Re: BootBlocks.

[John Von Essen <john@xxxxxxxxxx>]

The KISS method is something rarely followed these days.

One of the reasons why I liked FreeBSD's sysinstall is that it is 
simple, I rarely use all of its features. Just Custom -> Minimal -> 
Partition -> Commit -> and Reboot!

The install takes about 3 minutes! I can rollout 20 FreeBSD systems in 
just over an hour. Pkg level customization can easily be scripted after 
that via ports. Compare that to a Redhat installer which can take 45 
minutes from start to finish.

I hope the DFly installer remains simple. It just has to setup boot 
loader, allow user to partition disk to their liking, then make a 
filesystem and install core OS. Thats it.

Its up to debate what the "core" OS will include (sendmail vs exim vs 
nothing, will perl be there, and so on). But trying to make an 
installer that will allow you to add/remove pkg's during the initial 
setup is a complete waste of time. Just reboot and do it once the 
system is up via the DFly pkg system - whatever that may be.

In a perfect world, I would love the installer to remain text-based and 
simple. And in an even more perfect world, have something similiar to 
/usr/ports to work from once the system is up.

-John

On Jun 3, 2006, at 2:12 PM, Matthew Dillon wrote:

>     It's irrelevant.  You can hardly expect a small project like ours 
> to
>     cover all the bases.  Those linux installers have large groups of
>     people DEDICATED to just working on the installer.  Short of us 
> dropping
>     everything and putting all our resources for the next year into the
>     installer (at which point I might ask: Why are we bothering if the 
> rest
>     of the system is static because of it).... short of doing that, 
> DragonFly
>     will *NEVER* have an installer that achieves the same level of user
>     friendliness.
>
>     Our current installer is wonderful.  Not because it is user 
> friendly
>     (it isn't, necessarily), but because a small group of people were
>     able to work it up in a few months and because it does precisely 
> what
>     we intended for it to do.  Nothing fancy, nothing spectacular... it
>     just does its job and does it reasonably well.
>
>     For that matter, you have to ask yourself:  Why use DragonFly if 
> you
>     aren't interested in its clustering and filesystem goals?  Well, 
> ok,
>     reliability is important to.  I consider reliability very 
> important, in
>     fact, but if all I wanted was reliability we would still basically 
> just
>     be running FreeBSD-4.x.
>
>     I want clustering.  I want ZFS.  I want OS virtualization 
> (Zen-like, but
>     Dragonfly on top of DragonFly rather then DragonFly on top of 
> Linux).
>     Those are the things I am working towards.
>
> 						-Matt