Path: utzoo!utgpu!news-server.csri.toronto.edu!clyde.concordia.ca!uunet!cs.utexas.edu!samsung!uakari.primate.wisc.edu!ames!sgi!calcite!vjs
From: vjs@calcite.UUCP (Vernon Schryver)
Newsgroups: comp.unix.i386
Subject: Re: UUCP over ethernet
Summary: rcp can "store and forward"
Message-ID: <91@calcite.UUCP>
Date: 4 May 90 06:48:39 GMT
References: <70400005@m.cs.uiuc.edu> <1990May2.044758.22817@cs.columbia.edu> <899@sixhub.UUCP>
Organization: Rhyolite Software, Mountain View, CA
Lines: 25

In article <899@sixhub.UUCP>, davidsen@sixhub.UUCP (Wm E. Davidsen Jr) writes:
> 
>   Not true in mant cases. uucp and uux are store and forward protocols,
> which do not require that both machines be up at the instant you enter
> the command. NFS, rcp, rsh all fail if the other system is down.

There is a fair chunk of news shuffled among a some large sites in the San
Francisco Bay Area using rcp/rsh/TCP/IP/leased-lines.  That is, a trivial
protocol using news batch files and rcp act as a "store and forward
protocol," doing the correct things when one end or the other is down.  My
introduction to the idea was indirectly from Brian Reed of DEC.  It is very
easy to implement.  I much prefer the scheme to NNTP/TCP as a transport
mechanism because NNTP uses an incredible ihave/sendme, one article at a
piddling, context switching, tiny packet time.  I confess that I have not
actually measured the relative CPU costs or network costs.  Looking for
facts when you have a perfectly good theory would be against the spirit of
usenet.

It must be noted that there is some pressure to replace the rcp scripts
with "normal" UUCP/TCP from someone at one of the sites who would rather
reduce the number of mechanisms he has to maintain.


Vernon Schryver
vjs@calcite.uucp