[IronPython] IronPython with CherryPy through WSGI Memory issue

[Can Gencer]

Hello,

I am trying to use CherryPy through IronPython. I am using a custom
web server written in C# and I am using the NWSGI handler available
here (http://nwsgi.codeplex.com/) with some modifications to work with
my custom web server instead of IIS.

Everything works after some tweaking done to CherryPy. I re-use the
application callable that is retrieved from the main script that is
compiled into CompiledCode . However the memory usage never seems to
go down, and goes up with every HTTP request, even if I force
GC.Collect() after every request.

The WSGI handler is invoking the delegate returned from the main
python script with some parameters that are standard in WSGI, such as
a start_response delegate  Everytime a request is made, the handler
will invoke the application callable that is stored as a reference.

CherryPy also has a built in web server that is using WSGI that can be
invoked with CPython. When testing that, there are no memory leaks.

Any ideas on what the problem could be?