Re: Ada PIWG

[jjs@sei.cmu.edu (Jeffrey Stewart)]

In article <14062@grebyn.com> ted@grebyn.com (Ted Holden) writes:

>The literature and every conversation I have ever had with real-world 
>people who have been forced to actually attempt to USE Ada have totally 
>convinced me of the failure of this project.  By the same token, C++ 
>could very nearly, if not entirely, fill the bill.

Well Ted, I guess you never talked to me or the group I worked with at
Hercules Defense Electronics.

We have just finished a project to implement all of the control software
plus some signal-processing software, in Ada, for a Millimeter-Wave radar
seeker for the Maverick missile.  

Its about 7500 lines of code, based on an object-oriented design, does 
signal-processing, provides dual-loop control for the antenna, provides 
mission control for the entire missile, plus some other things.  And it 
does it at 600Hz, on a 10MHz 80286.  There is NO ASSEMBLER.  The goal 
was a complete Ada artifact, and the goal was achieved.

We wanted a complete Ada artifact because part of the goal was for the
implementation to be as portable as possible.  We achieved this, in part,
by rewriting some of the system requirements to be event, rather than timing,
driven.  There was zero loss of functionality and performance due to these
changed requirements.

We had NO problems with the language.  We did have one problem with the
compiler generating incorrect code for a representation clause, but this was
a compiler implementation problem which was easily fixed by the vendor.

Projects that DO have a problem with Ada tend to fall into the following 
categories :

     The classic "10 pounds of stuff into a 5 pound bag" problem.  You won't 
     solve this with C either.  Many times, the requirements stack the deck 
     against you.  Period.

     Less than production-quality compilers.  I'll admit this was a problem, 
     and is less so every day.

     Whiners who will use every excuse to blame the language rather than the
     design.  The two ARE different, design and implementation, don'cha know?
     Many of these projects fail due to poor design, regardless of
     implementation language.

I refuse to assign percentages to these categories.  Let that be an exercise
for the reader.

By the way Ted, part of the rationale for Ada was that the techniques and
languages in current use (mid 70's) would be inadequate for the million-plus
SLOC systems anticipated for the 90's.  What are the successful C-language
projects of this size?

I like Ada not because it is Ada, but because, in my estimation, it allows
the best use to-date of all the software engineering principles I know and
have seen used effectively.  Most C code still looks like line-noise to me.

To sum up, Ada DOES work, for its intended purposes.  Maybe you should
consider becoming an equal opportunity basher, i.e., cite some C/C++
failures as well.  (And don't tell me there aren't any.  Your use of
absolute terms is part of what detracts from your credibility.)

Jeff