Skip site navigation (1)Skip section navigation (2)
Date:      Sun, 19 May 2002 10:25:18 -0700
From:      Michael Sierchio <kudzu@tenebras.com>
To:        freebsd-net@freebsd.org
Subject:   Re: HEADS UP: ALTQ integration developer preview
Message-ID:  <3CE7DFFE.2090809@tenebras.com>
References:  <Pine.BSF.4.10.10205170216500.29826-100000@ady.warpnet.ro> <Pine.LNX.4.44.0205171056200.2091-100000@scribble.fsn.hu> <3CE55A9B.73EA3DE4@mindspring.com> <Pine.LNX.4.44.0205181018300.10011-100000@scribble.fsn.hu> <3CE61675.BCE2A9E1@mindspring.com> <1021717195.1466.4.camel@gurney.reilly.home> <3CE6D592.DCF73743@mindspring.com> <20020519001249.GA24012@roughtrade.net> <3CE6F653.CDE9D2B4@mindspring.com> <20020519010703.GE24468@roughtrade.net> <3CE7508D.36568484@mindspring.com>

next in thread | previous in thread | raw e-mail | index | archive | help
Terry Lambert wrote:

> UDP is still a bad bet for reliable request response.  It's
> really dumb to effectively reimplement TCP without windows on
> top of UDP just to avoid using TCP.

Speaking as someone who has implemented reliable message protocols
over UDP about a dozen times, I can affirm Terry's point.  Once
you find yourself reimplementing TCP, it's time to use TCP.  I
think that there are some very good uses for UDP-based protocols
(gee, DNS seems to work, but falls back to TCP for responses larger
than size N), but it's tricky.  If it's not an authenticated protocol,
you leave yourself open to a whole class of DOS attacks, akin to
RFC 1644 T/TCP.  Datagram reassembly in user space is for the birds.
Consuming resources for pending operations can choke you to death.

TCP is constantly improving,  though the improvements sometimes
have unwanted side-effects -- congestion-control algorithms
break for wireless, where packet loss might be due to a neighbor
using a cordless phone rather than congestion.

Back to problem of NFS over UDP -- it's not so stateless, is it? ;-)
Remote disk access is mostly bulk transfer operations anyway,
why wouldn't you use TCP?


To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-net" in the body of the message




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?3CE7DFFE.2090809>