[1089] in linux-net channel archive
Different behaviour of 1.2 and 1.3 netcode concerning lost connections
daemon@ATHENA.MIT.EDU (Lutz Pressler)
Sun Sep 10 19:53:40 1995
Date: Sun, 10 Sep 1995 22:05:40 +0200
From: Lutz Pressler <Lutz.Pressler@AMS.Med.Uni-Goettingen.DE>
To: linux-net@vger.rutgers.edu
CC: iialan@iifeak.swan.ac.uk, Lutz.Pressler@AMS.Med.Uni-Goettingen.DE
Hello,
in the following typical situation the behaviour of 1.2 and 1.3 kernels
differ:
"client" running that kernel version (and the same network tools in both
cases) has a PPP dialup connection to "server" (running Linux 1.2.x or 1.3.x,
that makes no difference for users on "client"). User on "client" has multiple
TCP connections (especially rlogin) to hosts reachable via that link.
If the modem drops carrier (and thus the ppp link is lost) under 1.2
the tcp connections are kept alive for a certain amount of time. The
ppp connection can be redialed and everything continues..
With 1.3 this is _only_ the case if there has been nothing in the Send-Q
for that connection. If the user (has) type(s/d) anything in an rlogin
connection, for example, while the ppp link goes down, that connection
is closed.
The 1.2 behaviour is really more conveniant. Is there any reason it has
changed? Is there maybe even a socket option the rlogin/telnet/...
client should set?
Have a nice day,
Lutz
--
Lutz Pre"sler <URL:http://www.AMS.Med.Uni-Goettingen.DE/~lpressl1/>
Systemverwaltung -- Abt. Medizinische Statistik, Universit"at G"ottingen
Humboldtallee 32, D-37073 G"ottingen, Tel.: +49(0551) 39-9774 FAX: -4995
<Lutz.Pressler@AMS.Med.Uni-Goettingen.DE> [PGP-key:WWW&Keyserver] IRC:lp
