Though I've never used the PM for UUCP, let alone done any measuremnets,
I think this is wrong - or rather a misunderstanding. Certainly
"portmaster protocol" should be more efficient per se than telnet or
rlogin (though I'd guess the difference is smaller for something like
UUCP than for human interactive use), but the question was about "telnet
to port 540", where a "uucp daemon" is presumably listening.
This has got to be the most efficient alternative, as uucico will end up
sitting directly at the other end of the TCP connection - whereas in the
other cases (or for that matter telnet to the standard port 23), the
bits go from the kernel to a user-level daemon (rlogind/in.pmd/telnetd),
then back into the kernel through a pseudo-tty, and finally back out
again to the user-level uucico - i.e. three context switches instead of
one, and lots more code executed.
Whether this will actually be noticeable unless you have lots of UUCP
connections, and whether there are other problems such as transparency
(i.e. is the "telnet" client side in the PM transparent when talking to
a non-standard port - I'd guess so, but don't know), is another thing of
course... There is also the "netdata" alternative, don't know exactly
what it is and don't have the docs at hand, but from what I remember
reading about it, it might be The Right Thing for this (pointed at port
540, that is).
--Per Hedeland
per@erix.ericsson.se