Re: SGI GNAT Question? (Long)

[dewar@gnat.com]

In article <7bov12$r8o$1@nnrp1.dejanews.com>,
  dennison@telepath.com wrote:

> The company I work for makes no secret of using all sorts
> of "unsupported" Open Source tools (gcc, emacs, perl,
> tk/tcl etc), and apparently makes a pretty good living
> for itself doing so. I suspect you'd probably be
> hard-pressed to find a fourune 100 company that *didn't*
> have someone using a public version of one of those tools
> to do production development.

Whether you find it useful to use unsupported software
depends on your circumstances. Our policy at ACT is not
to use unsupported software for any critical purposes,
and that is the advice we pass on to our customers.

> Generally the support you get from fellow users on usenet
> is far superior to what any company could provide.

That may be true in general, but certainly our customers
do not consider it is true for the support that ACT
provides which is certainly NOT what is generally typical
in the field. Most certainly Ted cannot make a relevant
judgment here since he does not have ACT support. We have
certainly seen a number of occasions on which Ted has been
frustrated to run into problems that would have been solved
immediately if he had support. It all depends on how you
value your time, and how much of a problem it is if you
run into a blocking problem.

> And in
> a real pinch, you can go into the source and fix a
> problem yourself.

That's not really true, not "in a real pinch". Yes it is
certainly true that you can support GNAT yourself, and
indeed this possibility is important to many of our
customers. But there is a learning curve here, and you
need people who have some real knowledge of the GNAT
sources.

The scenario where you run into a problem, and then quick,
in emergency mode burrow into the sources. Furthermore if
you are selling this as a kind of insurance policy to your
management, it is bogus, given the need to ramp up in
people and experience before you could effectively solve
the problem.

Sure there may be limited cases in which this case be done,
and often customers for example make fixes or modifications
to run time units, but digging into the visibility circuits
of the compiler to figure out if a puzzling error message
is in fact a bug in GNAT or is a user error, and then
fixing it in the former case is not something you can
reasonably expect to do in an emergency.

> If we do *need* guaranteed support from experts for a
> particular project we would be happy to pay for it.

Of course one of the things here is that you don't really
know whether support would benefit you without trying it
out. This is why we discourage people from using the
public version of GNAT for evaluation. Instead, get an
evaluation contract from us, and evaluate for yourselves
how useful support can be.

A very large part of our support consists in helping people
with problems in their understanding of Ada or GNAT, or how
best to use these technologies. We also help people find
bugs in their programs in many cases as part of sorting
out what is going on.

I often see people flailing around on CLA, or chat@gnat.com
looking for help on problems that I am certain would be
solved quickly by our appropriate expert if they had an
ACT support contract.

As I say, it all depends on your requirements and needs,
and on how valuable your time is.

What we often see is serious projects trying to use GNAT
without support getting into a mess, and then deciding that
the mess means there is a problem with GNAT. Well that may
be true in some cases, but nearly always the problems could
be eliminated or at least managed in the context of proper
support.

That's why our position is that basically the public
version of GNAT is intended for student and research use.
If others find it useful fine, but that is not the audience
we are catering to with the public releases.

Robert Dewar
Ada Core Technologies


-----------== Posted via Deja News, The Discussion Network ==----------
http://www.dejanews.com/       Search, Read, Discuss, or Start Your Own