[3395] in linux-net channel archive
Re: your mail
daemon@ATHENA.MIT.EDU (Rob Janssen reading Linux mailingl)
Sat Jun 22 08:48:51 1996
From: linux@pe1chl.ampr.org (Rob Janssen reading Linux mailinglist)
To: olaf@bigred.inka.de (Olaf Titz)
Date: Fri, 21 Jun 1996 13:15:35 +0200 (MET DST)
Cc: submit-linux-dev-net@ratatosk.yggdrasil.com
In-Reply-To: <m0uWspc-000J4OC@bigred.inka.de> from "Olaf Titz" at Jun 21, 96 01:08:00 am
Reply-To: linux-vger@wab-tis.rabobank.nl
According to Olaf Titz:
> IMHO that would only need as much buffer space as would actually fit
> in the outgoing data rate. I.e. the (worst possible) maximum lifetime
> of an IP packet is 255 seconds, after that it gets killed anyway; at
> 14.4kpbs you can (nominally) transfer 459000 bytes in this time.
> Down to the common TTL of 64 it's only 115200 bytes.
While the TTL in theory (and in the specs) is expressed in seconds, in
practice it is a hop counter.
You normally just set a "reasonable" time limit on an outgoing queue by
limiting its size to a bitrate-dependent value.
(normally this is forgotten in the "first try" implementation of IP routing
software, and the system will crash when someone pumps data into the
router at a higher rate than the outgoing connection allows...)
Rob
--
+------------------------------------+--------------------------------------+
| Rob Janssen pe1chl@amsat.org | BBS: +31-302870036 (2300-0730 local) |
| AMPRnet: rob@pe1chl.ampr.org | AX.25 BBS: PE1CHL@PI8WNO.#UTR.NLD.EU |
+------------------------------------+--------------------------------------+
