Re: Ada95 references needed.

[mfb@mbunix.mitre.org (Michael F Brenner)]

In the last year I converted several projects from Ada-83 to Ada-95,
including HYTLIST (the hypertext link insertion tool, about 20,000 LOC),
a command and control system (800,000 LOC, of which 250,000 were
non-comment lines), the Braid Group Laboratory Software (about 10,000 LOC),
and some smaller tools. The largest of these (the command and control
software) required 300 lines of code to bring it up from a dialect
of Ada compilable by only one Ada-83 compiler to good Ada-95. HYTLIST
required no changes. The Braid Group Laboratory had problems with
the late changes to Ada-95 regarding unsigned numbers and wound up
costing a few weekends. In general, conversion from Ada-83 to Ada-95
does not take very much time for fully tested Ada-83 code. If you do not
have test programs to see how the Ada-83 code is supposed to work,
then you may have to write the test programs first. 

The biggest question you probably need to ask is not which programs
are 100 percent Ada, but rather, how easy is it for Ada to use
software and global variables from C, Java Byte Code, and Assembler
on the systems you will be connecting to. The big advantage of Ada-95
is the ease of substituting code in one language for another. I had
a contest with a veteran C lover writing a device driver (they wrote
it in C and I wrote it in Ada); the Ada code wound up running faster
(because of arcane rules of the hardware that I happened to know
more about); the Ada device driver was plugged into a system with
more lines of code in assembler and C than in Ada. 

Conversions from C to Ada might be replaceable by Ada calling C or
C calling Ada. That is one of the real strengths of Ada-95.

Mike Brenner mikeb@mitre.org