Telnetd flow control

Doug Graham dgraham at nortel.com
Tue Feb 3 00:57:13 UTC 2009


On Tue, Feb 03, 2009 at 12:45:14AM +0100, Denys Vlasenko wrote:
> On Sunday 01 February 2009 19:48, Doug Graham wrote:
> > Bottom line is that the server should not be enabling the TELNET_LFLOW
> > option unless it is prepared to actually implement the rest of RFC 1080.
> 
> Done. TELNET_LFLOW is removed:
> 
>                 static const char iacs_to_send[] ALIGN1 = {
>                         IAC, DO, TELOPT_ECHO,
>                         IAC, DO, TELOPT_NAWS,
>                 /* This requires telnetd.ctrlSQ.patch (incomplete) */
>                 /*      IAC, DO, TELOPT_LFLOW, */
>                         IAC, WILL, TELOPT_ECHO,
>                         IAC, WILL, TELOPT_SGA
>                 };

I think this is good enough.

I've just been experimenting with your patch and mine, and the normal
flow-control behaviour in response to a ^S is definitely much snappier
with my patch than yours.  But I've also been experimenting with the
openssh 5.1p1 client and server on x86 boxes, and it seems that they
don't implement local flow control either, at least not by default.
In fact, they seem even less responsive than telnet without local
flow control, possibly because they use bigger buffers somewhere.
Since ssh's lack of local flow control has never bothered me before,
I think I can live with busybox's telnetd not implementing it either.
As long as emacs works, I'm happy!

Thanks,
Doug.


More information about the busybox mailing list