Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!usc!barnard.usc.edu!jconti
From: jconti@barnard.usc.edu (John Conti)
Newsgroups: comp.dcom.sys.cisco
Subject: Re: replace config
Message-ID: <28451@usc>
Date: 27 Nov 90 21:44:59 GMT
References: <1990Nov19.205855.18509@arcturus.uucp> <12641311988012@heap.cisco.com>
Sender: news@usc
Lines: 52
Nntp-Posting-Host: barnard.usc.edu

JOHN@heap.cisco.com (John Wright) writes:

>...... The
>problem with a replace command would be having to erase the active
>configuration and therefor halt the running of the system while we
>parse the brand new configuration information, not something you
>normally want to do on an active running production network. It is
>more likely you would load a file containing commands with both
>disabling some services with "NO ..."  commands, and then turning
>services on with perhaps difference addresses or parameters later
>in that file.  The contents of the active configuration file in RAM
>are written to NVM with the write memory command.

I often have to erase the NVRAM, reload, and config from scratch
when I make signifigant changes to a configuration.  Because of the
above it involves downtime which neither the users, nor the
administrator (me) like.

I was wondering if a diff program could be written for the cisco
configs.  The program would know enough so that it could take a
human readable (comments, command differences, and different order
of presentation) config file and generate a "diff" between that
and another file that contained the active config or the config in
NVRAM (previously downloaded from the router).

This way you could change your config and just generate the "NO"
commands neccessary to change the active config correctly.  This
first occured to me because I used to try to do the above by hand
and would always make some little mistake, and my boss would get
all over my case because the running config wouldn't match our copy
on disk somewhere.

This is made worse because we don't use tftp to download the
configs.  All of our boxes have NVRAM and are expected to work
independant of supporting hosts.  I have to use serial lines to
re-gen our routers.

I think all you would need to do this would be a complete listing
of the parse tree for the configuration language.  This looks
simple and maybe cisco even has a yacc file or something. Would
cisco feel like releasing and updating such a document?

--
	,John

John Conti	{ jconti@usc.edu | jconti@xenon.usc.edu (NeXT mail) }
Computing Services, University of Southern California, (213)740-2957
-- 
	,John

John Conti	{ jconti@usc.edu | jconti@xenon.usc.edu (NeXT mail) }
Computing Services, University of Southern California, (213)740-2957