Another thing that might be hampering your survey is that we had a
similar survey in this list last year, organised by Carlos Cantu. You
may find it helpful to look at that survey as part of your research.

You may also want to look a similar survey by Evans Research, done
last year. The report is not available for free, but there are many
press clippings about it.

> I think you have a very valid point. I agree that the survey is end-
> user centric. We are basically trying to assess the reach and
> acceptance that Firebird has achieved so far in the marketplace and
> amongst developers, its strengths and weaknesses as per the actual
> users and its readiness for Enterprise level applications.

Can you disclose who this survey is for? If not by name, perhaps by
anonymised description?

As you probably realised, answering your question through a survey is
somewhat problematic. You may recall the 1997 paper by Bob Young
(then the CEO of Red Hat, these days the man behind lulu.com) "Sizing
the Linux market", which discusses the problems in finding a workable
methodology.

Customer and instance counts are not easy to find in the RDBMS world.
For a example, I have not found a single public reference for those
two numbers for Oracle. However, you can find many snippets of
information in the investor presentations and annual reports of MS,
IBM and Oracle.

Measuring open source RDBMS numbers is even harder, as it combines
industry secrecy with Bob Young's problem. People tend to go with
proxies. For instance, MySQL acceptance was long derived by counting
internet-facing web servers running PHP and doing some math. As
MySQL's share in PHP is falling, this no longer works.

For a long time, measuring DB2 meant counting the number of AS/400
boxes and doing some math. With the AS/400 world shrinking and DB2
moving to other platforms, this no longer works.

The general proxy for counting Firebird would be to count the number
of Delphi and C++Builder developers and applications and doing some
math. With Firebird deployments shifting to Java, C# and PHP, this no
longer works.

So, problems abound. My estimates would be:
MySQL: 5 million instances, falling
SQLServer: 4 million instances, rising
FB/IB: 3 million instances, stable
Oracle: 1 million instances, rising
DB2: 1 million instances, falling

Hope the above helps.

Paul