Tom Moran wrote:
> What might cause a programmer, when given a new assignment,
> to say to his boss "Let's try Ada for this"? What might
> cause him to spend his own money on a book to learn something
> about Ada, or to spend his own time to, say, download Gnat or
> ObjectAda, and experiment?
A fair question. You could ask the same thing about
Java. Why are they downloading Java tools and playing
with them? Why are they buying Java books by the
It's because they 1) have been persuaded by the hype
that Java is the a) the next big thing and/or b) cool;
or 2) are attracted by this or that feature of Java.
The point of my last few messages about marketing and
Ada have been that now is the right time to "claim"
real-time systems specifically for Ada. If our hype
was directed towards this end, then a certain amount
of momentum might be generated to persuade a new
generation of developers that Ada 95 is the "big thing"
for real-time systems.
People can also be attracted to try Ada because of
some of its features. I always emphasize tasking
features when I "sell" Ada. It's a very, *very* cool
topic. At Tri-Ada '96 when Tucker Taft had a face-
off forum with a Java dude from Sun, Tucker stated
that he believed that Ada had a lot of "steak",
whereas Java had a lot of "sizzle". I interpreted
him to mean that Java had "sex appeal" to programmers
but that Ada did not, and mentally I said "what
about tasking?". Whenever I teach Ada, everyone
wants to learn about tasking. It's got panache.
(Under the "tasking" heading I am including all aspects
of concurrency and real-time features.) It's a world
of fun to learn and play with. I can envision a number
of articles in pop magazines like ESP or JOOP that
highlight different aspects of Ada concurrency features,
specifically protected records, asynchronous transfer
of control, rendezvous variants, etc.
Programmers are attracted by features and the extra
level of power that they feel they gain when having
access to and understanding of a feature. When
promoting Ada, it's important not simply to claim
"Ada is so wonderful, it's so great, you should
really consider Ada". Get down to specifics by
describing a single feature in detail. I've done
this a number of times recently in my office with
the ATC ("select ... then abort ...") feature of
Ada 95, to great effect.
Generally, it's necessary to walk someone through
the semantics of the feature and given them a few
examples of how it can be used to solve this or
that problem. In a way, you must be the teacher in
a mini Ada 95 class focused on a single feature.
This is how almost all of the "practical programmer"
columns are written in magazines like "C++ Report".
Introducing and illustrating a feature has a
tendency to fire the imaginations of programmers.
Their fingers begin to itch -- they want to use
the new, fun feature in their own work.
> Imagine yourself one of the best programmers in your shop,
> your boss and colleagues know it, and you feel confident in
> your abilities in your current language. What would cause
> you to spend time and perhaps even money investigating Ada?
For programmers in a "shop" ("we're a C shop, we don't use
Ada/Java/Eiffel around here"), you've got to show them how
to use Ada (or any new technology, for that matter) without
disturbing the world they are already working in. This
means that you have to be able to show them that they
can use Ada for some things even if they can't use it
for everything. The first step is to illustrate how
some Ada feature would be useful to them, but which is
unavailable in whatever language they are using. After
this, it's necessary to explain how to do the multi-
language interfacing (I've done this for a number of
people -- it's always a real eye-opener for the
un-initiated) and provide good examples. This is
a heck of a lot easier with Ada 95 than it used to
be in Ada 83, so this approach is even better these
days than it once was.
Let's take an example. Let's say that we get a
programmer in a C shop excited about the possibility
of using ATC ("select .. then abort"). This feature
is not available in C except via some convoluted
POSIX signal mess. Then we show this programmer how
to use this in the context of their current system.
All he or she must do is write an Ada procedure
that "imports" a C function; this is then called in
the "then abort" section of the ATC. Demonstrate
how to link it all using adainit and adafinal.
Voila; you have now initiated them into the world
BTW, Ada's tasking/concurrency features are ideal
for this kind of blending of systems; almost no
other language provides multi-tasking as a part
of the language (Java excepted), and in a large
project on a Unix system direct use of pthreads
and signaling is often confusing and messy. Ada
can be introduced in such an environment as the
"tasking harness" from which the C system is
invoked via imported function calls. Our company
has such an Ada harness for real-time simulation
systems and has delivered it for use in other
companies which are generally C or C++ "shops".
A project at one of these companies decided
to switch to using Ada for its own development
effort after receiving our harness!
Most shops will not be so extreme, but once Ada's
foot is in the door, it will find other users.
In spite of the snide remarks you always hear
being bandied about in the popular press and on
Usenet, I find there's not much deep prejudice
against Ada among C programmers I have worked with.
Of course some people are just that way, but I rarely
encounter it on a person-to-person basis.
Apart from the perception that "Ada is dead/dying"
(which is simply the result of poor marketing)
that causes programmers to shun Ada and acquire
skills that they think will get them good jobs,
the major barrier to introducing Ada into a lot of
"shops" is the lack of compilers for the specific
hosts/targets that they need to use.
For a C/C++ shop which produces a product that must run
on 12 different platforms, Ada cannot be an option if
only 6 of those platforms have Ada compilers. This
will only change if the Ada market gets big enough to
motivate vendors to port their compilers, or if
Intermetrics' new Ada-to-C converter has a highly
portable run-time library.
mailto:[log in to unmask]