loadable modules ?

Hi

I've implemented a sms service type that loads and calls a library
dynamically (using dlopen).

The configuration would look something like:

group = sms-service
keyword = default
name =
"mod"
# name of function called, append ".so" to get filename    (Is it ok to
use the name property this way? What is the intended purpose of name?)
load = "this string is passed to the library function %p %P %a"
# passed to function "mod"
omit-empty = 1

The library call protocol is:

Octstr* func( Msg* msg, Octstr* pattern );          /* return reply
text, msg == NULL -> library is about to be unloaded. */
where func is the name specified in the sms-service group.


The purpose is to offer a way of extending kannel functionallity without
changing kannel source code.
It can be considered a plugin architecture for sms service types.
It is usefull for sms service types that benefit from reusing resources
having the same lifetime as the process that use it (in my example, a
mySQL connection).

Currently the implementation is a bit "hacky" and needs cleaning and
"packaging" before I can submit it.

My question is: should I bother? Is there a need/want for this kind of
thing?



         /Dennis

  Dennis Malmström
  Erda Technology AB
  S:t Larsgatan 12, 582 24 Linköping, Sweden
  http://www.erda.se
  mail: [email protected]
  phone: +46 (0)13 377218
  gsm: +46 (0)706483090