Any expected improvements in startup times?

Sun Feb 22 21:24:00 GMT 2004

Hi,
I think I've finally been assimilated. I've found a good use for shared libraries!
I'm working on an application where each .cgi page is a gcj program sitting in
cgi-bin, behind apache. Each .cgi page would be way too large to link statically;
therefore, I use as many shared objects as possible.
But then again, statically linked, the application runs much faster than dynamically
linked. Why?
By the way, I'm having way having worse startup times than purely native .cgi
programs such as cvstrac. Cvstrac has a truly extreme kickstart performance and
absolutely minimal memory usage. Idem dito for qmailadmin and nullmail. These C-based
.cgi programs seem to outperform any competition.
However, I can live with *some* inflated memory usage because cgi processes clean up
after them by dying after serving a request, returning resources to the OS. There is
actually not even a need for garbage collection in such scenario, but I absolutely
need faster startup times.
Short of moving the whole lot to C (printf...printf...duh!), is there a way in which
I could get dramatically better startup times for GCJ-compiled .cgi programs?
Erik