I feel you can solve this problem in an easier way, like this:
- Place your regexes and actions in a config file. This would be Perl code placing the regular expressions in a data structure (hash, array, whatever).
- When the script starts, it sets up the sockets and eval()s said config file, thus learning the last set of rules.
- Install a signal handler for SIGHUP. Upon receiving this signal, you can wipe the data structure and re-eval() the config file, thus learning the new rules and actions.
The user only needs to modify the config file and kill -HUP
the process id of your script. You can also get fancy and detect changes on the script, though I do not advise this. One of the advangtages of this method, is that the config file can be reasonably checked by using perl -c config-file
. You can also do this by eval()
ing the config file in a separate namespace first (checking for errors in $@
) and if all goes well, proceeding to the real namespace.
In fact, you could accomplish the last phase by placing your script in a separate namespace, say:
# Your script goes here
The config file could be explicitly placed in the main
namespace. You can then delete all non-built-in symbols in main
deleted. This would get rid of even the actions defined as sub
s in the config file.
I have a similar thing in production which works fine. It handles around 20 RADIUS authentications per second :)