Re: What's stopping you from using Ada for your next commercial project?

[Stephen Leake <stephen_leake@stephe-leake.org>]

Gerd <GerdM.O@t-online.de> writes:

> 2. Pricing for GNAT PRO would be to much for such a small development
> team (I think they offer support for at leat 10 person teams). Our
> customer will not pay more, only for the software to be written in
> Ada.

I don't understand this argument. Using Ada instead of C, I get easily a
factor of 2 (I usually say 10, but let's be conservative) productivity.

So you can use 2 people instead of 4, freeing up 2 for another project.

People cost at least $100k per year, so you have that to spend on an
Ada toolchain.

GNATPRO is around $25k per year.

What is the problem? I'm guessing you don't really believe the
productivity factor.

There is a learning curve with any new toolchain, but you imply that's
not an issue.

> 3. Software development is not only done on Windows and Linux.

Just out of curiosity, what development OS are you using? Or are you
refering to the target OS?

> 4. Ada is not available for many of the processors that are used in
> embedded range (e.g. NEC 850, Infineon TriCore, or - still used - 8051/
> HC11), and if there would be an Ada compiler (in fact a complete tool
> chain is needed, including RT support for the bare board), then the
> costs are far beyond of what would be acceptable (see Green Hills fo
> example).

This is a real problem, although the SoftCheck method of compiling to C
and then using the C toolchain for a final step is an option. You will
probably have to use a subset of the Ada runtime, but that's still far
better than C.

> 5. For a lot of projects not only a tool chain is needed, but a
> "safety certified" tool chain is required.

You seem to be implying that you can get a safety certified C toolchain
for less than you would need to pay for GNATPRO. I simply don't believe
that! But maybe there is an economy of scale at work; if someone can
sell 100k copies of a certified C toolchain, it will be cheaper than
GNATPRO. 

> 7. There is no experience how Ada would fit into the Autosar
> architecture, that is used in automotive range.

Yes, this is a learning curve, which requires up front investment. It
will pay off (factor of 2 productivity).

> 8. Ada tools are not adopted to systems often used in embedded range
> like Windows CE or OSEK.

Same as point 4.

> PS: Developers with Ada know-how are not a problem here.

That's good news!

-- 
-- Stephe