Need an advice : how to design the support of several DB connections ?

Hi,

   I need my radius server to support several DB connection (probably having
different setup and calling different SQL statment). Those connections will
be
choosen according to the radius request content. (My english is not fluent
enough so that I am mind is clear. I call here a "connection" the content of
the "sqlserver" file).

   The reason of my need is the following : I use VoIP boxes that have only
one
IP entry for a radius server. On a logicial point of view, each box can
implement different kind of services. The software implementing the
"intelligence" of those services are different so that I cannot send all the
radius request to the same DB server and execute the same SQL statment.

   I suppose that this kind of feature could be useful to several gnu-radius
users so that I purpose to develop it cleanly and send it to Sergey so that
it'll be included in future gnu radius releases.

   Before starting to code that feature, and because I am not an expert of
the
gnu radius philosophy, I prefer to wait for an advice on the implementation
design.

   So, people coding on gnu radius, would you have an idea to about how to
design implementation of such a feature, using a "natural way" according to
gnu
radius allready implemented code ?

   Regards,

   Georges