Re: Ada Book

[eachus@spectre.mitre.org (Robert I. Eachus)]

In article <3gh8t9$cak@cronkite.seas.gwu.edu> dobrien@seas.gwu.edu (David O'Brien) writes:

 > One nice thing about Ada (I believe) was that the standard was out
 > before there were any/many implementations.  Thus everyone was playing
 > from the same sheet of music.  In a way this was bad though.  I believe
 > the language designers would have gotten a little insight by first
 > implementing an Ada compiler before unleashing that job to others.  I
 > may be wrong, but an Ada compiler can be *quite* hard to implement and
 > wasn't that way it has taken so long to get good compilers (especially on
 > small machines like the PC and Mac).

    Gee, you would think that the government would have been smart
enough to require a compiler as part of the langauge development
contract.  (For the humor impaired, read the fifth paragraph of the
Forward to the 83 RM.)

   The problem, and it was a major problem, was that the test
translator was developed on Multics, which was a very different
environment from most systems of the day.  In particular the automatic
support of dynamic binding made the correct implementation of (Ada 80)
libraries easy.  With the wide review as part of the Ada 83
standardization process, the Ada library "evolved" to something easier
to implement on other OS's of the day, but lost a lot of the
functionality of the original.  (For example, Multics "first reference
traps" allowed the order of elaboration of library units to be
determined dynamically during execution.  Ada 83, if you read the RM,
still allows this, but it practice the ACVC tests forbid the most
elegant solution.)

     I hope now that most serious OS's support dynamic linking, some
Ada 95 compilers will take advantage of this and eliminate any linking
or binding step.
--

					Robert I. Eachus

with Standard_Disclaimer;
use  Standard_Disclaimer;
function Message (Text: in Clever_Ideas) return Better_Ideas is...