From mboxrd@z Thu Jan  1 00:00:00 1970
From: Matti Aarnio <matti.aarnio@zmailer.org>
Subject: Re: why NFS is UDP based
Date: Wed, 15 Jan 2003 23:18:30 +0200
Sender: linux-fsdevel-owner@vger.kernel.org
Message-ID: <20030115211830.GY27709@mea-ext.zmailer.org>
References: <Pine.GSO.4.51.0301151505110.17019@mamba.cs.Virginia.EDU>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Cc: linux-fsdevel@vger.kernel.org
Return-path: <linux-fsdevel-owner@vger.kernel.org>
To: Ronghua Zhang <rz5b@cs.virginia.edu>
Content-Disposition: inline
In-Reply-To: <Pine.GSO.4.51.0301151505110.17019@mamba.cs.Virginia.EDU>
List-Id: linux-fsdevel.vger.kernel.org

On Wed, Jan 15, 2003 at 03:12:11PM -0500, Ronghua Zhang wrote:
> I wish I am posting the message to the right place.  I am just curious
> what's the most important reason to choose UDP instead of TCP. And when
> some error happens, will NFS do retransmission ? thanks.

You need to familiarize yourself with the history of NFS.
Who, when, and in what kind of environment did invent NFS.
(Sun, late 1980es, machines with _very_ low power compared
to todays standards, and very slow networks!)

Doing NFS over TCP has been a lot more recent development.

Yes, when errors (like UDP datagram loss -> response timeout)
occur, NFS must do retrying / other error processing.

> Ronghua

/Matti Aarnio