...........
More specifically, with CGI the webserver starts a new process for every single request. That's bad enough for a light C program,
but it's certainly not a good idea to start a whole new Python process for every request. At least not for any production website
or web service that serves any real amount of traffic.
That is, for those who didn't know, the reason why CGI fell out of use quite
some time ago.
Well I'm afraid I cannot agree with that reasoning. ReportLab used cgi exactly because it starts a new process; most of the
reports that are generated take more than a second to generate so the startup time is of less importance. Starting up in a clean
state is of importance because the sharing of resources across different reports often leads to buggy report code such as relying
on an earlier report to load fonts, define colours etc etc. This could perhaps have been done with multiple interpreters eg
mod_python, but I think the isolation there is not perfect especially with C extensions.
It might be that the new 'cgi' doesn't use the older api and in that sense we could just use wsgi or whatever the new interface
is, but I would still use cgitb to provide nicely formatted traceback html.
--
Robin Becker
--
https://mail.python.org/mailman/listinfo/python-list
