Re: Ada2005 clear screen etc.

["Robert I. Eachus" <rieachus@attbi.com>]

Preben Randhol <randhol+news@pvv.org> wrote:

>As a external library. I don't see the need for it. I mean next will be
>that we should have some sort of GUI standard in the Ada standard.

Let me try to answer this from the ARG standpoint, plus some history. 
Way, way, back in the Mil-Std 1815 (Ada 80) days, I/O was an integral 
part of the language.  By ANSI/Mil-Std 1815A (Ada 83) I/O required no 
special features within the language, and there was a set of standard 
packages, Direct_IO, Sequential_IO, Text_IO, and Low_Level_IO, that were 
required by the standard.

In ANSI/ISO 8652 (Ada 95) these packages were moved into Annex A, 
predefined language environment, along with other old language features 
and a few new ones.  In addition several specialized needs annexes were 
added for things that were only of interest to small segments of the Ada 
population.  The idea was that compilers need not implement all of the 
specialized needs annexes.  In practice, the major front ends, including 
GNAT have tended to implement all of the annexes, it really isn't all 
that hard.  (All you need is someone who can read and understand each 
annex. ;-)

But the annexes in my mind have been a success, because the design of 
the annexes forced us to really work hard on limiting any low-level 
extensions required to implement an annex. For example, the Information 
Systems Annex requires a lot of compiler independent software which need 
only be written once.  (Ben Brosgal, Dave Emery, and I wrote a prototype 
of those operations under contract to the AJPO, and the source code was 
made public domain.) It also requires supporting a decimal type with at 
least eighteen decimal digits, and that's about it.

Also the there are some dependencies between special needs annexes, but 
with a structure that makes sense.  (For example, it makes no sense to 
implement the Real-Time Systems annex without the Systems Programming Annex.

But when Ada 9X was in development we spent a lot of effort deciding 
what should and should not be added to the standard.  The NRG had done a 
lot of work over a decade developing math packages.  In 1982, the 
consensus was that adding a bad numerics library to the standard would 
be a mistake, and we couldn't delay the Ada standard until they were 
ready.  By the early nineties they were ready, consistant with the IEEE 
work, and there was little argument that they should not be added. 
Similarly, for political reasons, support for other character sets was a 
requirement.

But even though we put a significant amount of effort into developing an 
SQL interface, the consensus was that it wasn't something that belonged 
in the standard.  (And in my opinion a good thing too.  There were three 
proposals each with its strengths and weaknesses, and I have never 
really used any of them for more than a two or three page program. 
However there are now some very nice database bindings that don't 
require a blinding headache--and that were not available in time for Ada 
95.  If someone were to take the GNADE work: 
http://gnade.sourceforge.net/doc/gnade.html and propose all or part of 
it as part of the new standard, go ahead.  Personally, I find the thin 
ODBC binding easy enough to use, but I think that a thinker binding 
would be more appropriate for adding to the standard.

So what will be added to Ada 0Y?  The only real consensus at this point 
is that there are some container libraries that should be considered. 
Check the Ada Issues database http://www.ada-auth.org/~acats/ais.html 
for details.