>translating from german to english what i want to say is hard business ...
>i know that telnet from H0 to H1 and then to get a config from server
>F1 per ftp
>is a very, very stupid example.
>but this is not my focus.
>(for the test i've made few minutes ago i decided to telnet to a host and then
>send a "uname -a" on this host. ;-))
>my focus is on the fork-> Maybe somebody can say due to his/her
>experience if my
>code is right or false?
>i've never tested this script with forking with 500 routers. it's only in my
>testlab with the 3 suns.
>i ve tested a similar script with 30 machines but the load of the
>expect runs on increases much.
>page 446: i will think about it. i can't understand it not yet.
To my shame, I don't read German.
I see a couple of issues. One is how to fork. Another
is how to manage network monitoring of hundreds of hosts.
All my experience is that [fork]ing is NOT part of the
solution of the latter. Yes, you need some sort of con-
currency to monitor all your hosts satisfyingly; forking,
though, or even a thread-based concurrency model, will
NOT give the performance and reliability you're after.
You need an event-based model, or, in Unix programming
terms, one founded on select().
That's the subject of page 446.
A couple of tasks remain. One is to analyze your source
code, and explain why [fork] gives you results that
surprise you. More urgent than that is to write an
[expect_background] example or two to help you make good
use of this interface in your development work.
It might be a week before Don, I, or someone else can help
you more with these items.