Client Logging

I sometimes need to be able to identify what client application is causing
messages that are appering in the backend log file.  To do this I want the
client to identify itself in the log file.

I can see 3 ways this could be implemented:
1) An identification string passed when a connection is established.  This
string would either be immediatly logged or included in the log entry for
any subsequent messages for that connection.
2) A sql statement to print to the log file.
3) A function that takes a string argument and calls elog.

#1 would be ideal but requires digging into the source code.  #2 seams wrong
headed.  I can implement #3 locally myself.

Before I go any farther, has something like this already been invented?