Hi Alexey,
Yes, I agree and understand what the RFC 793
mentions in this regard. As Andy clarified, RFC takes
the view that the SYN is an old one... so it doesnt
work correctly in the case when the SYN is a new
one. Ofcourse, if you had the TimeStamp with SYN
packet, the problem would be solved ..
perhaps, cleaning up the half-open connection
(as opposed to waiting for 7 retries before freeing it)
would be a good idea, in addition to sending an RST
back..
I am basically trying to argue that we should try to
bring the system back to a usable state as quickly
as possible. (currently,
what happens is untill the server side TCP times
out the open-req, the server port is unusable for
that client.. eg if the server was a web server, that
client wont be able to reach the web server untill
the server times the port out or client uses a different
local port, which sometimes you dont have a control over)
I hope I made the problem clear.
Thanks,
Rajeev
----- Original Message -----
From: <kuznet@ms2.inr.ac.ru>
To: "Rajeev Bector" <rajeev@akamai.com>
Cc: <ak@suse.de>; <linux-net@vger.rutgers.edu>
Sent: Monday, July 17, 2000 11:12 AM
Subject: Re: multiple TCP sockets with same (srcip,dstip,sport,dport)
| Hello!
|
| > Perhaps, this behaviour should
| > be configurable/sysctable ??
|
| The behaviour is prescribed by TCP specs. Nothing to sysctl.
|
| Look at all those nice pictures in rfc793, which explain what occurs
| with half-open connections and how tcp resynchronizes in this situation.
| Do you have something to add to this picture?
|
| Alexey
| -
| : send the line "unsubscribe linux-net" in
| the body of a message to majordomo@vger.rutgers.edu
-
: send the line "unsubscribe linux-net" in
the body of a message to majordomo@vger.rutgers.edu
