The latest instance of this is at www.pypi.info, and it isn't the first time I have seen this person making incorrect statements about mod_python and how it works.
So, let me set the record straight on one such myth about mod_python. That myth is that mod_python creates a separate Python interpreter instance for each request. This is just not the case.
What actually happens is that the first time a request arrives for an application which hasn't yet been loaded, and which hasn't been marked to run in the main Python interpreter, is that a new sub interpreter is created. That sub interpreter is then persistent within that process though, not being destroyed until the process itself is destroyed. Any subsequent requests for that application are then handled within the context of that same sub interpreter which has already been created, so no new sub interpreter is created. Even when multithreading is used, all the request handler threads execute within the context of that one sub interpreter. So, neither is a separate sub interpreter created for each distinct request handler thread in the Apache thread pool, as is also sometimes claimed.
As to why that poster had the problems he had with mod_python, they are more likely because of which Apache MPM was selected when Apache was compiled, as well as not changing the default MPM settings used. Especially with prefork MPM, the default settings are more appropriate for static file serving and PHP. If you do not change the default MPM settings as well as tweak HTTP keep alive settings to make them more appropriate for a large persistent Python web application, you will see load spikes and memory issues, especially when a system is put under load.
From what I have seen, the majority of people setting up mod_python don't understand the need to change the default Apache configuration. It may well have been the case that back in time when mod_python was somewhat newer that you didn't have to, but this is because back then Python web applications were typically much smaller and so didn't have significant start up times or use large amounts of memory. These days though they do, so you need to change how Apache is configured to have it perform adequately.
Now, can we please stop trying to do performance and memory usage comparisons if you don't know how to setup the systems being compared properly. :-(
BTW, everything above also applies to mod_wsgi when using embedded mode, as it is implemented in a similar manner to mod_python. Because I'd rather not see these misconceptions about mod_python transfers onto mod_wsgi, I will blog later more about the real source of these load spikes and memory issues. In the mean time, if using mod_wsgi use daemon mode and you will not have to think about it, as these issues in the main only relate to embedded mode.