a part of
the language as anything else.

As I mentioned before, the experience with ALTER in COBOL 8X should
show just how hard it is to remove features. In this case the COBOL
committee was pretty much unanimous in agreeing that this feature
should be removed, but a large insurance company (name withheld :-)
threatened to sue ANSI and the individual members of the committee
if they took this action, on the grounds that there were commercial
interests represented on the committee which were trying to destroy
the integrity of their legacy code -- ALTER stayed in :-)