python connect to server using SSH protocol

[Toby Dickenson]

On Tuesday 08 February 2005 13:26, Simon Anders wrote:

> This is what I was about to reply as well. But I did a short test 
> program and encountered a problem:
> 
> import os
> fi, foe = os.popen4 ('ssh dopey')
> print >>fi, 'ls'
> fi.close ()  # <-- this is annoying
> for line in foe:
>     print line,
> foe.close ()
> 
> The above connects to a server, passes the command 'ls', which is 
> executed there, and prints the returned result.
> 
> However, reading from foe succeeds only if fin has been closed before. 
> An fi.flush() seems to be not sufficient. 

But this version below does work. Im not sure whats happening when using the 
file as an iterator to make a difference.

import os, sys
fi, foe = os.popen4 ('ssh xxxxx')
print >>fi, 'ls'
fi.flush ()  # <-- this is annoying
while 1:
    b = foe.readline()
    sys.stdout.write(b)


> But if one wants Python to  
> interactivly communicate with some shell on a remote machine, it is 
> inconvenient to have to close and reopen the connection all the time.

But this route carries a big deadlock risk. the rsync program can tunnel over 
ssh this way, but have first hand experience of
http://www.google.com/search?q=rsync+ssh+deadlock

In the script above, if 'ls' is replaced with a longer input then there is 
every chance that the fi stream will block before all of it is written, 
because this script hasnt started draining foe.

For a real python program using ssh (but not 'interactive'... data written to 
fi does not depend on data read from foe) see
http://dirstorage.sourceforge.net/replica.html
-- 
Toby Dickenson