Table of Contents
The htpipe handler implements piping of ashd(7) requests over a named Unix socket. The main reason for doing so would be to isolate an ashd(7) handler program from the effects of restarting its parent process. In particular, it is useful for isolating the entire handler process tree from restarting htparser(1), which may be done particularly often for such reasons as certificate renewal.
If the -S flag is given to start htpipe in server mode, htpipe will start listening to a socket named by SOCKET-PATH, start a handler program as specified by CHILD and ARGS, accept connections on said socket, and pass requests received on such connections to the handler program. It can handle an arbitrary amount of such connections, but it is not implemented for high performance with a large number of connections. It is expected that the ordinary will involve very few connections, usually only one at a time.
If the -C flag is given the start htpipe in client mode, htpipe will connect to the socket named by SOCKET-PATH, accept requests from standard input, and pass such requests over said socket, presumably to an htpipe instance running in server mode on the other end.
By default, when neither the -S nor the -C option is given, htpipe will attempt to connect to the socket named by SOCKET-PATH and, if that succeeds, go on running in client mode, just as if the -C option had been given. If, however, the connection fails, it will fork a copy of itself to run in server mode, and then reconnect to the socket presumably created by that new process.
htpipe is a persistent handler, as defined in ashd(7), and the specified child handler must also be a persistent handler.
Fredrik Tolf <fredrik@dolda2000.com>