sysinstall colours

Randi Harper randi at freebsd.org
Sat Oct 17 05:27:45 UTC 2009


On Fri, Oct 16, 2009 at 6:00 PM, Michiel Overtoom <motoom at xs4all.nl> wrote:

>
> On Friday 16 October 2009, Randi Harper wrote:
>
> > Personally if I spent a lot of time on such a project, I'd be sure to
> have
> > the "is this going to make it into freebsd base?" conversation first.
>
> I think there's no doubt about it that 'sysinstall' will feature in the
> next
> FreeBSD too.  It will!  Keep up the good work!  It's worth it.
>

Thank you for the kind words. :)


>
> The sysinstall manual page makes two apocalyptical remarks about itself:
>
> 1. "This product is currently at the end of its life cycle and will
> eventually
> be replaced."
>
> 2. "This utility is a prototype which lasted several years past its
> expiration
> date and is greatly in need of death."
>
> These doomsayings are wrong. To date no serious contenders have surfaced
> and
> up until that time sysinstall does its job, underappreciated perhaps, but
> it
> does it reasonably well, and adequately.  Now that it is back in the focus,
> we can look to a bright, evolutionary future for sysinstall.
>

As much as I want to agree with you, I can't quite do so.

There are (to the best of my understanding) solid reasons why no other
installer has made it into base. This doesn't necessarily mean that
sysinstall is the final answer, though. Eventually, in my opinion,
sysinstall needs to be replaced. It tries to do more than it should. For
example, one of the things I'd like to see removed is the upgrade option -
although I'm expecting quite a bit of backlash on that, so we'll see if that
happens. I also don't think it should manage configuring rc.conf beyond
network interfaces/hostname. The network services configuration is a mess. I
don't think enabling the NFS server via sysinstall even works at this point.

That aside, the code for sysinstall isn't really that bad, although it's
been more of a history lesson than I initially expected. It was clearly
written with the restrictions of older technology in mind. Bringing it
completely up to date with current technology (devfs, gpt, zfs, whatever) is
going to be such an extensive rewrite that it's true, one might as well
write a new installer altogether.

Then again, maybe I just like playing devil's advocate. :)

-- randi


More information about the freebsd-questions mailing list