Cannot bind any address

Markus Werle numerical.simulation at web.de
Fri Oct 5 19:35:52 EST 2001


Markus Friedl wrote:

> i think the message is already very appropriate.

Yes, I agree, for someone who knows what it is all about
the message is _very_ appropriate.

So do not hesitate to change it ;-)

I just remember a lecture about computer interface
ergonomy, where we found out that most error messages are
perfect, but sometimes do not help the user to fix the problem
anyway.

We learned about several accidents in nuclear power stations
where correct error messages were misinterpreted by
humans in situations where one does not think clearly anymore.

Several airplane crashs led to the conclusion that
in case of an engine failure the pilot is _not_
interested in a message about pressure loss in tube 233-56,
but rather in a red highlighted engine on his display with
a single "engine failure" message.

We do not have to get paranoid about it, but
I see some analogy for open-ssh:
The bind problem has a reason, but the error message
does not tell me this reason. I have to think about it.
So something is - well - not wrong, but improvable.

I have to know that child processes do not have a bind problem
when the father process is killed. I have to know they
can live even when another sshd comes up. I have to
know that only the first sshds conflict with each other,
not the forked ones.
Since I am not very interested in implementation details
of ssh I just overlooked this and got it plain wrong for
15 minutes.

Conclusion:

Although the message is correct and already very appropriate
it lacks some ergonomy from the viewpoint of people who
investigate human-machine-interaction .


Markus


P.S.:
In the actual discussion about airplane security some people
would like to see airplanes under tele-control in
critical situation. I hope they use open-ssh for that purpose,
just to achieve maximum security.

Since airplane developers take a deep look at software
ergonomy, it still may be a good idea to assume
maximum stupidity on user's side (that's me :-))..

I can imagine critical situations where an ergonomic
error message may save hundreds of lives.

> > feature request:
> > sshd detects a concurring sshd is running and gives an
> > appropriate error message.

Like this (takes 5 minutes to change the source code):

Cannot bind any address. Maybe another sshd is up and running.





More information about the openssh-unix-dev mailing list