On Mon, Oct 19, 1998 at 06:29:24PM +0200, Kragen wrote:
> On Mon, 19 Oct 1998, Andi Kleen wrote:
> > On Mon, Oct 19, 1998 at 01:47:57AM +0200, Alan Cox wrote:
> > > 2. Killing connections on a drop. This potentially violates the RFC
> > > check rules on time wait unless you are very careful. Also tell me
> > > why it cant be done in user space by turning /proc/net/ into a set
> > > of temporary 'reject' filter rules
> >
> > The RFC TIME_WAIT rules only make sense when the connection endpoint
> > (address/port) still exists. For a dynamic address that is gone that isn't
> > the case.
>
> Until a few months ago, I could frequently get the same IP address at
> my ISP by redialing within a few seconds. (Now I have a static IP and
> don't have to worry about this.)
>
> It would be a shame if losing the ppp0 interface had killed all my
> telnet sessions.
No problem, just don't set the IFF_DYNAMIC flag. It is just that there
are other people who aren't that lucky. You have the choice.
> (Interestingly, data I send over a TCP connection while ppp0 is down
> seems to get lost. When I reconnect, I can send more data, but the
> data I sent while the connection was down seems to have disappeared
> into a black hole. This causes problems with trn. I'm using 2.0.30.)
There were a few bugs in 2.0 TCP related to retransmit and route changes
because of some design problems. I believe they were mainly kludged around
in later 2.0.x releases (x>30). In 2.1 the problem was solved completely
by a rewrite.
-Andi
-
To unsubscribe from this list: send the line "unsubscribe linux-net" in
the body of a message to [EMAIL PROTECTED]
