On 21 Nov 2007, at 1:35 am, Kyle McDonald wrote:

> Why Packages? Why response files?
>
> Why not have the finish script key on domain name, or IP network, do  
> a lookup based on hostname, and then have it create the ntp.conf (or  
> any other config file) on the fly?

Separation of content in install mechanism.

We kick off our own install process from the finish scrip.  It reads a  
definition of third party or in-house packages and patches that need  
to be install on top of the OS (+ patches), as well as - yes - scripts  
acting on input set before the jumpstart.  The mechanism has changed/ 
improved over the last 8 years only slightly, but the definitions of  
content changed a lot, moved from Solaris 8 to 10.  We were even able  
to call it as the none existent finish script for zone installs.

Sun is now supporting packages and patches outside the OS distribution  
in the jumpstart profile recognising the above need.  But no  
equivalent to sysidcfg so far.

Another example.  NFS v4.  The nfsv4 domain id.  We all had to put our  
finish.sh "hacks" in to stop the early Solaris 10 releases from going  
interactive at first boot (which is the equivalent to "run this at  
first restart" which was mentioned by Bart).  That isn't really any  
good if you want to have automated installs, is it?  So the NFS v4  
domain had to be moved into sysidcfg.  Sun can do that.  We can't.

Again, even if I say "sysidcfg API", that's not what I mean.  I'd like  
the sysidcfg problem to be given some thought in the brave new world  
and design it such that it accessible to the end user.  I haven't read  
everything, but the bits that I have read didn't cover this.

- mo


Reply via email to