A few random observations...

   * Perhaps Ada is not shrinking, but the rest of the software industry is
growing so fast it is tough to see it as the same.  I would hope you can
build 100 web sites in the same time it takes to build a single airplane.
In reality, Ada has not made nearly the effort to be a super-web language
as Java has, and an enormous percentage of software work is either building
web stuff, or making existing stuff web enabled.   CGI is fine, but does
not have nearly the comfy standardized framework as Servlets, JSPs, and
EJBs provide.

   * Language conversion is very seldom a 'good timesaving choice'.  Been
there, hated that.  While it may be possible to translate to erect similar
or even identical compiled code, it does not usually help at the source
level, and hardly leads to architectural improvement.  I would love to hear
an XP methodologist's opinion on code translation, Ha!  Refactor this!

   *  The most disappointing thing about promoting Ada is the dramatic leap
from OO concepts to source code.  While it is unfortunate that UML choose
the word Package for a grouping mechanism, it is still a great boon to
users of Java and to some degree C++ to have instant translation from model
to code.  Syntactically, I love Ada's readability, but am left wanting with
the way it describes OO features from a largely structured architectural
mechanism.  I would love to see other opinions of this (probably off-line
to be polite) but after teaching Ada for 3 years, a year with Java has done
more to help my understanding and growth in OO concepts since it really
requires you to buy in or always swim upstream.

Thanks everyone for waking up and making this list start buzzing again!