> Does anyone know if it's safe to draw power from DTR with a circuit made of
> L78S05, MAX232 and (of course) 16F84.


"Safe" as in "can't damage the RS232 port":

Yes.  RS232 ports are designed to tolerate shortcircuits between any of their
pins.  Most RS232 port drivers are damaged by ESD events, which CAN happen
at insertion of ANY equipment.



"Safe" as in "works correctly with all RS232 ports":

No.  Simple reason:  The MAX232 is supposed to talk to the RS232 port correctly.
To generate the weakest allowed signals, it needs minimum X mA at 5V.  X is
the bare minimum to comply with RS232 specs.  Confirmed?

Now, the host RS232 port is allowed TOO to generate the weakest signals that
comply with the SAME specs.  Imagine, the PC designer has chosen a MAX232,
and supplies it with exactly X mA at 5V.  This is not the most common situation,
but perfectly legal with the RS232 specs. Confirmed?

Finally, the X mA current drive the host MAX232, the RS232 cable, your 78x05,
and your PIC.  What is LEFT of that current is surely less than X, isn't it?
Your MAX232 will receive X-Y mA and NOT comply with RS232 specs anymore.


If a RS232-powered MAX232 would give a valid RS232 port, I could use THAT to
run another MAX232, and hey why not run another MAX from that, and ...
It would give a nice 8-way MIDI interface with no power supply required!

Reminds me of that perpetuum mobile thing.



You should change your design to require less power!

For example, use an LP2950-CZ50 as regulator, and a low frequency XTAL
at the PIC. Put it asleep where possible.  For the RS232 side, use a
discrete transistor or FET interface, which either shorts the RX line
or DTR to TX.  Program the host software to put DTR+RTS high, and RX low
(half duplex).

This will work on many ports, but still NOT ALL.



Or, simply use a power supply. Then you are perfectly in-spec!