Roman Semenenko wrote:
> Volker , is that something like FAI in debian ?
>   

Well, not quite, since I am working towards a different goal. Let me 
explain.

I see FAI as an infrastructure for mass deployment - the more boxes you 
have the better. FAI is excellent software, and I would not see any 
point in writing something new for the same purpose.

The itch that I am scratching deals with the first (and maybe only) 
server, the one you are going to install without having any 
infrastructure, no DHCP server, no nothing. While I was still using 
SuSE, I found myself entering IP addresses and user names into YaST over 
and over. It got better with FreeBSD since I could use an "install.cfg" 
floppy, and with RedHat since I could prepare a "ks.cfg" floppy. 
AutoYaST didn't help, because it is more like a macro recorder. All this 
only got me over installation, not configuration or customisation. I 
still had to prepare backups and restore them selectively on new 
versions, since I never knew what would have changed. It is my credo to 
RESTORE user data from backpu media and to CONSTRUCT configuration data 
from a data model. That is why I like CFENGINE so much and use it at work.

What my little program (written in C so I don't need any interpreters, 
currently around 40 to 50 C files, one exectuable, called ConfGen) does 
is to execute the functions /arch/setup does (fdisk, pacman.static, 
grub). Then it reboots the machine and hooks itself early into the boot 
process and adjusts all relevant configuration files (right now: apache, 
dcron, cups, dnsmasq, dovecot, dyndns, ethers, fetchmail, fstab, hosts, 
imap, networks, nisdomainname, ntp, postfix, ppp, pppoe, printers, 
procmail, rc.conf, lmhosts, samba, squid, ssl, ssh, udev.rules, users 
incl. passwords, vsftp, xinetd, yp, plus working on a few others) in a 
*consistent* way. The input information is made available via a set of 
XML input files without redundancy (e.g. the machine name is given only 
once in system.xml). Documentation is only in the source code for now, 
but I have started to write up something since I wanted to start an OSS 
project *some day* anyhow. And the XML files come with an integrated 
DTD. I am working on this only in my copious spare time, you know :-)

If there is interest for it, I'd be more than happy to make it 
available. Anything I write is under GPL anyway. But I have no idea what 
would be the best way of contribution. Setup a project on GNU Savannah? 
Is there a user contribution area with actual project space or a CVS in 
AUR? I could setup a CVS repository on my own DSL-connected box, but 
that would certainly require some time. Some guidance here is highly 
welcome.

Kind regards
Volker


_______________________________________________
arch mailing list
[email protected]
http://www.archlinux.org/mailman/listinfo/arch

Reply via email to