Re: Proving Correctness (was Java Portability)

["Marin David Condic" <dont.bother.mcondic.auntie.spam@[acm.org>]

Well, the situation probably lies somewhere between the two extremes. If the
software library we are talking about is fairly general-purpose in nature,
this is one thing. If it addresses some narrow niche, then its another.

I get the impression that we were talking about some semi-general-purpose
library that would be distributed with a compiler. Something similar to the
MFC - interfaces to the OS, graphics, data structures, etc. Notice that
Microsoft didn't go to one or more customers and say "Please fund the
development of this end product". (Of course at the size they are, there is
a lot more capability to develop things like this on speculation.)

If "Ada-Libraries-R-US" wanted to build a similar library of development
tools - one that has a market beyond one company or even one compiler
company - I don't think you'd find very many companies that would contract
for its development and expect to pay the full development cost. Someone
might partially fund it, realizing that a) they need it and b) there has to
be some seed money to get it going and c) this ends up cheaper than doing it
themselves. Even this, I think is doubtful and as evidence, I would observe
that there are companies that would benefit from having such a library and
there are any number of eager entrepeneurs, start-ups or existing hungry
companies that would want to build such a thing. Yet no such relationship
exists. Why?

I think it is because most customers are not willing to totally or even
partially fund a development of this kind. Its too risky & they have
alternatives. Why pay all or some of the cost of developing a library in Ada
if they can go to C++Compilers-R-Us and buy the whole thing for a fraction
of what it costs to get it built from bottom-dead-center in Ada? Because
then it would be in Ada? Why does a business care if it is in Ada, C++ or
Lisp? They care about how it impacts the bottom line & (say) $50,000 to
develop it in Ada vs $500 to buy it COTS in C++ has to be justified in those
terms.

Put yourself in the shoes of the customer (as I'm sure you are from time to
time) and ask why it is you buy or don't buy some software product. When
does your company out-source the development of any software? Why do they do
it? What conditions are placed on it? What advantages do they get and what
do they give up when they do this? These are the important questions to ask
if we want to understand why/why-not some company would want to fund the
development of an Ada library of some sort.

BTW: My experience has been that very few companies ever outsource custom
software development of any kind. (By this I mean contracting with an
outside company to build something they want that they can't get
off-the-shelf. Not the business of hiring consultants or other indirect
labor to work on it in-house.) Why is that? I can speculate, but perhaps its
more valuable to get the speculation of others.

MDC
--
Marin David Condic
Senior Software Engineer
Pace Micro Technology Americas    www.pacemicro.com
Enabling the digital revolution
e-Mail:    marin.condic@pacemicro.com
Web:      http://www.mcondic.com/


"Larry Elmore" <ljelmore@home.com> wrote in message
news:3B75C79C.38C670F2@home.com...
>
> I think your TV example was flawed because developing a software library
> has nowhere near the risk if the company you're dealing with has a good
> track record, and nothing has happened to it recently that might affect
> its ability to carry out the job. Software libraries can also usually be
> used in a restricted fashion long before they're complete, which is a
> far different situation than a factory, which must be functionally
> complete before it does anything useful.
>