Wouldn't it be better to create a lightweight server à-la Mongrel, and have Nginx proxy_pass requests to that server? Hmmm nekovm_cluster (insert homer's voice here).
This way we could monitor the server, so if it crashes we get notified. If it is embedded in Nginx as a module, it would be harder to monitor it. -- Neko : One VM to run them all (http://nekovm.org)
